Total Heat Recovery Heat and Moisture Recovery From Ventilation Air

TOTAL HEAT RECOVERY:
HEAT AND MOISTURE RECOVERY

FROM VENTILATION AIR
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
TOTAL HEAT RECOVERY:
HEAT AND MOISTURE RECOVERY
FROM VENTILATION AIR
LI-ZHI ZHANG
Nova Science Publishers, Inc.

New York
Copyright © 2008 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
NOTICE TO THE READER

The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.
Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.
This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Zhang, Li-Zhi.
Total heat recovery : heat & moisture recovery from ventilation air / Li-Zhi Zhang.
p. cm.
Includes bibliographical references and index.
ISBN 978-1-60876-275-0 (E-Book)
1. Heat exchangers. 2. Ventilation. 3. Heat recovery. 4. Condensation. 5. Moisture. 6.
Humidity--Control. 7. Water harvesting. 8. Buildings--Energy conservation. I. Title.
TH7683.H42Z48 2009
621.402--dc22 2008036864
Published by Nova Science Publishers, Inc. New York

CONTENTS
Preface vii
Chapter 1 Total Heat Recovery in Air-conditioning 1
Chapter 2 Energy Recovery Potentials 9
Chapter 3 Estimation of Sorption and Diffusion Properties of Hygroscopic
Materials 15
Chapter 4 Performance of Energy Wheels 55
Chapter 5 Heat Mass Transfer in Bended Sinusoidal Narrow Ducts 75
Chapter 6 Convective Heat Mass Transfer in Plate-fin Channels 99
Chapter 7 Effectiveness Correlations of Total Heat Exchangers 131
Chapter 8 Numerical Simulation of Total Heat Exchangers 155
Chapter 9 Novel Membranes for Total Heat Exchanger 187
Chapter 10 Heat Mass Transfer in Cross-corrugated Triangular Ducts 229
Chapter 11 Applications of Total Heat Recovery 269
Index 307
PREFACE
Energy and environment are two hot topics world wide today. Energy has been described
as “that which makes things go.” It is seen clearly in the transfer of heat and work inside
environmental systems. Heat occurs due to a temperature difference between the system and
its surroundings, while work makes use of that difference to perform a function. The
depleting nature of primary energy resources, negative environmental impact of fossil fuels
and low exergetic efficiencies obtained in conventional space heating and cooling are the
main incentives for developing alternative heating, ventilating and air-conditioning (HVAC)
techniques which can either save energy or employ low-grade thermal energy sources. Today,
air conditioning has accounted for 1/3 of the total energy use by the whole society. The
percentages are still rising in the fast developing economies like China.
The design of HVAC systems for thermal comfort requires increasing attention,
especially in light of recent regulations and standardization on ventilation, so that an optimal
level of indoor humidity may be reached and maintained to ensure a comfortable and healthy
environment and to avoid condensation damage to the building envelope and furnishings.
Fresh air ventilation is necessary, not only for breathing purposes, but also for the prevention
of deadly epidemic diseases like bird flu and SARs. Energy expenses from ventilation are
very heavy. Ventilation air accounts for 20-40% of the cooling load for HVAC industry. The
ratio can be even higher in hot and humid regions where latent load from fresh air is as heavy
as 50% of the cooling load. To reduce this part of energy is very crucial for the reduction of
energy consumption of the whole HVAC system.
Total heat exchangers (enthalpy exchangers, or the so-called energy recovery ventilators)
could save a large fraction of energy for cooling and dehumidifying the fresh air since cool air
and dryness would be recovered from the exhaust stream to the fresh air in summer. Similarly
in winter, the total heat exchangers could also save energy because exhaust moisture and heat
can be recovered to save heating and humidification energy for fresh air. With total heat
exchangers, the efficiency of the existing HVAC systems can be improved too. When they are
combined with independent air dehumidification units, some alternative cooling technologies
like deep well water cooling, chilled-ceiling, phase-change material cooling storage can be
used in practice. They represent a novel trend in the sustainable development of HVAC
industry.
Social resources spent in energy conservation and environmental protection have been
increased substantially. Scientists and engineers worldwide are in active pursuit of novel
energy recovery technologies. The topic has drawn my attention, too. In fact, my research
viii Li-Zhi Zhang
career has been centered around energy recovery and air dehumidification for more than a
dozen years. From these years’ work, I have the deep feeling that total heat recovery, though
very significant and interesting, is not an easy task.
The first obstacle comes from the materials. The novel total heat recovery technology
requires novel adsorbent materials or highly vapor-permeable membrane materials to fulfill
this task. However, there has been only limited information disclosed from public sources
until now. Most of the novel materials are unavailable commercially and even if they are
available, their cost is a problem. The second obstacle comes from the insufficient
accumulation of heat mass transfer data for the total heat exchangers. Novel materials have
led to new heat mass transfer phenomenon. Simultaneous heat mass transfer is the major
phenomenon for total heat exchangers. However, information on mass transfer in heat mass
exchangers is still scarce. The available information in text books and references is for
sensible heat transfer only. Mass transfer in the total heat exchangers, especially with the
novel materials, has been not considered sufficiently. As a result, system design and
performance analysis of total heat exchangers are difficult, which has prevented their market
penetrations.
To overcome these problems, in these years, I have conducted many fundamental
researches on novel total heat recovery systems, from materials fabrication to heat mass
transfer analysis. Numerical modeling has provided me an efficient tool. These results are
helpful to advance this technology from a notion to a real application. However, these data
are still quite sporadic. It’s not easy to have full access to them. On the other hand, a
systematic introduction on this technology is highly desired, because energy saving in the air
conditioning industry has become a hot issue in these days, with oil prices skyrocketing. This
background has prompted me to write this book. I hope this technology can be systematically
introduced to the public through this book.
In this book, the systems and performances used for total heat recovery are introduced.
Energy wheels and membrane based total heat exchangers are specially described. Heat and
mass transfer modeling of the systems are performed. Influences of key material and design
parameters on the system performance are discussed. Novel membranes including
hydrophobic-hydrophilic composite membrane and composite supported liquid membrane are
developed for total heat exchangers and are characterized. Besides material side
intensification, air side intensification measures are taken as well. Plate-fin and cross
corrugated triangular ducts are two important structures that are introduced. The basic
transport data in these structures are provided. Convective heat and mass transfer coefficients
in plate-fin ducts of finite fin conductance with various cross sections are numerical obtained.
Fluid flow and heat transfer in cross-corrugated triangular ducts are estimated by considering
laminar, transitional, and turbulent complex flow regimes.
Based on the fundamental heat mass transfer data, the book illustrates some examples of
the applications of total heat recovery in novel HVAC systems. Chilled-ceiling combined
with desiccant cooling and independent air dehumidification are two pioneering trends in air
conditioning industry. They overcome the shortcomings of conventional all air systems by
decoupling the treatment of latent load from sensible load. Partial or full total heat recovery
are realized in combination with these novel systems, which contribute to reduced energy use
with increased indoor humidity control, even in transit seasons when traditional air
conditioning systems fail to control humidity. The component modeling of various key
equipments like refrigeration cycle, heat pumps, regenerative wheels, heat exchangers,
Preface ix
cooling coils, are conducted to estimate their energy performance and their effects on indoor
thermal and humidity performance.
The book combines theoretical analysis with engineering practices. It covers a wide range
of knowledge from fundamental heat mass transfer to novel systems design and performance
analysis, from materials synthesis, characterization to heat exchanger thermodynamics and
fluid dynamics. I hope the book may give some insight and design guidelines for the total
heat recovery oriented air conditioning energy conservation.
When writing the book, I am keeping in mind that the fundamentals and methodologies
be given the priority. The goal is not only to tell engineers and students what to do, but how
to do. As a basis, first priority is given to the synthesis and characterization of novel materials
for total heat recovery. Another emphasis is on energy system modeling. To evaluate an
energy system, detailed steps from physical model setting up, mathematical model setting up,
solution method, experimental and numerical validation, and parametric studies, are
illustrated. Generally, it is a cross-discipline endeavor which relies heavily on numerical heat
mass transfer for system modeling. I hope peer engineers and research students could benefit
from the methodologies exhibited in this analysis and extend them to the analysis of other
energy systems.
At last, I would like to thank my family for their long-term support of researches on this
topic. I would also like to thank the Natural Science Foundation of China (NSFC) for their
continuing financial support in my researches of total heat recovery. I am indebted to my
colleagues at South China University of Technology, and others all over China and
throughout the worldwide who have provided suggestions and ideas which, in no small way,
have contributed to the fabric of this text. I have always strived to remain conscious of
student learning needs and difficulties, and I am grateful to my many research students, at
South China University of Technology and elsewhere, who have provided positive
reinforcement for my efforts.
Chapter 1
TOTAL HEAT RECOVERY IN AIR-CONDITIONING
ABSTRACT
In most industrialized countries, energy consumption by HVAC sector accounts for
1/3 of the total energy consumption of the whole society. Cooling and dehumidifying
fresh ventilation air constitutes 20-40% of the total energy load for HVAC in hot and
humid regions. Total heat recovery-heat and moisture recovery from ventilation air has
become a hot topic in these years. In this chapter, the research backgrounds are
introduced. A description is given to the most commonly encountered total heat recovery
equipments: energy wheels, membrane stationary total heat exchangers.
1.1. INTRODUCTION
Modern buildings and their heating, ventilating, and air-conditioning (HVAC) systems
are required to be more energy efficient, while considering the ever-increasing demand for
better indoor air quality, performance and environmental issues. The goal of HVAC design in
buildings is to provide good comfort and air quality for occupants during a wide range of
outdoor conditions. There are many researches aimed at improving the HVAC systems in
buildings while reducing the energy costs and environmental impacts. Some studies
concentrate on control strategies and protocols like VAV (Variable Air Volume), VRV
(Variable Refrigerant Volume) and others focus on the analysis of specific components like
refrigerators, cooling coils, etc. In order for these systems to have the greatest impact, it is
important for the energy needs of the building to be reduced as much as practical.
Energy has been described as “that which makes things go.” It is seen clearly in the
transfer of heat and work inside environmental systems. Heat occurs due to a temperature
difference between the system and its surroundings, while work makes use of that difference
to perform a function. Energy and environment have become the major two issues around the
globe. In most industrialized countries, energy consumption by HVAC sector accounts for 1/3
of the total energy consumption of the whole society. Cooling and dehumidifying fresh
ventilation air constitutes 20-40% of the total energy load for HVAC in hot and humid
regions. The percentage can be even higher where 100% fresh air ventilation is required [1],
2 Li-Zhi Zhang
such as kitchen, hospital, factories. To reduce this part of energy are very crucial for the
reduction of energy consumption of the whole HVAC system.
Air-conditioning in hot and humid environments is an essential requirement for support
of daily human activities. Humidity problems can be found in many applications including
office buildings, supermarkets, art galleries, museums, libraries, electronics manufacturing
facilities, pharmaceutical clean rooms, indoor swimming pools and other commercial
facilities. For thermal comfort reasons, indoor air conditions around 27°C temperature and
10g/kg humidity ratio are the accepted set points. However, the Southern China and other
Southeast Asia countries have a long summer season with a daily average temperature of
30°C, and humidity ratio above 20g/kg. Outdoor relative humidity often exceeds 80%
continuously for a dozen days, leading to mildew growth on wall and furniture surfaces,
which affects people’s life seriously. In spring in Southern China, there is a period named
“Plum raining seasons” when it rains continuously for one to two months. People can not see
sun for a long time and stuff from quilts to grains gets moldy easily. Consequently,
mechanical air dehumidification plays a major role in air conditioning industry in these
regions.
1.2. OPPORTUNITIES AND CHALLENGES

Ventilation air is the major source of moisture load in air conditioning. As shown in
Figure 1.1 for a moisture load estimation of a medium size retail store, ventilation air
constitutes about 68% of the total moisture load in most commercial buildings [2]. As a
consequence, treatment of the latent load from the ventilation air is a difficult and imminent
task for HVAC engineers, especially in hot and humid climates like South China.
Normally the water vapor content of atmospheric air is small, some tens of grams per kilo
of air. Nevertheless, due to the very high heat of vaporization, the latent heat content in air
conditioning is of the same order of the sensible one. Due to the fact that the fresh air latent
load accounts for 20-40% of the total load for air conditioning and air conditioning accounts
for 1/3 of the total energy use in society, to reduce energy use in treating fresh air a crucial
step to the whole society’s sustainable development. Total heat exchangers (or the so-called
energy recovery ventilators, or enthalpy exchangers) could save a large fraction of energy for
cooling and dehumidifying the fresh air since cool and dryness would be recovered from the
exhaust stream to the fresh air in summer. Figure 1.2 shows a schematic of an energy wheel.
Figure 1.3 shows a schematic of a stationary total heat exchanger. It is a cross-flow
membrane based total heat exchanger. With total heat exchangers, the efficiency of the
existing HVAC systems can be improved either. The reason is that normally the fresh air is
dehumidified by cooling coil through condensation followed by a re-heating process, which is
very energy intensive. This part of energy can be saved if total heat exchangers are installed
to reduce the dehumidification load. Besides energy conservation, the total heat exchangers
have the additional benefits of ensuring sufficient fresh air supply, which is crucial for the
prevention of epidemic respiratory diseases like SARS and Bird flu.
Total Heat Recovery in Air-conditioning 3
Domestic loads 0.7

Humid materials 0.3
Wet surfaces 0
Doors 7.7
Infiltration 11.8
Ventilation 59.5
Permeance 0.9
People 8.2
0 10 20 30 40 50 60 70
Loads (kg/h)
Figure 1.1. Sources of moisture loads in a medium size retail store.
Fresh air out

Fresh air in
Exhaust air in
Exhaust air out
Figure 1.2. Schematic of an energy wheel.
Exhaust in
Plates
Fresh out
Fresh in
Duct Sealing
Exhaust out
Figure 1.3. Schematic of a stationary total heat exchanger.
The depleting nature of primary energy resources, negative environmental impact of

fossil fuels and low exergetic efficiencies obtained in conventional space heating and cooling
are the main incentives for developing alternative heating, ventilating and air-conditioning
4 Li-Zhi Zhang
(HVAC) techniques which can either save energy or employ low-grade thermal energy
sources. Novel air conditioning systems with total heat recovery are the directions for
sustainable development of HVAC industry [3-6].
Besides temperature, humidity is another important parameter influencing people’s
feeling of thermal comfort. Figure 1.4 shows the comfort zone in a psychrometric chart [7].
As seen, in summer, the narrow zone between operative temperature 24°C and 27°C,
humidity between 4g/kg and 20°C wet bulb are the acceptable levels of thermal comfort.
People will feel uncomfortable whether it’s too dry or too humid. The design of HVAC
systems for thermal comfort requires increasing attention, especially in the light of recent
regulations and standardization on ventilation [8], so that an optimal level of indoor humidity
may be reached and maintained to ensure a comfortable and healthy environment and to avoid
condensation damage for building envelope and furnishings.
Part load is another problem. In hot and humid climates, air conditioning is an
indispensable component to maintain a comfort indoor environment with lower temperature
and humidity than outdoor conditions. Operating under hot and humid outdoor conditions, air
conditioning has to deal with both sensible and latent loads in a space. In many cases to deal
with space latent cooling load using a small HVAC system is often challenging and difficult.
The air conditioning system’s design load is calculated based on the number of occupants and
their level of activity, types and quantity of equipments used in space, solar irradiation
experienced, heat transmission through the building materials, heat gained from infiltrated
outdoor air and many other factors. In reality, the space loads are always below their design
values. Under part-load conditions, the common practice is to employ control method to
maintain the space temperature while allowing the space humidity to vary. In part load
conditions, supply air temperature is reduced. It is still enough to extract sensible load, but is
insufficient to extract latent load. Because at the rised temperature, air cannot be
dehumidified by dew-point condensation. The indoor humidity is out of control. In full load
seasons from June to September, humidity is controlled very well, but in other transit seasons,
humidity of the space may drift towards a high value that causes human discomfort while
supporting the growth of pathogenic or allergenic organisms. It is also believed that the
emission rate of formaldehyde from furniture and building materials is higher when humidity
rises, resulting in poor indoor air quality.
Stringent ventilation regulations make the humidity problem more serious. In modern
society, people spend most of their time in built environments. More attention has been paid
to indoor air quality and indoor thermal comfort. HVAC systems are necessary for almost all
buildings. However, conventional air conditioning modes, such as constant air volume (CAV)
systems and variable air volume (VAV) systems, face great challenges in effective outdoor air
ventilation and precision indoor air humidity control. From the view points of ventilation, the
main problems with conventional air conditioning systems are analyzed as follows [1,4,5].
(1) The outdoor air in conventional air conditioning systems mixes with the re-circulated
air, which causes transmission of bacterium and virus among multiple zones. In this
situation, occupants are at high risk of infection when diseases breakout, like SARS
and bird flu. Human’s expectation of effective ventilation with 100% outdoor air has
been increasingly rising.
Humidity ratio (g/kg)

0 5 10 15 20
35
18
20
°C
RH
RH
°C
WB
％
WB
50
60
Operative temperature (°C)
30
25 Summer
Winter RH
0 ％
10
20
H
30％R
15
Figure 1.4. Comfort zone in a psychrometric chart.
(2) The indoor relative humidity tends to rise under part load operation because the air
conditioning systems usually control the indoor temperature by reducing their
cooling capacities. To control load, cooling-reheating processes are required, which
are very much energy intensive. This problem is very serious in hot and humid
regions, like Canton. To improve humidity control, the method of decoupling
temperature control and humidity control has attracted much attention. To realize
independent humidity control, an independent fresh air conditioning system, or
known as dedicated outdoor air system, is always required.
(3) New technologies for a more comfortable and energy efficient indoor environment,
such as chilled ceiling/beam, thermal storage and VRV (Variable Refrigerants
Volume), require parallel independent fresh air conditioning systems to meet
demands on effective ventilation and removal of latent load. However, the energy
consumption for dehumidifying fresh air is huge, which often accounts for 20%-40%
of the total energy for air conditioning in hot and humid areas. The unaffordable
energy cost for treatment of fresh air, particularly for fresh air dehumidification
seriously restricts the application of independent fresh air conditioners. Total heat
recovery becomes a necessity.
Conventionally cooling coils are used to cool and dehumidify supply air. It is called the
coupled cooling since cooling and dehumidification are accomplished simultaneously in a
coupled way. To dehumidify air, air temperature must be cooled to below dew point
6 Li-Zhi Zhang
temperature like 10°C. Dehumidified air of such low temperature cannot be supplied to the
space directly because people may feel cold draft under the cold air stream. Reheating has
been widely used in many applications behind a cooling coil to prevent this problem.
However this cooling-reheating process is energy intensive. Energy is needed to overcool the
air across the cooling coil and also to reheat the off-coil air to the desired space humidity and
temperature. Although reheating is able to maintain a space at its design temperature and
humidity during pert-load conditions, it is not often a recommended practice chiefly due to its
high energy use.
To solve this problem, nowadays there is a trend to separate the treatment of sensible load
from latent load. This is the so called independent humidity control [3,4]. According to this
scheme, sensible load is treated by chilled-ceilings, cooling coils, or air handling units which
still cools the supply air but doesn’t necessarily cool it as low as to dew point. Supply air
temperature can be adjusted as sensible load requires. The latent load is accomplished by an
independent dehumidification unit, which is to treat all latent load alone. How to combine
these systems with total heat recovery is a challenging yet interesting work.
The novel total recovery systems can be classified into two categories: energy wheels and
stationary total heat exchangers. In the following chapters, energy wheels and membrane
based total heat exchangers are specially described. Heat and mass transfer modeling of the
system are performed. Influences of key material and design parameters on the system
performance are discussed. Sorption and diffusion of moisture in hygroscopic materials are
the key parameters influencing latent heat recovery capability. Their appraisal methods are
provided and improved. Novel membranes including hydrophobic-hydrophilic composite
membrane and composite supported liquid membrane are developed for total heat exchangers
and are characterized. Besides materials side intensification, air side intensification measures
are taken as well. Plate-fin and cross corrugated triangular ducts are two important structures
that are introduced. Plate-fin is compact and mechanically strong. Cross-corrugated triangular
ducts are a new type of primary surface heat mass exchanger. The basic transport data in
these structures are provided. Convective heat and mass transfer coefficients in plate-fin ducts
of finite fin conductance with various cross sections are numerical obtained. Fluid flow and
heat transfer in cross-corrugated triangular ducts are estimated by considering laminar,
transitional, and turbulent complex flow regimes. Some application examples of total heat
recovery in combination to novel independent air dehumidification units and chilled-ceiling
panels are provided.
1.3. CONCLUSION
Energy expenses in air conditioning has rise to 1/3 of the total energy expenses in the
whole society. Conditioning of ventilation fresh air accounts for 20-40% of the total energy
cost in air-conditioning industry. Stringent ventilation regulations make the problem more
serious. Total heat recovery from ventilation air has become one of the most important faction
of HVAC energy conservation. Present technologies in total heat recovery fall into two
categories: energy wheels and stationary total heat exchangers. The stationary total heat
exchangers uses hygroscopic materials like paper and membranes as the heat and moisture
transfer media. There are full of challenges and opportunities in this area.
REFERENCES
[1] Niu, J.L.; Zhang, L.Z.; Zuo, H.G. Energy savings potential of chilled-ceiling combined
with desiccant cooling in hot and humid climates. Energy and Buildings, 2002, 34,
487-495.
[2] Zhang, L.Z.; Zhu, D.S.; Deng, X.H.; Hua, B. Thermodynamic modeling of a novel air
dehumidification system. Energy and Buildings, 2005, 37, 279-286.
[3] Zhang, L.Z.; Niu, J.L. Performance comparisons of desiccant wheels for air
dehumidification and enthalpy recovery. Applied Thermal Engineering, 2002, 22,
1347-1367.
[4] Zhang, L.Z.; Niu, J.L. A pre-cooling Munters Environmental Control cooling cycle in
combination with chilled-ceiling panels. Energy, 2003, 28, 3, 275-292.
[5] Niu, J.L.; Kooi vd, J.; Ree, H.vd. Energy saving possibilities with cooled-ceiling
systems. Energy and Buildings, 1995, 23, 147-158.
[6] Zhang, L.Z.; Niu, J.L. Indoor humidity behaviors associated with decoupled cooling in
hot and humid climates. Building and Environment, 2003, 38, 99-107.
[7] ASHRAE. 2005 ASHRAE Handbook - Fundamental. Atlanta (GA): American Society
of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (ASHRAE), 2005.
[8] ASHRAE. ANSI/ASHRAE Standard 62-2001, Ventilation for acceptable indoor air
quality. Atlanta (GA): American Society of Heating, Refrigerating and Air-
Conditioning Engineers, Inc. (ASHRAE), 2001.
Chapter 2
ENERGY RECOVERY POTENTIALS
ABSTRACT
Stringent ventilation regulations make the humidity problem more serious. In
modern society, people spend most of their time in built environments. More attention
has been paid to indoor air quality and indoor thermal comfort. Fresh air centered HVAC
systems are necessary for almost all buildings. In this chapter, a quantitative analysis is
provided for the energy expenses and possible savings in fresh air ventilation. More
specifically, hot and humid climates like Canton are selected as the calculating sample.
Under such climates, moisture recovery is more important than sensible recovery.
2.1. INTRODUCTION
Air-conditioning in hot and humid environments is an essential requirement for support
of daily human activities. Humidity problems can be found in many applications including
office buildings, supermarkets, art galleries, museums, libraries, electronics manufacturing
facilities, pharmaceutical clean rooms, indoor swimming pools and other commercial
facilities. For thermal comfort reasons, indoor air conditions around 27°C temperature and
10g/kg humidity ratio are the accepted set points. However, the Southern China and other
Southeast Asia countries have a long summer season with a daily average temperature of
30°C, and humidity ratio above 20g/kg. Outdoor relative humidity often exceeds 80%
continuously for a dozen days, leading to mildew growth on wall and furniture surfaces,
which affects people’s life seriously. In spring in Southern China, there is a period named
“Plum raining seasons” when it rains continuously for one to two months. People can not see
sun for a long time and stuff from quilts to grains gets moldy easily. Consequently,
mechanical air dehumidification plays a major role in air conditioning industry in these
regions.
How much energy that can be recovered? The first impression may be that it’s trivial.
However, the question is meaningless unless we take a quantitative analysis of the energy
uses in air conditioning. Normally the water vapor content of atmospheric air is small, some
10 Li-Zhi Zhang
tens of grams per kilo of air. Nevertheless, due to the very high heat of vaporization, the latent
heat content in air conditioning is of the same order of the sensible one.
2.2. ENERGY CALCULATIONS

Sensible load for fresh air can be calculated by
Qs = c pa (To − Ti ) (2.1)
where cpa is specific heat of air (equal to 1.005 kJkg-1K-1), To is outside temperature (°C), and
Ti is indoor set point temperature (°C).
Latent load for fresh air
QL = Lw (ω o − ω i ) (2.2)
where Lw is latent heat of water evaporation (2501 kJkg-1), ωo is outside humidity ratio
(kg/kg), and ωi is indoor set point humidity ratio (kg/kg).
South China is a typical sub-tropical climate, where the dry bulb temperature is high, but
not as high as 40°C. However, it’s humid almost during the whole year. Table 2.1 lists the
hourly mean outdoor dry bulb temperature and humidity ratio in each month for the city of
Hong Kong. As seen, average relative humidity is above 70% during the whole year. The
calculated sensible load and latent load for fresh air is also listed in the table. A comparison
between the sensible load and latent load is given in Figure 1.1.
The set points for indoor air are: winter, 18°C DB, 0.50 RH, 6.4g/kg HR; Summer, 27°C
DB, 0.50 RH, 10g/kg HR. As seen, sensible load for fresh air in winter (January, February,
and December) is negative, meaning heating in these three months are required. Sensible load
from March until November is positive, meaning cooling is required in this long summer
period. They are for the ventilation air only.
During the whole year, the latent load for fresh air is positive. This indicates that air
dehumidification is required 12 months a year in this region. Even in winter and in transient
seasons like April, when it’s cool outside, but the outdoor air is very humid with relative
humidities above 70%, see Table 1. Therefore air dehumidification is a necessity here. When
it’s in hot and humid months from June to September, the latent load is around 25 kJ/kg,
almost 5.5 times higher than sensible load. Therefore in these regions air dehumidification is
more important than air cooling. Moisture load accounts for 80% of the total load of cooling
and dehumidification. Sensible heat recovery is meaningless if latent heat is not recovered.
Each occupant requires 35m3/h fresh air in an air-conditioned space. Considering an
office of 20m2, the mean sensible cooling load (from various sources like heat gains form
surroundings, sunlight, computers, human body heat dissipation, etc) for the office building is
100W/m2. The total sensible load is 2.0kW. If there are 5 people in the office, in August, the
sensible load for fresh air is 0.23kW and the latent load for fresh air is 1.53kW. As a result, of
the total air conditioning load of 2.0+0.23+1.53=3.76kW, the sensible load for fresh air only
accounts for 6%, while the latent load for fresh air accounts for 41%. That’s very impressive.
Energy Recovery Potentials 11
The sensible-only heat exchangers like heat pipes, run-around heat exchangers and
regenerative heat exchangers [2-4] have little use now.
Table 2.1. Hourly mean sensible and latent load for fresh ventilation air
DB HR Qs QL
Month RH
(°C) (g/kg) (kJ/kg) (kJ/kg)
1 (Jan) 14.6 8.3 0.814 -3.417 4.764
2 (Feb) 15.1 7.9 0.750 -2.915 3.764
3 (Mar) 20.3 11.6 0.794 2.311 13.017
4 (April) 20.9 13.7 0.904 2.915 18.269
5 (May) 25.6 17.1 0.849 0.603 18.078
6 (Jun) 28.1 19.9 0.853 3.116 25.081
7 (July) 29.3 20.3 0.812 4.321 26.081
8 (Aug) 28.9 20.4 0.835 3.920 26.331
9 (Sep) 28.1 19.9 0.853 3.116 25.081
10 (Oct) 25.8 15.3 0.751 0.804 13.576
11 (Nov) 21.1 11.2 0.730 3.116 12.017
12 (Dec) 16.2 9.1 0.806 -1.809 6.765
Notes: DB, dry bulb temperature; HR, humidity ratio; RH, relative humidity; Qs, sensible load, minus
for heating, positive for cooling; QL, latent load, minus for humidification, positive for
dehumidification. Set points for indoor air: winter, 18°C DB, 0.50 RH, 6.4g/kg HR; Summer, 25°C
DB, 0.50 RH, 10g/kg HR.
30
Qs
25
QL
20
Load (kJ/kg)
15
10
0
1 2 3 4 5 6 7 8 9 10 11 12
-5
Month
Figure 2.1. Comparison of sensible load and latent load for fresh air on each month.
12 Li-Zhi Zhang
In summary, generally, fresh air moisture load accounts for 40% of the total load for air
conditioning in hot and humid regions. Considering energy consumed by air conditioning
industry has accounts for 1/3 of the total energy use by the whole society, energy
conservation from ventilation fresh air is very significant and effective. Yet how to save this
part of energy is an interesting and difficult task.
The general idea to save sensible and latent load from ventilation air is to use a total heat
exchanger. This device is also called as the enthalpy exchanger, or the energy recovery
ventilator [5,6]. Figure 2 shows a schematic of a stationary total heat exchanger. As seen, the
device is like a parallel-plates air-to-air heat exchanger. However, in place of common metal
foils, some new materials with vapor-permeable capabilities are used as the plates. Therefore
both the sensible heat and the latent heat (moisture) can be exchanged between two air flows.
Due to the sensible heat and moisture exchange, heat and humidity would be recovered from
the exhaust stream in winter (especially in cold climates like in Beijing) and excess heat and
moisture would be transferred to the exhaust in order to cool and dehumidify the incoming
fresh air in summer.
Exhaust in
Plates
Fresh out
Fresh in
Duct Sealing
Exhaust out
Figure 2.2. Schematic of a cross-flow parallel-plates total heat exchanger.
Return air
Total Heat Exchanger
Exhaust air
Supply air
Fresh air
Sensors Cooling coils Heater Humidifier
Figure 2.3. An air handling unit (AHU) with total heat exchanger.
Energy Recovery Potentials 13
The total heat exchanger can be used as an independent stand alone ventilator for a room.
In such cases, it is often used in combination with a VRV (variable refrigerant volume)
refrigeration system, where the cooling coils are used to treat the sensible load. For other
traditional all-air central air conditioning systems, it can be combined to the existing air
handling unit (AHU), as shown in Figure 3, to save fresh air load.
2.3. CONCLUSION
A sample calculation indicates that in hot and humid regions air dehumidification is
required 12 months a year. When it’s in hot and humid months from June to September, the
latent load is around 25 kJ/kg. It is almost 5.5 times higher than sensible load. Therefore in
these regions air dehumidification is more important than air cooling. Moisture load accounts
for 80% of the total load of cooling and dehumidification. Sensible heat recovery is
meaningless if latent heat is not recovered. Energy conservation from ventilation fresh air is
very significant and effective.
REFERENCES
[1] Zhang, L.Z.; Niu J.L. Energy requirements for conditioning fresh air and the long-term
savings with a membrane-based energy recovery ventilator in Hong Kong. Energy,
2001, 26, 119-135.
[2] Dhital, P.; Besant, R. W.; Schoenau, G. J. Integrating run-around heat exchanger
systems into the design of large office buildings. ASHRAE Trans., 1995, 101, 979-991.
[3] Johnson, A.B.; Besant, R.W.; Schoenau, G. J. Design of multi-coil run-around heat
exchanger systems for ventilation air heating and cooling. ASHRAE Trans., 1995, 101,
967-978.
[4] Manz, H.; Huber, H.; Schalin, A.; Weber, A.; Ferrazzini, M.; Studer, M. Performance
of single room ventilation units with recuperative or regenerative heat recovery. Energy
and Buildings, 2000, 31, 37-47.
[5] Kistler, K.R.; Cussler, E.L. Membrane modules for building ventilation. Chemical
Engineering Research & Design, 2002, 80, 53-64.
[6] Zhang, L.Z.; Jiang, Y. Heat and mass transfer in a membrane-based energy recovery
ventilator. J. Membrane Sci., 1999, 163, 29-38.
Chapter 3
ESTIMATION OF SORPTION AND DIFFUSION

PROPERTIES OF HYGROSCOPIC MATERIALS
ABSTRACT
Sorption and diffusion properties of moisture in hygroscopic materials are the basic
properties for moisture recovery. This chapter introduces some methodologies for the
correct and convenient estimation of moisture sorption and diffusion in novel materials.
The first method is a simple one: sorption and diffusion of moisture in a thermo-
hygrostat. The second method is relatively complicated, however it’s more accurate. The
method uses an emission cell to measure moisture permeation through membranes. It
simultaneously considers convective mass transfer resistance on membrane surfaces.
NOMENCLATURE
At transfer area (m2)
C shape factor for the isotherm
Dva vapor diffusivity in air (m2/s)
Dvm moisture diffusivity in material (m2/s)
Ev local emission rate (kgm-2s-1)
Hd duct height of air stream (m)
K total moisture transfer coefficient (kgm-2s-1)
k convective mass transfer coefficient (m/s)
kp partition coefficient (kg air/kg membrane)
N air exchange rate (s-1)
NTU Number of Transfer Units
p Pressure (Pa)
r radius coordinate (m)
r0 cell radius (m)
rm extent of drying
RH relative humidity
Re Reynolds number
16 Li-Zhi Zhang
Sc Schmidt number
Sh Sherwood number
T temperature (K)
t time (s)
u velocity (m/s)
um air bulk velocity (m/s)
V volume of the cell (m3)
Va volumetric air flow rate (m3/s)
w moisture uptake in material (kg moisture/kg dry material)
W total weight of membrane including moisture (kg)
Wmax maximum water uptake of membrane material (kg/kg)
z coordinates in thickness (m)
z0 Half the desiccant sheet thickness (m)
Greek letters
ν kinematic viscosity of air (m2/s)

δ air slit height (m)
ω humidity ratio (kg moisture/kg air)
φ angle
ε moisture exchange effectiveness
θ dimensionless humidity ratio
ρ density (kg/m3)
Superscripts
* dimensionless
Subscripts
i inlet
L Lower chamber
o outlet
s surface
v vapor
m material
3.1. INTRODUCTION
Treatment of moisture is the fundamental aspect of total heat recovery. Hygroscopic
materials provide the media for moisture absorption and removal. In a desiccant energy wheel
system, the desiccant wheel rotates between the outside fresh air and the exhaust air from
room. Heat and humidity would be recovered from the exhaust in winter and excess heat and
moisture would be transferred to the exhaust to cool and dehumidify the process air in the
Estimation of Sorption and Diffusion Properties … 17
summer. In a membrane total heat exchanger system, moisture is adsorbed by membrane

surface in fresh air side. The adsorbed moisture then permeates to the exhaust side, and
desorbs from the membrane surface in exhaust side. Energy wheel and membranes use
desiccant materials. They can treat moisture because they are hygroscopic. For both
applications, the sorption and diffusion characteristics in the desiccant material are the basic
data for system performance. This chapter introduces the basic knowledge to measure and
estimate sorption and diffusion properties of desiccant materials. They are the fundamentals
for heat and moisture transfer in these systems, which provide the basics for system
performance analysis and optimization.
3.2. SORPTION AND DIFFUSIVITY IN A THERMO-HYGROSTAT

The measurement of sorption properties of desiccant materials is well-known, but the
simultaneous measurement of sorption and diffusion properties of desiccant materials can not
be easily found from the published literature. This may be due to the fact that it is only
recently that the mass transfers inside the desiccant are considered with most of the previous
studies confined to heat transfers. To better estimate the wheel and membrane performance, a
method for the simultaneous measurement of sorption and diffusion properties for desiccant
sheets, with which a new polymer material is measured, is proposed in this section.
Sorption Experiments
HUTC-MEM02, a novel hydrophilic desiccant material sheet is considered for air

dehumidification since previous studies have already found that the diffusivity of this
desiccant sheet is very high. Equilibrium sorption measurements of water vapor in the
desiccant sheet material are performed at various temperatures and humidities in a thermo-
hygrostat. An automatic thermo-hygrostat (WS-97) is designed and constructed for sorption
experiments in air-conditioning applications. Its photographic view is shown in Figure 3.1.
The dimension of the working chamber in the thermo-hygrostat is 500mm×500mm×800mm.
The temperatures in the chamber are measured with platinum resistance sensors and the
humidities are measured with chilled-mirror dew point meters. The operating parameters in
the working chamber can be varied in the following ranges: Humidity, 30-90%RH;
Temperature, 20-50°C (the upper limit is controlled by the characteristics of humidity
sensors, while the lower limit is determined by the temperature of cooling bath water). The
non-uniformity in the chamber is: Temperature, ≤±0.1°C; Humidity, ≤±2%RH. The precision
of measurement is: Temperature, 0.1°C; Humidity, 2%RH. The precision of the control is:
Temperature, ≤±0.1°C; Humidity, ≤±2%RH. The air velocity in the chamber is less than
0.2m/s.
The temperature control system of the thermo-hygrostat consists of temperature-constant
water tank, water circulation pumps, electric heater, cooling coils, heat exchanger for the
chamber, heat exchanger for the outer cavity, and electric-magnetic valves. Before the
experiment, the temperature of the water tank is first heated by the electric heater. At the
same time, driven by the circulation pumps, hot water in the tank circulates through the two
18 Li-Zhi Zhang
heat exchangers, to heat the chamber and the outer cavity to a set value. When the chamber
temperature is greater than the prescribed value, the hot water valve closes and the hot water
circulation comes to a halt. When the chamber temperature is smaller than the prescribed
value, the cool-water valve opens and the cooling water begins to circulate in the cooling coil,
driving the chamber temperature down to its set point. Thus the temperature in the chamber
can be kept nearly constant throughout the test by such repeated heating and cooling
processes.
The humidity control system consists of humidification water tank, electric heater,
cooling coil placed in the humidification water tank, humidification cylinder, fans and
dehumidification column. When the humidity goes below the set value, the humidification fan
begins to work, which drives ambient air to flow through the humidification water tank,
where it absorbs water and is humidified. Then the moist air is transported to the working
chamber, after it is mixed with the circulation air from the chamber. The extent of
humidification is adjusted by raising or decreasing the temperature of humidification water
tank. In some cases such as when the environment is very dry, it is very difficult to obtain a
high humidity only by the above humidification method. At these times, an auxiliary
ultrasonic humidifying system will be applied. When the humidity reaches the set point,
humidification process stops.
On the other hand, when the humidity in the thermo-hygrostat is greater than the set
value, dehumidification process begins. The circulation air from the chamber is first directed
to flow through a silica gel column, where it is dehumidified, before it is returned to the
working chamber. The humidity in the thermo-hygrostat is kept nearly constant by these
humidification and dehumidification processes. All these temperature and humidity control
processes are performed by a microcomputer.
Before the experiments, the temperature in the working chamber is first set to a
prescribed value. Then the humidity in the chamber is set to a start value around 20%. This
process usually needs 2.5 to 3.5 hours. During this process, the desiccant sheet sample (4.2g,
0.52mm thick) is dried in a hot-wind drying chamber of 80°C until the weight of the desiccant
sheet becomes unchanging. After above preparations, the desiccant sheet is placed in the
working chamber and hooked to a strain gauge whose other end is connected to the chamber
roof. The increase of the weight of the desiccant sheet is detected by the resistance changes of
the strain gauge. The variations of the resistance are then recorded by a computer with the
help of an additional circuit that converts resistance to voltage. When the difference of
desiccant sheet weight is less than 0.1mg for every 0.5 hour, it is assumed that the equilibrium
has been reached for the given state, and one sorption experiment is finished. Then the
humidity of the chamber is increased to a new value while the temperature is kept unchanged.
Another sorption test is conducted at this set temperature and the above procedures are
repeated. Altogether more than five tests are needed for one sorption isotherm at a given
temperature.
Figure 3.1. Photographic view of the thermo-hygrostat.
Drying Experiments
Saturated desiccant sheet samples are regenerated in a hot wind drying box manufactured
by Shanghai Instruments Corporation. The temperature in the box can be as high as 300°C.
The wind speed in the chamber measured is 0.25m/s. The heating power of the drying box is
3.3KW. The fluctuation of temperature at a set point is ±1°C.
Before the regeneration process, the desiccant sheet sample is placed in a chamber of
100% Relative Humidity for 24 hours to reach saturation. Then it is moved into the drying
box that has been pre-set to a given drying temperature. Since most of the water in desiccant
sheet can be air dried after 25 minutes, the weight of the desiccant sheet is measured every 5
minutes during the initial 25 minutes, and every 15 minutes during the rest of time
respectively, by an electronic scale to minimize errors. The Relative Humidity of the
environment during the test is 45%.
Calculation of Diffusivity
The activated diffusion process of water vapor in desiccant sheet follows three stages.
First, water vapor is adsorbed on the sheet surface at the side of higher vapor partial pressure;
second, adsorbed water diffuses through the desiccant sheet, driven by a concentration or
activity gradient; and third, water desorbs from the other side of the sheet [1]. Usually, Fick’s
law applies to the diffusion of the vapor or gas flow in desiccant sheet, which provides the
basis for the method of slops [2] in diffusivity calculations. Assuming a constant density in
dry desiccant sheet, the unsteady Fick’s equation [3] can be expressed as
∂w ∂ ∂w
= [ Dvm ] (3.1)
∂t ∂z ∂z
20 Li-Zhi Zhang
where z is the coordinate in thickness direction (m), and the zero point locates at half of the
desiccant sheet thickness. The initial and boundary conditions for this equation are
t=0 0<z<z0 w = win (3.2)
t > 0 z= 0 ∂w / ∂z = 0 (3.3)
t > 0 z= z0 w = we (3.4)
where win is the initial water concentration (kg moisture/kg material), we is the equilibrium
concentration, and z0 is half the desiccant sheet thickness.
The solution of the Eqs.(3.1)-(3.4) in the case of a constant diffusivity is given by
Crank[3] as
w − we 8 ∞
1 ⎡ (2n + 1) 2 π 2 Dvm t ⎤
Ω= = 2
win − we π
∑
n = 0 ( 2n + 1)
2
exp ⎢−
4 z 02
⎥ (3.5)
⎣ ⎦
where Ω is the dimensionless moisture uptake, and w is the mean value of the moisture
2
uptake of the desiccant sheet. In the range of the Fourier number of Fo(= Dvm t / z 0 )>0.3, the
first term (n=0) of Eq.(3.5) is far greater than the other terms. Then the above equation can be
simplified as
w − we 8 ⎡ π 2 Dvm t ⎤
Ω= = 2 exp ⎢− ⎥ (3.6)
win − we π ⎣ 4 z 02 ⎦
To apply the method of slops, the experimental drying/sorption curves (lnΩ versus t) are
2
compared to the theoretical diffusion curves (lnΩ versus Fo= Dvm t / z 0 ) for the desiccant
sheet tested. The slops of the experimental drying/sorption curves (dΩ/dt) and the theoretic
curves (dΩ/dFo) are estimated at a given moisture uptake, using numerical or graphical
differentiation. The effective moisture diffusivity at a given uptake Ω is calculated from the
equation,
Dvm = [(d (ln Ω) / dt ) exp /(d (ln Ω) / dFo) th ]z 02 (3.7)
Uncertainty Analysis
The accuracy of the calculated diffusivity could not be discussed without a through
understanding of the different contributions of the accuracy of water uptake, thickness, and
measuring time, since the diffusivity is calculated from these three parameters. It should be
known that the accuracy of diffusivity also relates to the number of terms in the series of
Eq.(3.5). However, when Ω>0.5, the truncation errors from Eq.(3.5) to (3.6) are within
0.1%[3], so they can be neglected in error analysis. With a logarithmic transformation on both
sides of Eq.(3.6), we have
⎛ π 2Ω ⎞ π 2 Dvm t
ln⎜⎜ ⎟⎟ = − (3.8)
⎝ 8 ⎠ 4 z 02
Another logarithmic transformation of the above equation gives
⎛ π 2Ω ⎞ ⎛π 2 ⎞
⎜
ln⎜ − ln ⎟
⎟ = ln Dvm + ln t − 2 ln z 0 + ln⎜⎜ 4 ⎟⎟ (3.9)
⎝ 8 ⎠ ⎝ ⎠
A differentiation of Eq. (3.9) suggests
π 2Ω
d ln
8 = dDvm + dt − 2 dz 0 (3.10)
π Ω2
Dvm t z0
ln
8
Thus the errors in the calculations of moisture diffusivity in desiccant sheet by the
method of slops can be estimated as
dDvm dt dz 1 dΩ
= +2 0 + (3.11)
Dvm t z0 ln(π Ω / 8) Ω
2
This equation clearly discloses the different contributions of the measurement errors of
time, desiccant sheet thickness, and water uptake to diffusivity.
Sorption Isotherms
The measured isotherms of the HUTC-MEM02 (a novel hydrophilic desiccant sheet

material) and water vapor are shown in Figure 3.2. They are typical III class adsorption
isotherms. This indicates stronger sensitivities of desiccant sheet to relative humidity at
higher humidities. Actually, the water uptake is an exponential function of relative humidity,
as shown in Figure 3.3. This phenomenon is attributed to the stronger interaction between
water molecules and adsorbed layers in the voids of desiccant sheet with an increase in water
uptake, which increases the difficulties of applying either Langmuir or Dual-sorption model
to the analysis of such isotherms. Nevertheless, an empirical correlation for the isotherms of
desiccant sheet studied is obtained with the help of least square fit of the experimental data.
22 Li-Zhi Zhang
0.3
0.25
Water uptake (kg/kg)

0.2
0.15
t=23.5C
0.1
t=30.1C
0.05
t=38.2C
0
0 1000 2000 3000 4000 5000 6000 7000
Vapor partial pressure (Pa)
Figure 3.2. Isotherms of HUTC-MEM02 and water vapor. The solid line: calculated; discrete dots:
experimental.
w = c1 RH + c 2 RH 2 (3.12)
p
RH = (3.13)
ps
ln ps = a + b / T (3.14)
where ps is the saturation pressure of vapor at temperature T. The values in the equation are:
c1=1.07908E-01; c2=1.50516E-01; a=20.5896; b=-5098.26.
0.25
0.2
Water content(kg/kg)
0.15
0.1
0.05 Experimental
Model
0
0 0.2 0.4 0.6 0.8 1
Relative Humidity
Figure 3.3. Water uptake in relation to relative humidity of air.

0.2
0.15
w (kg/kg)
0.1
0.05 23.5℃,75.5%RH
30.1℃,78.0%RH
Calculated
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (hr)
Figure 3.4. Sorption curves of the desiccant sheet at two conditions.
Sorption Curves
The sorption curves are shown in Figure 3.4. It can be seen that the moisture uptake
increases with the lapse of time. At the beginning, the sorption rates are very fast. After 2-2.5
hours, the water adsorbed amounts to more than 93% of the total moisture that could be
adsorbed by the desiccant sheet when it reaches equilibrium with vapor. During the rest of the
time, the sorption processes go very slowly and the desiccant sheet adsorbs only a small
quantity of water.
The sorption curves under a specified temperature and vapor pressure can be expressed
by the following equation
w = w0 + A1 (1 − e −t / a1 ) + A2 (1 − e −t / a2 ) (3.15)
The constants in the equation can be calculated by the technique of least square fit of
experimental data. The values for the two conditions in Figure 4 are listed in Table 3.1.
Table 3.1. Values in the model of uptake curves
Conditions w0 A1 a1 A2 a2 Uncertainty
(%)
(23.5°C, 5.49E-3 7.64E-2 0.55 7.94E-2 0.55 0.62
75.5%RH)
(30.1°C, 5.54E-3 7.91E-2 0.73 7.91E-2 0.73 0.43
78.0%RH)
24 Li-Zhi Zhang
0.25
55℃
0.2 60℃
70℃
80℃
0.15
w (kg/kg)
0.1
0.05
0
0 0.5 1 1.5 2 2.5
Time (hr)
Figure 3.5. Regenerating curves for HUTC-MEM02 and vapor.
Regeneration Curves
The drying temperature has a considerable influence on the time for the desiccant sheet to
be regenerated, which increases by 2 times when the drying temperature changes from 80°C
to 55°C. This can be clearly seen in Figure 3.5. Most of the water in desiccant sheet can be
dried after 20 minutes’ drying.
A model is summarized to express the regeneration curves of desiccant sheet as
w = ae t +c
(3.16)
The constants in the equation are listed in Table 3.2.
Table 3.2. The constants in the model of regeneration curves
Temperature a b c Uncertainty (%)

(C°)
55 6.00E-3 1.79 0.48 7.09
60 7.67E-3 1.26 0.37 3.80
70 5.48E-3 1.26 0.33 5.54
80 4.44E-3 1.10 0.28 9.09
Table 3.3. Diffusivity and errors calculated from drying curves at two temperatures
Time 5min 10min 20min 1hr 1.5hr

T=60°C
Moisture uptake (kg/kg) 0.13 0.079 0.044 0.021 0.016
Diffusivity (10-8m2/s) 3.98 3.11 2.05 0.67 0.38
Uncertainty 3.32% 7.93% 9.64% 8.88% 6.47%
T=80°C
Moisture uptake (kg/kg) 0.093 0.051 0.028 0.0094 0.0093
Diffusivity (10-8m2/s) 4.89 3.22 1.70 0.39 0.26
Uncertainty 3.21% 7.94% 8.86% 8.37% 7.46%
Moisture Diffusivity
The diffusivity of moisture in desiccant membrane is calculated with Eq.(3.7) as

previously deduced. In Table 3.3 are shown the results from two drying curves. The
uncertainties of the calculated diffusivity can be estimated from Eq.(3.11), as listed in Table
3.3. It is clear that the obtained diffusivity is in the order of 10-8m2/s, and the maximum
uncertainty is less than 10%.
To discuss the diffusivity more clearly, the results from four different drying curves are
plotted in Figure 3.6. The discrete dots in the figure represent the diffusivity calculated at the
corresponding water uptake and temperatures. It is seen that the moisture diffusivity increases
with an increase in water uptake in desiccant sheet, or an increase of sorption temperature.
Furthermore, when the uptake is below 0.07kg/kg, these dots are very densely plotted, which
means that the influence of temperature on diffusivity is negligible. However, as the uptake
increases above 0.07kg/kg, the effects of sorption temperature on diffusivity become larger.
The higher the water concentration in desiccant sheet, the greater the discrepancies of
diffusivity resulted from the sorption temperatures. On the other hand, when the uptake is less
than 0.05kg/kg, the diffusivity increases almost linearly with increasing uptake, while at
larger moisture concentrations, diffusivity becomes stable with variations of uptake. This
character indicates that a constant diffusivity can be assumed for isothermal moisture transfer
through a desiccant sheet, since in most cases, the water content in a hydrophilic desiccant
sheet is bigger than 0.05kg/kg at normal temperatures.
The diffusivity can also be obtained from sorption curves. They are similar to those
calculated from drying curves (for instance, at T=23.1°C, w=0.15kg/kg, Dvm=2.55×10-8m2/s
and at T=30.5°C, w=0.16kg/kg, Dvm=3.17×10-8m2/s). The diffusivity of moisture in other
materials is usually far lower than the values of this study (10-10 m2/s for polymer gel [4]; 10-
12
-10-13 m2/s for methylcellulose desiccant sheet [5]; and 10-12 m2/s for Poly Vinylchloride
sheet [1], to name but a few). It is no wonder that HUTC-MEM02 has very high performance
in air dehumidification.
26 Li-Zhi Zhang
9
55℃
8
60℃
7
70℃
6 80℃
DvmD (10-8m2 /s)
5
4
3
2
1
0
0 0.05 0.1 0.15 0.2 0.25
w (kg H2 O/kg dry membrane)
Figure 3.6. Calculated diffusivity of moisture in desiccant membrane.
A method of directly measuring the sorption, drying, and diffusion characteristics of a

desiccant sheet is offered. An equation for error analysis of diffusivity is also presented. The
sorption isotherms, sorption curves, drying curves, and diffusivity variations of a novel
hydrophilic polymer desiccant sheet used in air dehumidification are obtained by
experiments. The constants in the mathematical models are discussed.
The obtained isotherms are typical III class adsorption isotherms. The sorption curves
indicate that during the first 2-2.5 hours, the desiccant sheet adsorbs most of the water that
can be adsorbed at equilibrium. The drying temperature has a major effect on regeneration
time. Results also show that the moisture diffusivity increases with either an increase in water
uptake in desiccant sheet, or an increase in sorption temperature. At low water concentrations
in desiccant sheet, the diffusivity increases almost linearly with increasing uptake, but at
larger moisture concentrations, diffusivity tends to be stable. Therefore a constant diffusivity
can be assumed for most of the hydrophilic desiccant sheet in air dehumidification.
3.3. FLUID FLOW AND MASS TRANSFER IN A NOVEL EMISSION CELL

Measurement of mass transfer in a thermo-hygrostat is rather complicated and expensive.
To ease the job, in this section, a novel emission cell is proposed. It is very similar to a FLEC
cell, but air flow velocities are one order higher than those through a FLEC cell. It is used to
measure moisture diffusivity through hygroscopic material sheets.
FELC (Field and Laboratory Emission Cell) cell was in recent years used to measure
VOCs emission from a material surface. It is portable and user-friendly, thus it has become a
standard for emission testing in Europe [6-8]. However, our previous studies [7-9] have found
that air velocities through a FLEC cell are too small. Air is easily saturated with emitted gas
as soon as it enters into the cell, which makes it difficult to measure the emissions on the
whole surface. It is not appropriate to use it in the measurement of membrane permeability
directly. Therefore recently, we have modified it to fit the needs in diffusivity measurements.
Though cell structures are similar, the air velocities through the cell are increased by one
order. A methodology similar to previous studies is used to model the mass transfer in the
units.
A complete emission model should include two mechanisms: convective mass transfer on
surfaces, and diffusion in solids. As an essential part of this process, in this section the
convective mass transfer coefficients in the cell are estimated. In next section, the whole
emission model will be set up.
The flow geometry in the cell is shown in Figure 3.7. It is composed of two parts: cap
(Figure 3.8a) and lower chamber (Figure 3.8b). When testing, the planar specimen of the
emission material is placed in the lower chamber and becomes an integral part of the emission
cell. The upper surface of the specimen (the emission surface) and the inner surface of the cell
cap form a cone-shaped cavity. The air is supplied through the air slits in the cap. It is
introduced through two diametrically positioned inlets (symmetrically placed) into a circular-
shaped channel at the perimeter, from where the air is distributed over the emission surface
through the circular air slit. The air flows radially inward, until it exits the cell outlet in the
center.
In addition to emission experiments, this chapter uses the cell to measure moisture
diffusivity in desiccant plate materials. As a first step, forced convection mass transfer and
fluid flow in this cavity is of great interest.
Specimen
Figure 3.7. A schematic showing the flow geometry of the cell.

28 Li-Zhi Zhang
(a)
Emission material
Lower chamber
(b)
Figure 3.8. A view of the cell, showing the cap (a) and lower cavity (b).
Mathematical Models
CFD simulation is employed to calculate the convective mass transfer coefficients

between the fluid flow and emission surface. For the present situation, the flow is assumed to
be laminar and steady. Considering the fluid properties to be constant, the hydrodynamic and
mass transfer problem can be described by Navier-Stokes equations in cylindrical coordinates
as [9]
Conservation of mass
1 ∂ * * ∂u *z 1 ∂uφ
*
(
r ur + * + * ) =0 (3.17)
r * ∂r * ∂z r ∂φ
where r, z and φ are radial, axial and angle coordinates, respectively; ur, uz, and uφ are
velocities in r, z and φ directions (m/s), respectively; superscript “*” in this and the following
equations represents dimensionless form.
Conservation of r-Momentum
∂u r* * ∂u r
*
uφ* ∂u r*
u *
+ uz * + * =
∂r * ∂z r ∂φ
r
(3.18)
∂p * ⎡ 1 ∂ ⎛ * ∂u r* ⎞ ∂ 2 u r* ∂ 2 u r* u r* ⎤
− * + ⎢ * * ⎜r⎜ * ⎟
⎟ + + 2 − 2⎥
∂r ⎣⎢ r ∂r ⎝ ∂r ⎠ ∂z
*2
r * ∂φ 2 r * ⎦⎥
where p represents pressure (pa).

Conservation of z-Momentum
∂u *z * ∂u z
*
uφ* ∂u *z ∂p * ⎡ 1 ∂ ⎛ * ∂u *z ⎞ ∂ 2 u *z ∂ 2 u *z ⎤
u *
+ uz * + * = − * + ⎢ * * ⎜⎜ r * ⎟
⎟+ + 2 ⎥
∂r * ∂z r ∂φ ∂z ⎣⎢ r ∂r ⎝ ∂r ⎠ ∂z
r
*2
r * ∂φ 2 ⎦⎥
(3.19)
Conservation of φ-Momentum
∂uφ* ∂uφ* uφ* ∂uφ* 1 ∂p * ⎡ 1 ∂ ⎛⎜ * ∂uφ ⎞ ∂ 2 uφ* ∂ 2 uφ* ⎤

*
+u + =− * +⎢ ⎟+ +
⎟ ∂z * 2 r * 2 ∂φ 2 ⎥⎥
* *
u r
∂r * ∂z * ∂φ r ∂φ ⎢⎣ r * ∂r * ⎜⎝ ∂r *
r z
r* ⎠ ⎦
(3.20)
Conservation of water vapor
∂θ * ∂θ
uφ* ∂θ 1 ⎡ 1 ∂ ⎛ * ∂θ ⎞ ∂ 2θ 1 ∂ 2θ ⎤
u *
+ uz * + * = ⎢ ⎜r ⎟+ + ⎥ (3.21)
∂r * ∂z r ∂φ Sc ⎣ r * ∂r * ⎝ ∂r * ⎠ ∂z * 2 r * 2 ∂φ 2 ⎦
r
where θ is the dimensionless humidity ratio. The characteristic distance is selected as two
times the spacing between the emission surface and the cap at the cell perimeter. The mean
velocity at the air slit is selected as the characteristic velocity. The dimensionless forms for
the variables are expressed as
30 Li-Zhi Zhang
r
r* = (3.22)
2δ
z
z* = (3.23)
2δ
2uδ
u* = (3.24)
ν
2vδ
v* = (3.25)
ν
2 wδ
w* = (3.26)
ν
where δ is the height of space between the emission surface and the cell cap at air slit (m).
Contrary to FLEC where the air slit pitch is 1mm, in this system, it is 2mm; ν is the kinematic
viscosity (m2/s).
The dimensionless pressure is defined as
4 pδ 2
p* = (3.27)
ρν 2
where ρ is density (kg/m3). The dimensionless humidity ratio is given as
ω − ωs
θ= (3.28)
ωi − ωs
where ω is the air humidity ratio (kg vapor/ kg air); ωi represents humidity at the air slit, and
ωs represents humidity at the emission surface.
The Schmidt number is
ν
Sc = (3.29)
Dva
where Dva is vapor diffusivity in the air mixture (m2/s).

The Reynolds number used to characterize the airflow rate is given by
2u mδ
Re = (3.30)
ν
V
where um is the mean air velocity at the slit, and it is calculated by um = (3.31)
2πr0δ
where V is the volumetric air flow rate to the cell (m3/s); r0 is the maximum radius of the
emission surface, where air is distributed from the slit (m). The Reynolds numbers for the
flow in the cell chamber are very small, say, Re=42 when V=5L/min. Since Re<<2300, the
flow is thought to be laminar.
The Sherwood number is
2kδ
Sh = (3.32)
Dva
where k is the convective mass transfer coefficient (m/s).

Now, considering a control volume in the radial direction, the mass balance has
kAt (ω s − ω b ) = −u m Ac dω b (3.33)
where ωb is the bulk humidity ratio on a cross-section at radius r.

For reasons of symmetry, only half of the cell geometry is taken into account, the mass
transfer area is
At = πrdr (3.34)
The cross-sectional area is
Ac = πr0δ (3.35)
Substituting Eqs.(3.34) and (3.35) into (3.33), the local mass transfer coefficient is
u m r0δ dω b
k=− (3.36)
r (ω s − ω b ) dr
The local Sherwood number is
r0* ReSc dθ b
ShL = (3.37)
2θ b r * dr *
32 Li-Zhi Zhang
by considering
u m* = Re (3.38)
Similarly, the mean Sherwood number is
r0* ReSc
Shm = − ln θ b (3.39)
*2 *2
r0 −r
where the bulk dimensionless humidity ratio is
∫ (u θ )dA
*
θb = (3.40)
u m*
The inlet and boundary conditions for mass transfer are
r*=r0*: θ=1; (3.41)
z*=0: θ=0 (3.42)
where Eq.(3.42) indicates that the boundary condition on the emission surface is a uniform
concentration condition. Other boundaries are adiabatic surfaces, have no mass transfer, and
are expressed as
∂θ
=0 (3.43)
∂n
where n is the normal direction.
Discretisation and Solution Strategy
As previous studies, for reasons of symmetry, only half of the cell is selected as the
modeling domain. The Navier-Stokes equations are solved in three-dimensional cylindrical
coordinates. The round channel, which has a rectangular cross-section, the air slit, and the air
inlet and outlet vents are meshed as a whole simultaneously. Totally, there are 71318
hexahedral cells in the geometry. The meshes at the entrance region of the flow on the
emission surface are finer than those in other locations, to reflect the drastic variations of
variables in the boundary layer. The discretised meshes are shown in Figure 3.9. The graph is
amplified vertically to view the meshes above the emission surface clearly. The computations
are performed using the finite volume technique. The derivatives of the diffusive terms in the
N-S equations are approximated by second-order central difference and the derivatives of the
convective terms by first-order upwind difference. The discretised equations are solved by an
iterative procedure and in each step they are solved by the alternating direction implicit (ADI)
method. The coupling between velocity and pressure is performed through the SIMPLE
algorithm [9]. A relaxation factor of 0.65 is always required in iterations.
The convergence of the iterative procedure is studied following the evolution of the
normalized residues. When the normalized residues for mass, velocity and vapor
concentration are less than 10-5 at every node, the iterations are considered to be converged.
The fact is that after 500 iterations, the solution is usually converged, regardless of Reynolds
number.
The accuracy of the numerical method is determined from solutions on successively
refined grids. The rms error defined by Fletcher [10], and based on the normalized velocity
components, is used for that purpose:
1/ 2
⎡⎛ ⎞ ⎤
(
rms(u * ) = ⎢⎜⎜ ∑∑∑ u i*,1j ,k − u i*, j ,k )2
⎟ / m⎥
⎟ (3.44)
⎢⎣⎝ i j k ⎠ ⎥⎦
where ui,j,k*1 and ui,j,k* represent quantities calculated with grids having m1 and m number of
nodes, respectively. With the grids mentioned above, the solutions at V=509ml/min have an
rms error lower than 0.011. This value increases with flow rate and is about 0.018 at
1000ml/min.
o
φ
r
Figure 3.9. Discretized meshes of the calculating domain, half of the cell, amplified vertically.
34 Li-Zhi Zhang
Experimental Work
The cell is circular and made of stainless steel, with a diameter of 150mm and a
maximum height of 18mm. The deepness of the lower chamber is 10mm. A more detailed
description of the structure and parameters of the cell can be found in [9]. When testing, the
material is placed on the bottom surface of the lower cavity and becomes an integral part of
the emission cell. In this test, to ease the measurement of convective mass transfer
coefficients, distilled water is used as the loading material, instead of VOCs emission
material. Rather than directly measuring the emission rate profiles of VOCs from building
materials, convective surface mass transfer coefficients are calculated by measuring the
humidity differences between the inlet and outlet of an air stream, which flows through the
cell and exchanges moisture with water on the lower surface.
To investigate the local mass transfer coefficients at different cell radial locations, 7 glass
discs with thicknesses of 10mm and diameters ranging from 134 to 148mm are prepared. In
each test, a disc is placed in the lower chamber and on its bottom surface. Distilled water is
injected in the space between the chamber wall and the disc. When finished, the lower
chamber, the water, and the glass disk are on the same horizon. Special care is given to ensure
the water doesn’t wet the disc’s upper surface. With this method, the mass transfer area
between the water surface in the cell and the air is controlled. A picture showing the
placement of a glass disk in the cell lower chamber is shown in Figure 3.10. When the air
flows through the cell, it exchanges moisture with the water and is humidified. Since the RH
of water on the water surface is 100%, by calculating the humidity differences, the mean
convective mass transfer coefficients across the water surface can be calculated. Different
discs therefore provide the mean mass transfer coefficients at various cell radial locations. To
investigate the influences of different gases, both air and O2 are used in the experiment.
Figure 3.10. A picture of the placement of glass disk in the chamber.

Pressure Meter Pump/Flow Meter/Controller Tem/RH sensors

Valve
Bypass
Water
Compressed air
Bubbler
Cell Manometer
Flec
Figure 3.11. The set-up of the test apparatus.
The cell is supplied with clean and humidified air from an air supply unit. The complete
test rig is shown in Figure 3.11. The supply air flows from a compressed air bottle and is
divided into two streams. One of them is humidified through a bubbler immersed in a bottle
of water, and then re-mixed with the other dry air stream. The humidity of the mixed air
stream is controlled by adjusting the proportions of air mixing. The airflow rates are
controlled by two air pumps/controllers at the inlet and outlet of the cell. To prevent outside
air from infiltrating into the cell, a manometer is installed to monitor the pressure inside the
cell and ensure that it is positive. The humidities and temperatures inside and outside the cell
are measured by the built-in RH and temperature sensors, which are installed in the
pumps/controllers. The measuring accuracies are respectively 2% for relative humidity, 0.2°C
for temperature, and 2.5% for airflow rate.
Once the humidity differences are measured, the mean mass transfer coefficient is
calculated by
VΔω
Shm = (3.45)
At Δϖ
where Δω is the humidity ratio differences between the air inlet and outlet (kg/kg), At is the
transfer area between the air and water surface (m2), Δϖ is the logarithmic difference of the
humidity ratio between the water surface and the air in cell (kg/kg), and Shm is the mean
Sherwood number.
Local and Mean Sherwood Numbers
The variations in the measured outlet relative humidity from cell with disks of four
diameters are plotted in Figure 3.12. The inlet humidity is set to a prefixed value and the gas
is air. As can be seen, the outlet humidity decreases as the flow rate increases. However,
when the diameters are less than 135mm, the outlet humidity changes little.
36 Li-Zhi Zhang
100
90
80
Outlet RH
70
148mm
60
144mm
140mm
50
135mm
134mm
40
100 200 300 400 500 600
V (ml/min)
Figure 3.12. The outlet relative humidity of air from cell with 4 disks and various flow rates.
0.8
170ml/min
186ml/min
245ml/min
316ml/min
0.6 420ml/min
509ml/min
ShL
0.4
0.2
0
1 0.75 0.5 0.25 0
r/r0
Figure 3.13. Local Sherwood numbers along the cell radius for air.
1.2
170ml/min
186ml/min
1 245ml/min
316ml/min
420ml/min
0.8 509ml/min
Shm
0.6
0.4
0.2
0
1 0.75 0.5 0.25 0
r/r 0
Figure 3.14. Mean Sherwood numbers at various cell radii for air.
The calculated local and mean Sherwood numbers at different cell radius under various
volumetric flow rates for air are shown in Figure 3.13 and Figure 3.14, respectively. The
curves for nitrogen are similar to those for air. These two figures reveal the fact that the
Sherwood numbers decrease as the flow progresses. They are very large in the entrance
region, and they decrease as the radius decreases, asymptotically approaching the fully
developed values. The higher the flow rates (Reynolds numbers), the larger the Sh numbers.
For flow rates ranging from 170 to 509ml/min, the mean Sherwood numbers of the whole cell
emission surface change from 0.05 to 0.2.
The mean Sherwood numbers are also obtained from the experiments. Comparisons of
the mean Sherwood numbers between those calculated and experimentally obtained are
plotted in Figure 3.15. It is shown that the experimentally obtained values are in good
agreement with the numerical data. The largest deviation (24%) happens when the Sh is very
large, namely, at the position near the air slit, where the air begins to flow on the emission
surface. This phenomenon may result from the influences of the inlet flow conditions.
However, more than 90% of the numerical results are within ±5.5% deviation from the
experimental data.
Figure 3.16 represents the variation of the dimensionless bulk humidity against the cell
radius. The bulk humidity decreases very quickly after the air begins to make contact with the
water surface. This means that the convective mass transfer coefficients are very large in the
first quarter of the cell radius along the flow. Under current airflow rate conditions (less than
509ml/min), the air becomes nearly saturated in the remaining 3 quarters of cell radius along
the flow. This is due to the very small spacing between the emission surface and the cell cap
(minimum 1mm). When the air flow rate is increased, the exhaust air becomes less saturated.
38 Li-Zhi Zhang
2.5
Correlated Sh m Δ
1.5 Δ
Δ
1
Air
0.5 O2
N2
0
0 0.5 1 1.5 2 2.5
Experimental Sh m
Figure 3.15. Comparisons of Sh numbers correlated and experimentally obtained.
0.2
170ml/min
186ml/min
245ml/min
0.15 316ml/min
420ml/min
509ml/min
θb 0.1
0.05
0
1 0.75 0.5 0.25 0
r/r 0
Figure 3.16. Dimensionless bulk humidity ratios along the cell radius for air.
For ease of calculation, a multi-variable linear regression technique is used to analyze the
local and mean Sh numbers, and the bulk humidity. Three correlations have been obtained, as
follows:
−0.6761
⎛ r0 − r ⎞
Shm = 0.8271 Re 0.8578
Sc 0.6790
⎜ ⎟ (3.46)
⎝ 2δ ⎠
−0.834
⎛r −r⎞
ShL = 0.3259 Re Sc⎜ 0 ⎟ (3.47)
⎝ 2δ ⎠
−0.630
⎛ r0 − r ⎞
θ b = 0.2126 Re 0.358
Sc 0.806
⎜ ⎟ (3.48)
⎝ 2δ ⎠
where the validity is for gas with flow conditions of 0<Re≤200.

The maximum deviations between these correlations and the experimental and/or
numerical values are 6.8%, 6.4% and 7.7% for mean Sh, local Sh and θb, respectively.
Flow Patterns
Figure 3.17 gives an overview of the flow velocity vectors in the cell. The shapes are
similar to previous studies. The streamlines in a cross-section at φ=90° are shown in Figure
3.18. The flow inside the cell can be analyzed in three distinct regions: the impingement
region, where the flow extends from the inlet to the bottom surface and changes from axial to
radial due to the presence of the bottom surface; the radial flow region, where the air flows
inwardly on the emission surface and exchanges moisture with it; and the exhaust region,
where the air changes direction from radial to axial and is exhausted at the center of the cell.
φ
r
Figure 3.17. Three-dimensional velocity vectors in the cell for V=1.5L/min.

40 Li-Zhi Zhang
r o
Figure 3.18. Streamlines representation in a cross-section (φ=90°) for V=1.5L/min.
Impingement Region
The bottom surface imposes a shift in the air flow direction. The fluid decelerates in the
axial direction, losing kinetic energy that is converted into pressure energy. The deceleration
starts at the inlet exit and intensifies on approaching the lower surface. The increased pressure
is then primarily transformed into the radial momentum of the fluid, while some of it is
transformed into flows to other directions. Due to the confinement of the cell walls and the
small spacing between the cap and the lower surface, two vortices on both sides of the axis of
the air inlet are generated. In other words, in the impingement region and in the vicinity of the
inlet, the fluid is occupied by axisymmetric recirculating regions. The rotation axes of the
rolls are perpendicular to the inlet flow direction. The recirculating zones diminish with
increased angles from the air inlet. In this symmetry plane, the distance to the inlet is so long
that no rolls are generated. As a consequence, no vortices can be observed in this figure.
Radial Flow Region

The flow is distributed radially in the space between the bottom surface and the cap. The
cell is designed so that the radial-flow cross-section area is invariant with radial location in
this region. Therefore, the bulk velocity changes little along the radius. However, the local
velocity at the same altitude changes due to different duct heights. The streamlines and the
local velocity contours in a horizontal cross-section at half the spacing are observed. When
the air low rate is below 1000ml/min, the mean radial velocity above the emission surface is
relatively uniform. It does not vary much with regard to angle and radial locations. However,
when the air flow rate is further increased, the flow near the air inlet (in the area of φ=60-
120°) becomes demonstrably higher than at other positions. This is obviously the influence of
the air inlet.
Exhaust Region
The air flows from radial to axial locations and is exhausted in the center. In this region,
the radial air velocity is very small, and the resulting mass transfer between the air and the
emission surface is negligible.
Humidity Profiles
Due to the small spacing of 2mm between the cap and the bottom surface, the air will
become saturated as it flows along the cell radius if the air flow rates are below 1000ml/min.
To see the humidity profiles more clearly, the humidity contours for a larger air flow rate,
namely, 1500ml/min, are discussed. The calculated humidity distribution in a horizontal plate
at half the spacing is plotted in Figure 3.19. The humidity contours in a vertical cross-section
at φ=90° are plotted in Figure 3.20. The lines in the figures are constant humidity ratio lines.
For this case, the inlet temperature and humidity conditions are 23.4°C and 0.0035kg/kg,
respectively. This figure shows that steep humidity gradients exist in the entry region on the
emission surface. When the radius decreases, humidity gradients decrease drastically. This
proves that the convective mass transfer coefficient has the largest value at the beginning of
the flow and decreases with the flow’s progress. Vertically, humidity gradients are very steep
near the emission surface. In addition, under such a flow rate, the air inlet will seriously
influence the uniformities of the velocity and humidity fields. As can be seen, unlike the
contours under small flow rates, the contours here are not shaped in concentric circles, but in
irregular curves.
In this section, the fluid flow and convective mass transfer coefficients are
experimentally and numerically investigated. Three correlations are summarized to calculate
the mean Sh, local Sh and dimensionless bulk humidity along the chamber radius. For air flow
rates below 1000ml/min, the influences of the inlet on velocity distribution on the emission
surface is negligible. The velocity and humidity distributions are uniform and the local mass
transfer coefficients are only functions of radial locations. For larger air flow rates, the
influence of the inlet on the velocity and humidity fields becomes substantial. Overall, the
local mass transfer coefficients are very large in the entrance region, and they decrease as the
radius decreases, asymptotically approaching zero at the center of the cell. The flow inside the
cell can be analyzed in three distinct regions: the impingement region, the radial flow region,
and the exhaust region. The flow in the impingement region is rather complex: around the
axis of the inlet, the flow is occupied by axisymmetric recirculating regions. The rotation axes
of the rolls are perpendicular to the axis of the inlet. The flow in the other two regions is
relatively simple. Under current design airflow conditions, the air becomes nearly saturated
shortly after it begins to flow on the emission surface, due to the small spacing between the
cap and the bottom surface.
These data provides the basics for moisture diffusivity measurement with the cell.
42 Li-Zhi Zhang
Figure 3.19. Humidity profiles in a horizontal cross-section at z=0.5mm for V=1.5L/min.
Figure 3.20. Humidity profiles in a cross-section (φ=90°) for V=1.5L/min.

3.4. MEASUREMENT OF MOISTURE DIFFUSIVITY WITH THE CELL

It is well known that hydrophilic polymer membranes, say, poly cellulose acetate, poly
vinylidene fluoride, polyethersulfone, Nafion, and polyvinyl alcohol, are useful for the
separation of water vapor from other gases in air mixture, because the water molecule is
easily incorporated into the hydrophilic polymer membranes, due to the strong affinity
between the water molecule and the hydrophilic polymers, which facilitates the transport of
water, while impeding the permeation of other gases through the membranes.
Moisture transport properties in such hydrophilic polymer membranes are the most
important parameters affecting the system performance and the proper design of the units.
Traditionally, the measurements of water vapor diffusivity in membranes are conducted by
two ways: transient drying experiments, as previously described, and permeation tests [11-
13]. In the transient drying experiments, transient losses of membrane weight are recorded to
calculate the effective moisture diffusivity, with the analytical solution of Fick’s second law
of diffusion. Though popular and extensively used, this technique has inherent problems: the
assumptions and the operating conditions for the analytical solution of Fick’s law are very
rigorous and any deviation from this would lead to substantial errors [14]. On the other hand,
the permeation tests, though directly measure the moisture transport through membranes at
steady state, are rather complicated in the set-up. Most importantly, the obtained data are
case-specific, since the convective resistance on both sides of the membrane usually plays an
important role in the moisture transport performance, but was always neglected, which would
make the results less accurate.
In previous section, the convective mass transfer characteristics in the cell are obtained.
This section will extend the work to measure the moisture diffusivity through hydrophilic
polymer membranes (or sheets) with the cell and the test rig. The difference between this
section and [15,16] is that the convective mass transfer coefficients are the above obtained
values. Though the methodologies have similarities.
The Whole Set-up
The cell has been illustrated in Figures 3.7-3.8. In this test, distilled water is poured into
the lower chamber of the cell. Then a hydrophilic polymer membrane is covered on the lower
chamber. Following this step, the cap of the cell is placed on the membrane to form a
sandwiched structure. The membrane holding module is shown in Figure 3.21. A 2 mm gap
between the water layer and the membrane tested is kept. The saturated solution in the lower
chamber supplies a constant humidity ratio below the membrane lower surface. When the
humidity ratio between the two sides of the membrane is different, moisture will diffuse
through the membrane. Humid air is supplied from the inlets of the cap, which will exchange
moisture with the membrane and the humidity ratio will change along the path. During the
test, the temperature is kept constant. The relative humidity of the inlet and the outlet air
streams are measured, and the moisture exchange effectiveness can be calculated.
44 Li-Zhi Zhang
Inlet Outlet Inlet

Cap
Membrane
Lower chamber
Distilled water
Figure 3.21. Arrangement of the cell with tested membrane and water.
The cell is supplied with clean and humidified air from an auxiliary air supply unit of
cell. The complete test rig is shown in Figure 3.22. The system is similar to Figure 3.11,
however a membrane is sandwiched by the cell. The supply air flows from a compressed air
bottle and is purified through an AC carbon column filter and is then split into two streams.
One of them is humidified through a bubbler immersed in a bottle of distilled water, and then
re-mixed with the other dry air stream. The outlet humidity from the bottle with bubbler
reaches nearly 100%. The humidity of the mixed air stream is controlled by adjusting the
ratios of air mixing by two needle valves on each stream. The desired relative humidity is
obtained with this method. The airflow rates are controlled by two air pumps/controllers at
the inlet and outlet of the cell. This kind of pump has a built-in digital controller which keeps
the flow rates to the set points. To prevent outside air from infiltrating into the cell, a
manometer is installed to monitor the pressure inside the cell and ensure that it is positive.
The humidities to and from the cell are measured by RH sensors, which are installed after the
pumps/controllers. The measuring accuracies are respectively 2% for relative humidity, and
2.5% for airflow rate. The total uncertainty is less than 7.5%. The design of the system
allowed versatility of membrane replacement and the use of existing knowledge of flow in the
cell.
Before each test, the humidity sensors and the flow meters are carefully calibrated with a
chilled-mirror dew point meter (accuracy 0.1°C) and a floating ball flow meter (5ml/min),
respectively. Then the distilled water is poured into the lower chamber. Following this step,
the tested membrane is placed on the lower chamber and a sandwiched structure is formed
with the cap placed on the membrane. Special care is taken to prevent the membrane be
wetted by the water. Then the cell’s outlet and inlet are closed for 24 hours to let the
membrane and the cell volume become fully equilibrium with the water.
Flow Meter and

Pressure Meter Controller/Pump
Valve
By pass
Distilled
Water
Flow Meter and

Compressed air Bubbler Humidity Sensor Controller/Pump
Humidity Sensor
The cell with membrane
and
andsolution
water
Figure 3.22. Experimental equipment set-up for moisture transport tests.
Two hydrophilic polymer membranes, the mixed cellulose (MC, mixed cellulose and
pyroxylin) and acetate cellulose (AC), supplied by a local company, are selected as test
material. The reason for selecting them is that they are cheap, highly hydrophilic, and have a
certain mechanical strength, which are the necessary virtue for industrial applications. A
general isotherm equation can be fitted to represent the equilibrium water uptake with
moisture air as
Wmax
w= (3.49)
1 − C + C / RH
where Wmax is the maximum water uptake of membrane material (kg/kg); C is a constant
named the shape factor for the material; RH is air relative humidity.
The relative humidity is calculated by humidity ratio and temperature as [16]
RH e 5294 / T
= − 1.61RH (3.50)
ω 10 6
where T is in K. The second term on the right side of the equation will generally have less
than a 5% effect, and it can be usually neglected. Within a certain low concentration range,
Henry’s sorption law can be used to simplify the isotherm as
w = k pω (3.51)
where kp is the partition coefficient, kg air/kg membrane.

Table 3.4 lists the physical properties for the membrane materials. Since the membrane
fabrications have great influences on the membrane properties, they are measured in the
laboratory after the materials are purchased.
46 Li-Zhi Zhang
Table 3.4. Physical properties of the tested materials
Properties Mixed cellulose (MC) Acetate cellulose (AC)

Pore size (μm) 0.21 0.13
Porosity 0.72 0.86
Membrane thickness (μm) 115 105
Thermal conductivity (W/mK) 0.95 0.81
Density (kg/m3) 760 866
Sorption potential, Wmax (kg/kg) 0.27 0.46
Shape factor (C) 0.88 0.74
After the 24 hours equilibrium stage, the supply air is adjusted through the controlling
valves to the desired flow rate and the desired relative humidity. Then the valves before and
after the cell are opened to let the conditioned air stream flow through the cell. In the cell, the
air stream exchanges moisture with the membrane, and consequently exits the system.
Relative humidities are recorded by a data logging system as the air stream begins to flow the
cell. This process continues for 1-2 hours, long after the moisture transfer becomes fully
steady and the outlet relative humidity reaches a stable value. The whole test is performed
under room temperature conditions and it is controlled to within 0.5°C variations during the
test. Constant inlet relative humidity and air flow rates are maintained by the air supply unit.
After each test, the inlet humidity and air flow rates are set to new values to perform the
next experiment. Modifications to the air supply unit are done to supply larger volumetric
flow rates than the original cell system can.
Model Development
The governing equations for predicting the moisture transport in the emission cell are
developed. The schematic of the problem is represented in Figure 3.23. A control volume
based mass balance method is employed to obtain the partial differential equations. To aid in
the analysis, some assumptions are made as following:
(1) Fick’s law applies to the moisture diffusion in membrane. The thermo-physical
properties of membrane keep constant throughout the experiment. The process is
isothermal.
(2) Moisture diffusion in the flow direction in the air stream is negligible compared to
vapor convection by bulk flow. This assumption is true for Peclet number greater
than 10.
(3) Vapor diffusion in membrane is one-dimensional and in thickness direction. This is
valid considering the large dimensional differences in membrane geometry.
(4) The saturated solution is in equilibrium with membrane lower surface throughout the
test process.
Air stream z
r δ Membrane
O
r0
Figure 3.23. A schematic of the problem.
The moisture conservation in air stream gives:
∂ω ∂ω Ev
+ um = (3.52)
∂t ∂r H d ρ a
where ω is the humidity ratio (kg moisture/kg air); t is time (s); um is air bulk velocity (m/s)
along radius; r is radius coordinate (m); Ev is the local emission rate from the membrane to air
(kgm-2s-1); Hd is duct height of air stream (m); ρa is density of dry air (kg/m3). The cell is
specially designed that a constant um along the radius is realized.
Moisture diffusion in membrane
∂w ∂ ⎛ ∂w ⎞
= ⎜ Dvm ⎟ (3.53)
∂t ∂z ⎝ ∂z ⎠
where w is moisture uptake in membrane (kg moisture/kg dry membrane); z is coordinates in

membrane thickness (m); Dvm is moisture diffusivity in membrane (m2/s).
Local moisture emission rate
∂w
E v = − ρ m Dvm (3.54)
∂z z =2z0
where ρm is density of membrane (kg/m3).

Initial conditions:
t=0, ω=ωL, w=wL (3.55)
where ωL is the humidity ratio determined by the saturated NaCl solution and temperature
(kg/kg); and wL is the water uptake of membrane in equilibrium with the solution vapor.
Boundary conditions for air stream:
r=r0, ω=ωi (3.56)

48 Li-Zhi Zhang
r=0, Outflow (3.57)
Boundary conditions for membrane:
z=2z0, ω=ωs (3.58)
z=0, ω=ωL (3.59)
where ωi is the set point of humidity ratio of inlet air, ωs is the humidity ratio on membrane
surface. The relations between the humidity ratio on surface and in air stream are:
∂w
kρ a (ω s − ω ) = − ρ m Dvm (3.60)
∂z z = 2 z0
where k is the convective mass transfer coefficient (m/s).

The convective mass transfer coefficients are related to the fluid dynamics in the cell,
which are studied in detail in a previous section. The deduced correlation of Eq.(3.47) are
used to calculate convective mass transfer coefficients.
Re-arrangement of Eq.(3.60) and substituting it into Eq.(3.54) gives
ωL − ω
Ev = (3.61)
1 2ρ a z0
+
k Dvm ρ m k p
We define the total moisture transfer coefficient as
1
K= (3.62)
1 2ρ a z0
+
k Dvm ρ m k p
Then the Number of transfer units is defined by
KAt
NTU = (3.63)
NV
where At is the transfer area of the membrane surface (m2); N is the air exchange rate (s-1); V
is the volume of the cell (m3).
Similar to the analysis of a heat exchanger, a moisture exchange effectiveness can be
defined as
ωo − ωi
ε= (3.64)
ω L − ωi
where subscripts o, i, refer to air outlet and inlet respectively, and L to saturated humidity of
liquid water in lower chamber.
The transient equations for air stream, membrane and air-membrane interface are solved
in a coupled way with finite difference techniques. Iterations are necessary to find a
converged solution. At each time step, mass balances between the air stream, the membrane,
and the air/membrane interface are ensured. The outlet air relative humidity can be calculated
with moisture distributions in the cell and the local moisture emission rates on the membrane.
With time lapsing, outlet RH first decreases and then reaches a stable value, indicating that
moisture transport becomes stable. This stable value is called the steady state outlet relative
humidity. With this value, the moisture exchange effectiveness is calculated. Figure 3.24 plots
the calculated moisture exchange effectiveness from the distributed model with various
Number of Transfer Units. When the NTU is from 0 to 6, ε rises sharply with NTU. When the
NTU is greater than 6, there’s little merit in further increasing ε by an increase of NTU. When
the membrane thermophysical properties and the experimental conditions are known, NTU
can be calculated from this curve with the measured moisture exchange effectiveness. Finally,
the diffusivity can be obtained.
A polynomial correlation can be fitted to represent this curve as
ε = - 0.0002 NTU 4 + 0.0052 NTU 3 - 0.0638 NTU 2 + 0.3759 NTU + 0.015 (3.65)
R2 = 0.9998 (3.66)
0.8
0.6
ε
0.4
0.2
0
0 2 4 6 8 10 12
NTU
Figure 3.24. The variations of the moisture exchange effectiveness with the number of transfer units.
50 Li-Zhi Zhang
Estimation of Moisture Diffusivity
On the other hand, when the measured moisture exchange effectiveness, the membrane
thermophysical properties, and the experimental conditions are known, NTU can be estimated
from this curve. Finally, the diffusivity can be estimated.
NTU is a dimensionless parameter that governs flow rates, exchange area, and transfer
coefficient. It determines the moisture exchange effectiveness. It simultaneously reflects the
characters of flow rates, area, surface resistance and membrane resistance.
Specifically in this test, under air flow rate 10L/min, the measured moisture exchange
effectiveness are 0.67 and 0.81, for MC and AC membranes, respectively. As shown in
Figure 3.25, the corresponding NTU for these two ε are 2.75 and 4.4, respectively. The
resulting diffusivity can be then calculated from Eqs.(3.62), (3.63) as 3.1×10-10m2/s and
2.1×10-10m2/s for MC and AC membrane respectively.
The benefits with this approach are that it simultaneously considers the cell fluid
dynamics, membrane configurations and operating conditions. The uncertainties of the
measured values are related to the air flow rates through the cell. Under current flow rates, the
estimated uncertainty is 7.5%.
Substituting parameters in this study into Eq.(3.63), the resistance for membrane is 46.3
m s/kg, the convective mass transfer resistance is from 3.6 to 86.2 m2s/kg under 10L/min air
2
flow rate, depending on the cell location. The two resistances are in the same magnitudes,
therefore the convective mass transfer resistance cannot be neglected. The diffusion resistance
in air gap below the membrane is 4 m2s/kg, one order lower than the other two, therefore this
resistance can be neglected.
0.8 AC
MC
0.6
ε
0.4
0.2
0
0 2 4 6 8 10 12
NTU
Figure 3.25. Estimation of moisture exchange effectiveness with number of transfer units.
0.8
AC
MC
Outlet RH 0.7
0.6
0 0.5 1 1.5 2 2.5
t (s)
Figure 3.26. Transient variations of outlet RH, flow rate 10L/min.
0.7
0.6 AC
MC
0.5
0.4
Air RH
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
1-r/r 0
Figure 3.27. Bulk relative humidity of air stream along membrane cell radius.
Transient lapses of outlet RH with time for the two membranes are depicted in Figure
3.26. The volumetric air flow rates are kept at 10L/min. The discrete dots are the measured
values, and the solid and dashed lines are calculated values. As can be seen from the figure,
the model predicts the tested data reasonably well. Generally it takes around 1 second for the
moisture transport becomes steady. After this transient period, stable transport of moisture
from the solution to the air stream is established. Contrary to the assumption of a stepwise
52 Li-Zhi Zhang
square wave change of air humidity, in the transient period, the outlet air humidity changes on
a slope manner. The transitional time and the degree of slope are co-determined by the initial
conditions: moisture in the cell volume, and in the membrane. This indicates that in real
situations, estimation of moisture diffusivity by Eqs.(3.5)-(3.7) is problematic.
Figure 3.27 shows the variations of local relative humidity of the air stream along cell
radius when outlet RH becomes stable. As seen from this figure, air humidity rises
continuously from inlet to outlet, indicating continuous moisture emissions from the
membrane to air stream.
The local emission rates along the membrane radius are plotted in Figure 3.28 for the
steady transport period. From inlet to outlet, the moisture emission rates decrease, almost
linearly, along the membrane radius. Near the inlet, the two membranes have demonstrable
different vapor diffusion rates, but near the outlet, the emission rates are almost the same for
the two membranes. This indicates that near the inlet, emissions are controlled mainly by
membrane itself, but when the air humidity increases, the influence from membrane will
decrease.
To deeply disclose the distributed character of the emission rates, the water uptake
contours in the membrane are drawn in Figure 3.29 for the two membranes for the steady
state transfer period. The dimensionless thickness coordinates are defined as
z
z* = (3.67)
2z0
1.0E-04
AC
8.0E-05 MC
E v (kgm s )
6.0E-05
-2 -1
4.0E-05
2.0E-05
0.0E+00
0 0.2 0.4 0.6 0.8 1
1-r/r 0
Figure 3.28. Distributions of vapor emission rate on membrane along cell radius.
(a)
(b)
Figure 3.29. Water uptake contours in the membrane, (a) Mixed Cellulose and (b) Acetate Cellulose.
From this figure, the two-dimensional water uptake fields are clear. Water uptake
gradients near the air inlet is higher than those near the air outlet. This also indicates that
membranes at inlet contribute more to moisture transfer than those at outlet. In other words,
large disparities of moisture transfer exist on different membrane locations.
3.5. CONCLUSION
Sorption and diffusion parameters are the key basic parameters influencing heat and
moisture transfer in materials. How to estimate vapor diffusivity in membranes is one of the
most important issues in membrane related technology. The measurement of sorption
isotherms and diffusivity in a thermo-hygrostat is traditional. It is bulky and influenced by
boundary and operating conditions. As a novel method, a standard filed and laboratory
emission cell has been used to predict the moisture diffusivity in hydrophilic polymer
membranes. The membrane thermophysical properties and the hydrodynamics in the cell
geometry are considered simultaneously. Based on this step, the NTU-effectiveness method is
then used to study the moisture transport characteristics in the unit.
With the model proposed in the study, distributions of air humidity and emission rates on
membrane surface demonstrate a non-uniform character. The tested membrane’s diffusivity
can be estimated from the response curve of outlet RH. For the mixed cellulose membrane,
the measured diffusivity is 3.6×10-10m2/s. The water permeation potential can be reflected by
an effectiveness-Number of Transfer Units curve.
54 Li-Zhi Zhang
REFERENCES
[1] Okuno, H.; Renzo, K.; Uragami, T. Sorption and permeation of water and ethanol
vapors in poly (vinylchloride) membrane. J. Membrane Sci., 1995, 103, 31-38.
[2] Vagenas, G.K.; Karathanos, V.T. Prediction of the effective moisture diffusivity in
gelatinized food systems. J. Food Engng., 1993, 18, 159-179.
[3] Crank, J. The Mathematics of Diffusion, Second Edition. London: Oxford University Press;
1975.
[4] Asako, Y.; Maeda, K.; Jin, Z.; Yamaguchi, Y. Effective moisture diffusivity of super
absorbent polymer gel and pearlite-mortar with gel. Third Asia-Pacific Symposium on
Fire Science and Technology, Singapore, 1998, 359-370.
[5] Debeaufort, F.; Voilley, A.; Meares, P. Water vapor permeability and diffusivity
through methylcellulose edible films. J. Membrane Sci., 1994, 91, 125-133.
[6] Uhde, E.; Borgschulte, A.; Salthammer, T. Characterization of the field and laboratory
emission cell - FLEC: flow field and air velocities. Atmospheric Environment, 1998,
32, 773-781.
[7] Zhang, L.Z.; Niu, J.L. Effects of substrate parameters on the emissions of volatile
organic compounds from wet coating materials. Building and Environment, 2003, 38,
939-946.
[8] Zhang, L.Z.; Niu, J.L. Mass transfer of volatile organic compounds from painting
material in a standard field and laboratory emission cell (FLEC). International Journal
of Heat Mass Transfer, 2003, 46, 2415-2423.
[9] Zhang, L.Z.; Niu, J.L. Laminar fluid flow and mass transfer in a standard field and
laboratory emission cell (FLEC). International Journal of Heat Mass Transfer, 2003,
46, 91-100.
[10] Fletcher, C.A.J. Computational Techniques for Fluid Dynamics I, Springer Series in
Computational Physics. Berlin: Springer-Verlag, 1988.
[11] Ye, X.H.; Levan, M.D. Water transport properties of Nafion membranes Part I. Single-
tube membrane module for air drying. Journal of Membrane Science, 2003, 221, 147-
161.
[12] Scovazzo, P.; Burgos, J.; Hoehn, A.; Todd, P. Hydrophilic membrane based humidity
control. Journal of membrane Science, 1998, 149, 69-81.
[13] Scovazzo, P.; Hoehn, A.; Todd, P. Membrane porosity and hydrophilic membrane
based dehumidification performance. Journal of Membrane Science, 2000, 167, 217-
225.
[14] Hernandez-Munoz, P.; Gavara, R.; Hernandez, R.J. Evaluation of solubility and
diffusion coefficients in polymer film-vapor systems by sorption experiments. Journal
of membrane Science, 1999, 154, 195-204.
[15] Zhang, L.Z. Investigation of moisture transfer effectiveness through a hydrophilic

polymer membrane with a field and laboratory emission cell. International Journal of
Heat Mass Transfer, 2006, 49, 1176-1184.
[16] Zhang, L.Z. Evaluation of moisture diffusivity in hydrophilic polymer membranes: a
new approach. Journal of Membrane Science, 2006, 269, 75-83.
Chapter 4
PERFORMANCE OF ENERGY WHEELS
ABSTRACT
When hygroscopic materials are in contact with humid air, moisture would be
adsorbed by the materials. When they are in contact with dry air, the adsorbed moisture
would be released. Based on this mechanism, desiccant wheels can be used in total heat
recovery. In this chapter, detailed modeling of an energy wheel is conducted. Effects of
material parameters on the sensible and latent effectiveness are discussed. The vapor
sorption curves have a great influence on system performance.
NOMENCLATURE
a Pore radius (m)
As Transfer area (m2)
Bi Biot number
C Constant in sorption curve
cp Specific heat (kJkg-1K-1)
de Hydrodynamic diameter of a channel (m)
DA Combined ordinary and Knudson diffusivity (m2s-1)
DS Surface diffusivity (m2s-1)
f Fraction of desiccant in the wheel material
h Convective heat transfer coefficient (kWm-2K-1)
hm Convective mass transfer coefficient (kgm-2s-1)
H Specific enthalpy (kJ/kg)
k Thermal conductivity (kWm-1K-1)
L Length of a channel (m)
Le Lewis number
M1 Molecular weight (kg/kmol)
md Mass of the wheel (kg)
m g Mass flow rate of air stream (kg/s)
N Rotary speed (rpm)
NTU Number of transfer units
56 Li-Zhi Zhang
P Pressure (Pa)
qst Isosteric adsorption heat (kJ/kg)
SDP Specific Dehumidification Power (gkg-1s-1)
T Temperature (K)
t time (s)
ug Velocity of air stream (m/s)
V Volume (m3)
w Water uptake in desiccant (kg water/kg dry desiccant)
wmax Maximum water uptake of desiccant (kgkg-1)
x, y Coordinates (m)
Greek Letters
α Angle (rad)
β, λ Coefficients
θ Dimensionless temperature
ε Effectiveness
εt Total porosity
φ Relative humidity
δ Half thickness of channel (m)
ω Moisture content (kg moisture/kg dry air)
ρ Density (kg/m3)
ζ Tortuosity factor
τ Dimensionless time
ξ Resistance coefficient
Superscripts
* Dimensionless form of the variable
Subscripts
a Air
c Cooling, exhaust air
d Desiccant, dehumidification
g Gas
h Heating, fresh air
i Inlet
L Latent
m Moisture
o Outlet
opt Optimum
s Surface, sensible
w Water
Performance of Energy Wheels 57
x In x direction
y In y direction
4.1. INTRODUCTION
A schematic of a desiccant wheel is shown in Figure 4.1. The wheel is comprised of
parallel narrow ducts of hygroscopic materials like silica gel and LiCl impregnated paper.
When wet air passes through the channels, moisture is adsorbed or desorbed from the solid
channel walls. Another air stream, regenerating hot air, passes through wheel channels to
regenerate the adsorbed moisture from the adsorbents.
Desiccant wheels can be used in two fields: air dehumidification [1,2] and enthalpy
recovery [3-5]. In the first case, process air is dried after it flows through the wheel, which
rotates continuously between the process air and a hot regenerative air stream, as described in
the above figure. The dried air can either be used directly or be employed to make cooling
following further psychrometric processes, i.e. the so-called desiccant cooling. In the latter
case, the desiccant wheel rotates between the outside fresh air and the exhaust air from room.
Here fresh air is hot and humid while the exhaust air is cool and dry. Heat and humidity
would be recovered from the exhaust in winter and excess heat and moisture would be
transferred to the exhaust to cool and dehumidify the process air in the summer. The latter
case is also called as passive dehumidification. The wheel is called as an energy wheel. This
chapter will focus on this topic.
Fresh air Dried air
Regenerating air
Figure 4.1. A schematic of a desiccant wheel.

58 Li-Zhi Zhang
Due to the complexity in heat and mass transfer, mathematical modeling has been an
efficient tool to evaluate system performance. There have been many works in modeling heat
and moisture transfer in desiccant wheels. As far as this topic concerned, Simonson and
Besant [6] adopted the one-dimensional solid heat conduction equation in their model to
account for the longitudinal thermal resistance inside the solid. However, the model of
Simonson and Besant neglected the internal moisture resistance in the solid. Zhang and Niu
[4,5] proposed a dual-diffusion model that takes into account both the heat and the moisture
resistance in two dimensions: axial and in thickness directions of the solid. Moisture transfer
is expressed in two forms: surface diffusion and gaseous diffusion (Knudsen and ordinary
combined). In this chapter, the model is similar to previous studies, but material is a newly
developed hygroscopic material for energy wheels.
4.2. MATHEMATICAL MODEL

The energy wheel is a rotating cylindrical wheel of length L and diameter dw and it is
divided into two sections: fresh air side cooling adsorption section (angle fraction α0) and
exhaust air side heating desorption section (fraction 1-α0), where the wet fresh air and dry
exhaust air streams are in a counter flow arrangement. The wheel generally consists of a
matrix of numerous flow channels which have, depending on the manufacturing process, a
rectangular, triangular or sinusoidal shape. The flow channels that are parallel to the axis of
the wheel usually have a base material with a desiccant material impregnated on their
surfaces. Modern manufacturing technology has made it possible for the desiccant material
closely cross-linked to the base material and distributed evenly into the macro-voids in the
base material. Therefore, in this model, the wheel is approximated by a flow channel of
homogeneous composite material which has a desiccant content of f =0.6∼0.8. Because the
cycles the wheel channels undergo in rotating are identical, wheel performance can be
represented by a single channel. The model used is transient and two-dimensional. Because of
symmetry, the mid-plane of a channel can be considered to be adiabatic, and a half-size
channel surrounded by dashed line is used as the physical model, as shown in Figure 4.2.
Performance Index
If the wheel is used for enthalpy recovery, two effectiveness is defined:

Sensible effectiveness, εs
m c (Tci − Tco )
εs = (4.1)
m min (Tci − Thi )
o δ x
Air Stream
de
Desiccant
Duct length, L
y
Figure 4.2. A side view of one of the channels in the wheel.
Latent effectiveness, εL
m c (ω ci − ω co )
εL = (4.2)
m min (ω ci − ω hi )
where m min is the least value of process and exhaust mass flows.
Heat and mass conservation for the air stream
1 ∂Tg ∂Tg
+ =
4h
(Ts − Tg ) (4.3)
u g ∂t ∂x d e u g ρ g cpg
1 ∂ω g ∂ω g
+ =
4hm
(ω s − ω g ) (4.4)
u g ∂t ∂x d eug ρ g
where ug is the velocity (m/s), Tg and ωg are temperature (°C) and humidity ratio (kg/kg)
respectively, t is time (s), x is axial coordinate (m), de is the hydrodynamic diameter of the
channel (m), ρg is the density (kg/m3) and cpg is the specific heat (kJ/kg/K), h and hm are the
convective heat transfer (kWm-2K-1) and mass transfer (kgm-2s-1) coefficients between the air
stream and the solid surface, respectively. In the equations, subscripts “s” and “g” refer to
“surface” and “gas” respectively. By using heat mass transfer analogy, the relations between
h and hm can be expressed as
h
hm = (4.5)
cpg Le
where Le is the Lewis number of air stream.

60 Li-Zhi Zhang
Air streams in the channels are fully-developed laminar flow. The convective heat
transfer coefficients are calculated by the peripherally averaged Nusselt numbers for tubes of
various cross-sectional shapes [8].
For cooling and adsorption section, i.e., 0≤α*<α0, the inlet conditions,
⎧Tg
⎪ x =0 = Tci
⎨ (4.6)
⎪⎩ω g x =0 = ω ci
For heating and desorption section, i.e., α0≤α*<1:
⎧Tg
⎪ x = L = Thi
⎨ (4.7)
⎪⎩ω g x = L = ω hi
where subscripts “c”, “h”, “i” and “o” refer to cooling air, heating air, inlet and outlet,
respectively. The dimensionless angle α*=α/2π, and α is angle in the wheel (rad).
Enthalpy conservation in the desiccant
∂Td ⎛ ∂ 2 T ∂ 2T ⎞ ∂w
ρ d c tot = k d ⎜⎜ 2d + 2d ⎟ + qst ρ d
⎟ (4.8)
∂t ⎝ ∂x ∂y ⎠ ∂t
where ρd is the density of dry desiccant (kgm-3), kd is the thermal conductivity of the solid
(kWm-1K-1), w is the water content in the desiccant (kg/kg), y is the coordinate in the
thickness (m), qst is the adsorption heat (kJ/kg), ctot is the total heat capacity of moist
desiccant, which includes two parts: dry desiccant and adsorbed water and is calculated by
c tot =c pd + wcpw (4.9)
where cpd and cpw are the specific heats (kJkg-1K-1) of dry desiccant and liquid water
respectively.
Two phases of water, namely, gas and adsorbed state, co-exist and diffuse in the pores of
the solid. There are three dominant diffusion mechanisms [5]: surface diffusion, ordinary
diffusion, and Knudsen diffusion. The first diffusion is in the form of adsorbed state and the
latter two are in gas state. If Fick’s law is used to express the diffusion dynamics, the
moisture conservation in the solid can be expressed as
∂ω ∂w ⎡∂ ⎛ ∂ω ⎞ ∂ ⎛ ∂ω ⎞⎤
ε t ρa + ρd = ρ a ⎢ ⎜ DA ⎟ + ⎜⎜ DA ⎟⎥
∂t ∂t ⎣ ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂y ⎟⎠⎦
(4.10)
⎡∂ ⎛ ∂w ⎞ ∂ ⎛ ∂w ⎞⎤
+ ρ d ⎢ ⎜ DS ⎟ + ⎜⎜ DS ⎟⎟⎥
⎣ ∂x ⎝ ∂x ⎠ ∂y ⎝ ∂ y ⎠⎦
where εt is the total porosity of the desiccant. On the right hand side of Eq.(4.10), the first
term is the moisture transfer in gas (combined ordinary and Knudsen diffusion), and the
second term is in the adsorbed phase, namely, surface diffusion. DA and DS are the effective
diffusivities (m2s-1) of the combined ordinary and Knudsen diffusion and surface diffusion,
respectively. They are calculated by the following equations [5].
−1
ε ⎛ 1 1 ⎞
DA = t ⎜⎜ + ⎟⎟ (4.11)
ζ ⎝ DAO DAk ⎠
DS =
1
ζ
(
D0 exp − 0.974 × 10 −3 qst / Td ) (4.12)
1.685
T −9
DAO = 1.735 × 10 d (4.13)
Pa
1/ 2
⎛T ⎞
DAK = 97 a⎜⎜ d ⎟⎟ (4.14)
⎝ M1 ⎠
where ζ is tortuosity factor that accounts for the increase in diffusional length due to the
tortuous path of the real pores, DAO is the ordinary diffusivity, DAK is the Knudsen diffusivity,
D0 is a constant for surface diffusion calculation, a is the pore radius of the adsorbent, Pa is
the pressure in atmospheres (atm), Td is in K, M1 is the molecule weight of water.
Water content in the desiccant is governed by a general sorption isotherm as
fwmax
w= (4.15)
1 - C + C/ φ
where wmax is the maximum water content (kg/kg), C is a constant that determines the
isotherm shape, φ is the relative humidity. Selecting T and ω as two independent variables, a
differential form of adsorption content can be written in terms of humidity and temperature as
dw = ψdω + ϕdT (4.16)

62 Li-Zhi Zhang
where
⎛ ∂w ⎞
ψ =⎜ ⎟ (4.17)
⎝ ∂ω ⎠ T
⎛ ∂w ⎞
ϕ =⎜ ⎟ (4.18)
⎝ ∂T ⎠ ω
Using Clapeyron equation to represent the saturation vapor pressure and assuming a
standard atmospheric pressure of 101325Pa gives the relation between humidity and relative
humidity [1] as
φ
= 10 −6 e 5294 / T − 1.61φ (4.19)
ω
where T is in K. The second term on the right side of the equation will generally have less
than a 5% effect, thus it can be neglected. Therefore
fwmax C
ψ = 10 −6 e 5294 / T (4.20)
(1 − C + C / φ )2 φ 2
5294φ fwmax C
ϕ =− (4.21)
T 2 (1 − C + C / φ )2 φ 2
where ψ reflects the slope of w to ω, ant it is a dimensionless variable. Partial differential ϕ

reflects the slope of w to T, and its unit is K-1. For enthalpy recovery wheel, the bigger the ψ,
the better the performance; for air dehumidification wheel, the greater the ϕ, the better the
performance.
Introducing the dimensionless temperature
T − Tci
θ= (4.22)
Thi - Tci
and the dimensionless time
tN
τ= (4.23)
60
as well as the dimensionless coordinates
x
x* = (4.24)
L
y
y* = (4.25)
δ
Energy equation (4.8) can be normalized to
∂θ ∂ 2θ ∂ 2θ ∂ω
= k x* *2 + k y* *2 + qst* (4.26)
∂τ ∂x ∂y ∂τ
where
60k d
k x* = (4.27)
NL ρ d (c tot − qstϕ )
2
60k d
k y* = (4.28)
Nδ 2 ρ d (c tot − qstϕ )
qst ρ dψ
qst* = (4.29)
ρ d (c tot − qstϕ )(Thi − Tci )
N is the rotary speed in rpm.

The mass equation (4.10) can be normalized to
∂ω ∂θ ∂ ⎛ ∂ω ∂θ ⎞ ∂ ⎛ ∂ω ∂θ ⎞
+ β 0* = β x * ⎜ λω * + λT * ⎟ + β y * ⎜⎜ λω * + λT * ⎟⎟ (4.30)
∂τ ∂τ ∂x ⎝ ∂x ∂x ⎠ ∂y ⎝ ∂y ∂y ⎠
where
ρ dϕ (Thi − Tci )
β 0* = (4.31)
ε t ρ a + ψρ d
λω = ρ a DA + ρ d DSψ (4.32)
λT = ρ d DSϕ (Thi − Tci ) (4.33)

64 Li-Zhi Zhang
60
βx = (4.34)
NL (ε t ρ a + ψρ d )
2
60
βy = (4.35)
Nδ (ε t ρ a + ψρ d )
2
Boundary conditions for the solid phase become:
∂θ ∂θ ∂θ
= = =0 (4.36)
∂y * y =0
* ∂x * x* = 0 ∂x * x* =1
∂ω ∂ω ∂ω
= = =0 (4.37)
∂y * y =0
* ∂x * x* = 0 ∂x * x* =1
∂θ
− = Bi (θ − θ g ) (4.38)
∂y * y =1
*
∂ω λT ∂θ
− − = Bim (ω − ω g ) (4.39)
∂y * y =1
* λω ∂y * y =1*
where, Bi is the Biot number for heat transfer
hδ
Bi = (4.40)
kd
and Bim is the Biot number for mass transfer
hmδ
Bim = (4.41)
λω
The heat and mass transfer equations for the air streams can be normalized as
∂θ g ∂θ g
c1 + = NTU (θ s − θ g ) (4.42)
∂τ ∂x *
∂ω g ∂ω g
c1 + = NTU m (ω s − ω g ) (4.43)
∂τ ∂x *
where
NL hAs
c1 = NTU =
60u g m g cpg
hm As
NTU m=
m g
4.3. PERFORMANCE ANALYSIS

The two dimensional heat and mass transfer equations of the desiccant are numerically
solved by means of ADI (alternating direction implicit) method [9]. Because the equations are
strongly coupled and nonlinear, iterations are necessary to get converged values for each time
step. Before numerical analysis can be performed, the physical domain of the problem as well
as the equations must be discretized. The whole calculating domain is divided into a number
of equal-step discrete elements. Each element is identified as a control volume by a nodal
point. The numbers of nodes are: 40 in axial, 5 in thickness, and 120 in time (angle). The
model has been validated by an experiment in [5]. In the following, performances with
varying operating conditions will be discussed. For enthalpy recovery, the inlet temperature
and humidity are set to: fresh air, 35 °C, 0.021 kg/kg; exhaust air 24 °C, 0.012 kgkg-1. The
base properties for the simulated desiccant wheel are listed in Table 4.1. The desiccant
material is a newly developed hygroscopic material. The geometry of the channels in the
wheel is sinusoidal with a width of 4.35 mm and a height of 1.74 mm. The sorption curve of
the wheel material is shown in Figure 4.3. As seen, it is a typical third class sorption curve.
This kind of material is appropriate for energy recovery.
Table 4.1. Base properties of the energy wheel
Symbol Unit Value

a m 11×10-10
C 6.0
D0 m2s-1 1.6×10-6
f 0.75
k Wm-1K-1 0.20
L m 0.1
md kg 15.0
m g kg/s 0.4
qst kJ/kg 2650

wmax kg/kg 0.92
66 Li-Zhi Zhang
Table 4.1. (Continued)
Symbol Unit Value

α0 0.50
ρd kgm-3 1129
ξ 2.8
δ mm 0.1
εt 0.70
1
w (kg/kg)
0.5
0
0 0.2 0.4 0.6 0.8 1
Figure 4.3. Sorption curve of the energy wheel material.
Temperature and Humidity Profiles
If the wheel is assumed stationary, then both temperature and humidity vary with wheel
angle. The profiles of temperature and humidity of air inlet, air outlet, air mean, and desiccant
mean across a channel with dimensionless time (or the wheel angle) during enthalpy recovery
are shown in Figure 4.4. As seen, both the mean temperature and the mean humidity of the air
stream are higher than those of the desiccant in the adsorption section (τ=0.5∼1.0, 1.5∼2.0),
and are lower than those of the desiccant in the desorption region (τ=0∼0.5, 1.0∼1.5), which
discloses a fact that both the moisture and the sensible heat are transferred from the fresh air
stream to the exhaust air stream. Contrary to in air dehumidification, in enthalpy recovery,
heating section is also the adsorption section, and the cooling section is the desorption
section. Furthermore, for balanced flows, when εS>0.5, the outlet temperature of the exhaust
will be higher than that of the process air. When εL>0.5, the outlet humidity of the exhaust
will be higher than that of the process air.
40
35
T (°C)
30
25 Air in
Air out
Air mean
Desiccant mean
20
0 0.5 1 1.5 2 2.5
τ (a)
0.03
Humidity (kg/kg)
0.02
Air in
Air out
Air mean
Desiccant mean
0.01
0 0.5 1 1.5 2 2.5
τ (b)
Figure 4.4. Profiles of temperature (a) and humidity (b) of air inlet, air outlet, air mean, and desiccant
mean across a channel with dimensionless time (wheel angle).
Effects of Rotary Speed
There exists an optimum rotary speed at which the efficiency reaches the climax. When a
desiccant wheel rotates much faster than the optimum speed, the adsorption and desorption
processes are too short, which results in poor performance. On the other hand, when the
rotary speed is lower than the optimum, the adsorption and desorption processes are too long
and wasting more energy in sensible heating/cooling rather than in sorption processes, and
therefore are less effective. The variations of performance with rotary speed are plotted in
Figure 4.5 for enthalpy recovery purposes. The wall thickness is 0.2 mm. The figure indicates
68 Li-Zhi Zhang
that the optimum rotary speed for sensible and latent heat recovery is faster. In the following
analysis, the wheels are operated at the optimum speeds.
0.8
εS
0.6
εL
0.4
ε
0.2
0
0 10 20 30 40 50
N (rpm)
Figure 4.5. Performance variations under various rotary speeds with a wall thickness 0.2mm.
Figure 4.6 shows the effects of wall thickness on optimum rotary speeds for total heat
recovery. The wheel usually reaches the highest sensible effectiveness and latent
effectiveness simultaneously at the same optimum speed. The resulting values for εS and εL at
the optimum speed are also shown in this figure. Similar to air dehumidification, the optimum
speed decreases as the wall thickness in the wheel increases. However, when the wall
thickness is bigger than 2.0mm, the optimum rotary speed becomes insensitive to wall
thickness. It is disclosed that only a fraction of the thickness takes part in the sorption-
desorption working cycles for thick wheels at the optimum speed. In other words, there is an
“active layer” for the wheel with thick channel walls. When the wheels are operated in the
optimal modes, desiccant in the “inactive layer” has no big use for the air dehumidification
and enthalpy recovery. Therefore, in practice, wheels with thin walls and large transfer areas
are recommended. That’s the reason why the honeycomb type desiccant wheels are advocated
in industry.
20 1
εS
0.8
15
εL
0.6
N opt (rpm)
L
εand
10
S
0.4
Nopt
5
0.2
0 0
0.0 1.0 2.0 3.0 4.0 5.0
Wall thickness (2δ, mm)
Figure 4.6. Optimum rotary speed for enthalpy recovery and the corresponding sensible and latent
effectiveness with various wall thickness. The number of channels is fixed.
Effects of NTU
The number of transfer units, NTU, has a great influence on the system performance.
There are many ways to modify NTU. When the wheel mass and channel size are fixed,
changing the wall thickness will change the channel numbers, and consequently the contact
area and NTU. When the wheel volume and wall thickness are fixed, changing channel sizes
will change the number of channels and the packing density, consequently the NTU. Figure
4.7 shows the influence of NTU on total heat recovery efficiencies. A NTU of 2.4 is needed
for sensible and latent efficiencies higher than 0.61. It is noted that the latent effectiveness is
usually smaller than the sensible effectiveness, because the moisture transfer resistance is
usually larger than the heat transfer resistance. In other words, mass diffusion in the solid is
far less than the thermal diffusion.
For desiccant wheels, when the total volume of the wheel is fixed, a higher specific area
can be achieved by constructing with smaller size channels and thinner channel walls. With
higher Av, the wheel will be more compact and both the transfer area and the heat mass
transfer coefficients will become larger. These factors all contribute to an increased
performance, as shown in Figure 4.8. As can be seen, the effectiveness rises almost linearly
with increasing specific area. Therefore, in practice, the honeycomb type desiccant wheels
which have large contact areas should be recommended. However, increased performance is
70 Li-Zhi Zhang
achieved at the price of increased pressure drop, as shown in Figure 4.9. The pressure drop
rises rapidly with Av. In other words, fan power requirement will be increased for honeycomb
wheels to realize higher performance. Eventually, a compromise between increased Av and
increased pressure drop has to be maintained. It is obvious that NTU increases with Av.
0.8
εS
0.6
εL
Effectiveness
0.4
0.2
0
0 2 4 6 8
NTU
Figure 4.7. Effects of NTU on performance for enthalpy recovery.
0.8 8
0.6 εS 6
Effectiveness
NTU
εL
NTU
0.4 NTU 4
0.2 2
0 0
0 500 1000 1500 2000 2500
2 3
A v (m /m )
Figure 4.8. Effects of specific area on performance.

250
200
150
P (Pa)
100
50
0
0 500 1000 1500 2000 2500
2 3
A v (m /m )
Figure 4.9. Pressure drop with various specific area, m g = 0.45kg/s, md=16kg.
Psychrometric Cycle
The cycle the desiccant undergoes during a revolving is certainly of interest. The
evolutions of the states of desiccant mean and air outlet in relation to wheel angle are plotted
in psychrometric charts, as shown in Figure 4.10. In the figure, the dashed lines are constant
enthalpy lines and the inlet states of air streams are represented by points Hi for fresh air and
Ci for exhaust air. The arrows indicate the directions of angle increasing or wheel revolving.
Points Ho and Co represent the two average outlet air states for fresh air stream and exhaust
air stream respectively. For the desiccant, three properties, namely, water content, air
humidity in the pores, and temperature, are plotted in the graph simultaneously.
In Figure 4.10, the desiccant mean states evolve along the same line (ac) during the
cooling process as during the heating process (ca), but with an opposite direction. The outlet
states of air distribute along line ac for cooling process whose inlet is Ci and mean outlet is
Co, and along line c’a’ for the heating process, where the inlet is Hi and the mean outlet is
Ho, respectively. The two inlet states Hi and Ci are on the same line of ac or c’a’. During
heating and desorption, both the temperature and the humidity increase, while during cooling
and adsorption, both the temperature and the humidity decrease with wheel revolving. As a
result, enthalpy increases for Ci→Co (cooling air inlet to outlet) and decreases for Hi→Ho
(heating air inlet to outlet). The variation of water uptake in desiccant during a wheel working
cycle is relatively very small. This is the reason why wheels for enthalpy recovery
applications should be rotated much faster than those for dehumidification to realize
optimized performances.
72 Li-Zhi Zhang
0.025 0.15
w (kg water/kg desiccant)

(kg moisture/kg air) a
a
c
0.02 aa
0.015 0.1
27 28 29 30 31
T T(°C)
(°C) (a)
0.03
Hi
a
ω (kg moisture/kg air)
Co a
0.02 a’
a'
Ho c
c’
Ci c'
0.01
0
20 25 30 35 40
TT(°C)
(°C)
(b)
Figure 4.10. Variations of the states of desiccant mean (a) and outlet air (b) across wheel angle for
enthalpy recovery: ac, during adsorption; ca or c’a’, during desorption; dashed lines, constant enthalpy.
4.4. CONCLUSION
A two-dimensional, dual-diffusion transient heat and mass transfer model presented in
chapter is superior to other one-dimensional ones. The advantages of such a model lie in the
fact that it considers the heat conduction, the surface and gaseous diffusion in both the axial
and the thickness directions simultaneously. The effects of the channel wall thickness can be
investigated. Many other structural and operating parameters of the wheel could also be
studied.
The temperature and humidity profiles in the wheel discloses a fact that unlike air
dehumidification, in total heat recovery, heat and moisture are in phase, i.e., heat and
moisture are transferred simultaneously from fresh air stream with high temperature and high
humidity to the exhaust air stream with low temperature and low humidity. Since the
operating conditions and the purposes are different, optimum rotary speeds for total heat
recovery are different from air dehumidification. Total heat recovery is less sensitive to rotary
speed than air dehumidification is. Besides, the optimum rotary speed for total heat recovery
is much faster than those for air dehumidification. There is a strong influence of wall
thickness on the system performance and the optimum rotary speed. The higher the thickness,
the lower the optimum speed. When the weight of the wheel is fixed, increased thickness
leads to poorer performance. For wheels with thick walls, only the “active layer” takes part in
the processes at the optimum rotary speeds. In other words, the thinner the wall, the more
effective is the desiccant used. An NTU of 2.5 is needed for desiccant wheels to have a good
performance.
REFERENCES
[1] Zheng, W.; Worek, W.M. Numerical simulation of combined heat and mass transfer
processes in a rotary dehumidifier. Numerical Heat Transfer, Part A: Applications,
1993, 23, 211-232.
[2] Tauscher, R.; Dinglreiter, U.; Durst, B.; Mayinger, F. Transport processes in narrow
channels with application to rotary exchangers. Heat and Mass Transfer, 1999, 35,
123-131.
[3] Simonson, C.J.; Besant, R.W. Energy wheel effectiveness: part I -development of
dimensionless groups. Int. J. Heat Mass Transfer, 1999, 42, 2161-2170.
1347- 1367.
[5] Niu, J.L.; Zhang, L.Z. Effects of wall thickness on the heat and moisture transfers in
desiccant wheels for air dehumidification and enthalpy recovery. International
Communications in Heat and Mass Transfer, 2002, 29, 255-268.
[6] Simonson, C.J.; Besant, R.W. Heat and moisture transfer in energy wheels during
sorption, condensation, and frosting conditions. ASME Journal of Heat Transfer, 1998,
120, 699-708.
[7] Majumdar, P. Heat and mass transfer in composite desiccant pore structures for
dehumidification. Solar Energy, 1998, 62, 1-10.
[8] Zhang, L.Z.; Niu, J.L. A Numerical study of laminar forced convection in sinusoidal
ducts with arc lower boundaries under uniform wall temperature. Numerical Heat
Transfer, Part A: Applications, 2001, 40, 55-72.
74 Li-Zhi Zhang
[9] Samarskii, A.A.; Vabishchevich, P.N. Computational Heat Transfer. New York: John
Wiley & Sons Inc.; 1995.
Chapter 5
HEAT MASS TRANSFER

IN BENDED SINUSOIDAL NARROW DUCTS
ABSTRACT
Convective heat mass transfer coefficients in narrow channels are the basic
parameters for performance analysis of energy wheels. Fluid flow and heat transfer in
channels of regular cross sectional shapes have existed in references for many years.
However narrow channels in energy wheels are irregular shapes. Heat mass transfer
coefficients in such channels are influenced by channel shapes. In this chapter, fluid flow
and mass transfer in bended sinusoidal ducts, which are the common duct shapes in
energy wheels, are numerically calculated. Sherwood numbers and friction coefficients
under fully developed laminar flow are calculated.
NOMENCLATURE
a Half duct height
As Cross-sectional area
b Half duct width
Dh Hydraulic diameter
Dva Diffusivity
e Bending ratio of duct
f Friction coefficient
hL Local convective heat transfer coefficient
J Jacobian operator
Nu Nusselt number
NuT Nusselt number for thermally fully developed laminar flow
P Pressure
Pe Perimeter of duct
Pr Prandtl number
Re Reynolds number
Sh Sherwood number
76 Li-Zhi Zhang
Shω Sherwood number for mass fully developed laminar flow

Sc Schmidt number
T Temperature
u Velocity
x, y Dimensional transversal coordinates
z Axial coordinate
R Radius of lower boundary
Greek Letters
τ Aspect ratio, duct height to width ratio

δ Height of lower limit of duct
ω Humidity
ρ Density
θ Dimensionless humidity
μ Dynamic viscosity
ξ, η Transversal coordinates in computational plane
Superscripts and Subscripts
* Dimensionless
b Bulk
i Inlet
m Mean
w Wall
ω humidity
5.1. INTRODUCTION
Rotary desiccant wheels are the hearts of various total heat recovery systems. Therefore,
much effort has been devoted to develop wheels of high performance combined with low
cost. The honeycomb type wheel, as shown in Figure 5.1, has drawn much attention due to its
coherent two advantages: large contacting area (3000m2/m3) and compactness [1-3]. A
honeycomb wheel is usually composed of numerous corrugated ducts where fresh air
exchanges moisture and heat with the solid adsorbent. Then the solid adsorbent exchanges
heat and moisture with exhaust air, with wheel revolving. In performances modeling,
convective heat mass transfer in ducts under uniform wall temperature or humidity boundary
conditions are calculated first. Then they are combined with heat mass diffusion in solid
walls.
There are various duct cross sectional geometries. The most commonly encountered
includes sinusoidal, triangular, rectangular, etc. Due to the small diameters, the Reynolds
numbers fall into the regime of a laminar flow.
Heat Mass Transfer in Bended Sinusoidal Narrow Ducts 77
Fresh air out

Fresh air in
Exhaust air in
Exhaust air out
Figure 5.1. Configurations of a honeycomb energy wheel.
Heat transfers of laminar flow have been extensively studied for regularly shaped ducts.
The work of Shah and London [4] contains a thorough review of heat transfer under
developing and fully developed laminar flow in ducts of many cross sectional shapes. In
recent years, with the progress on computational techniques, ducts of irregular shapes are
increasingly investigated. Sherony and Solbrig [5] investigated the heat and mass transfers in
a corrugated duct surrounded by a sine curve and a flat plate. Fischer and Martin [6] studied
the friction factors in ducts confined by corrugated parallel walls. Ebadian and Zhang [7]
studied the fluid flow and heat transfer in a crescent-shaped lumen catheter. Dong and
Ebadian [8] provided a numerical analysis of thermally developing flow in elliptic ducts with
internal fins. Fluid flow and convective heat mass transfer in ducts of various cross sectional
geometries are the basic data for performance analysis. Heat transfer coefficients are
relatively well documented, however mass transfer coefficients are less recorded.
Upper boundary
2a
Lower boundary
y
δ 2b
Figure 5.2. Cross-section view and geometry of a bended corrugated duct in the wheel.
78 Li-Zhi Zhang
5.2. FLOW AND HEAT MASS TRANSFER MODEL

The corrugated sinusoidal duct geometry is the most commonly used structure in
honeycomb wheels because it is advantageous in its simplicity of construction and large
surface area. In such small diameter ducts, laminar flow prevails. The cross-sectional
geometry of a corrugated sinusoidal duct is shown in Figure 5.2. It is not a common
sinusoidal duct. The duct is like a duct bended from a regular sinusoidal duct. It is formed
during the wheel making processes, when a stack of plate-fin sinusoidal channels are bended
to form a wheel. It is observed that the single tube can be approximated with a sine curve for
the upper portion and an arc for the lower portion [9,10]. Heat transfer coefficients in such
ducts have been calculated before. This chapter will focus on the calculation of mass transfer
coefficients in the ducts.
Because of the irregular geometry and the small size of the passages, it is very difficult to
directly measure anything but overall time-mean performances. Numerical solution method
becomes important and will supply much needed design information. To address the
complexity of the duct geometry, the numerically generated boundary-fitted coordinate
system is applied to discretize the computational domain. According to this technique, the
governing equations can be solved with regular geometric methods by transforming the
complex duct geometry to a regular square duct.
Basic Equations
The problem considered here is that of a duct shown in Figure 5.2. The upper boundary
can be expressed as a sinusoidal function
⎡ ⎛ π ⎞⎤
y = a ⎢1 − cos⎜ x ⎟⎥ (5.1)
⎣ ⎝ b ⎠⎦
where a is the half height of the sine duct, b is the half width of the duct.
The lower boundary is an arc, which can be expressed as
y = R (sin ω − 1) + δ (5.2)
⎛ x−b⎞
ω = arccos⎜ ⎟ (5.3)
⎝ R ⎠
1 ⎛ b2 ⎞
R = ⎜⎜ + δ ⎟⎟ (5.4)
2⎝ δ ⎠
where R is the radius of the arc (m), δ is the height of the arc (m). The value of δ can be
greater or less than 0. The signs of δ for two consecutive ducts in a wheel are opposite. When
δ=0, the lower boundary becomes a flat plane. The duct becomes a common flat plate
sinusoidal duct [5].
Aspect ratio of the duct
a
τ= (5.5)
b
Bending ratio of the duct is defined as
δ
e= (5.6)
b
The flow in the duct is considered to be laminar and hydrodynamically fully developed,
but thermally developing in the entrance region of the duct. The fluid is Newtonian with
constant thermal properties. Additionally, a uniform wall humidity boundary condition is
considered.
Momentum equation
For two-dimensional fully developed laminar flow (fluid has axial velocity only), the
Navier-Stokes equations reduce to [11,12]
⎛ ∂ 2 u ∂ 2 u ⎞ dP
μ ⎜⎜ 2 + 2 ⎟⎟ = (5.7)
⎝ ∂x ∂y ⎠ dz
where μ is dynamic viscosity (Pa.s), u is fluid velocity (m/s), P is the pressure (Pa), z is the
axial coordinate (m).
Energy equation
∂T ⎛ ∂ 2T ∂ 2T ⎞
ρc P u = k ⎜⎜ 2 + 2 ⎟⎟ (5.8)
∂z ⎝ ∂x ∂y ⎠
where T is the fluid humidity (K), k is thermal diffusivity(kWm-1K-1), ρ is density (kgm-3) and
cP is specific heat of air (kJkg-1K-1).
Mass conservation
∂ω ⎛ ∂ 2ω ∂ 2ω ⎞
u = Dva ⎜⎜ 2 + 2 ⎟⎟ (5.9)
∂z ⎝ ∂x ∂y ⎠
where Dva is vapor diffusivity in dry air (m2s-1)

This chapter only considers mass transfer. The above equations (5.7) and (5.9) can be
normalizes as
80 Li-Zhi Zhang
2
∂ 2u * ⎛b⎞ ∂ u
2 *
4b 2
+⎜ ⎟ + 2 =0 (5.10)
⎝ a ⎠ ∂y *
2 2
∂x * Dh
and
∂θ ∂ 2θ ⎛ b ⎞ ∂ 2θ
2
U * = +⎜ ⎟ (5.11)
∂z ∂x * ⎝ a ⎠ ∂y *
2 2
where, in the above equations, the dimensionless velocity is
μu
u* = − (5.12)
dP 2
Dh
dz
and dimensionless humidity is
ω − ωW
θ= (5.13)
ωi − ω W
where ωi is the inlet humidity, and ωw the humidity of duct wall.

Dimensionless coordinates
x
x* = (5.14)
2b
y
y* = (5.15)
2a
z
z* = (5.16)
Dh ReSc
where the hydraulic diameter
4 As
Dh = (5.17)
Pe
where As is the cross section area of the duct (m2), Pe is the perimeter of the duct (m).
In Eq.(5.11), U is a coefficient defined by
u * 4b 2
U= * (5.18)
u m Dh2
where u*m is the average dimensionless velocity on a cross section, and it is calculated by
∫∫ u dA
*
u *
m = (5.19)
As
The characteristics of fluid flow and mass transfer in the duct can be represented by the
product of the friction coefficient and the Reynolds number, the dimensionless bulk humidity,
and the Sherwood number.
⎛ dP ⎞
⎜ D ⎟⎛ ρu D ⎞ 1
( fRe) = ⎜ − h dz2 ⎟⎜⎜ m h ⎟⎟ = * (5.20)
⎜ 2 ρu m ⎟⎝ μ ⎠ 2u m
⎜ ⎟
⎝ ⎠
∫∫ u θdA
*
θ b (z *
)= (5.21)
∫∫ u dA
*
Local Sherwood number
k L Dh
ShL = (5.22)
Dva
where kL is local mass transfer coefficient (m/s). As will be discussed later, the local
Sherwood number decreases asymptotically from a very high value near the entrance of a
tube to the fully developed value Shω at the end of mass entry length. It can be used to
calculate the local convective mass transfer coefficient and to estimate the mass entry length
of a tube.
Considering the mass balance in a control volume of length Δz, mass transferred through
convection
Q 1 = k L ρPeΔz (ω W − ω b ) (5.23)
Mass change of vapor in the fluid in the control volume
Q 2 = ρu m As c p Δω b (5.24)
82 Li-Zhi Zhang
Since
Q1 = Q 2 (5.25)
and
ρu m Dh2
Δz = Δz * Dh ReSc = Δz * (5.26)
Dva
thus
1 Δθ b 1 Dva Δθ b
kL = − ρu m Dh =− (5.27)
4θ b Δz 4θ b Dh Δz *
Substituting above equation to Eq.(5.22), and considering the control volume to be

infinitely small, then we obtain the local Sherwood number as
1 dθ b
ShL = − (5.28)
4θ b dz *
Average Sherwood number from 0 to z*
1 z*
Shm =
z* ∫0
Nu L dz * (5.29)
Substituting Eq.(5.28) to (5.29), it is obtained
1
Shm = − ln θ b (5.30)
4z*
Boundary Conditions
The flow is assumed hydrodynamically developed and mass developing. This means that
the cross-sectional velocity field doesn’t change with tube length, while the humidity fields
vary with tube length. For the honeycomb type desiccant wheels, strictly speaking, the tube
wall is neither an ideal uniform humidity nor an ideal uniform mass flux boundary condition.
However, the humidity difference on the wall is relatively small, compared with fluid
humidity variations [5]. Therefore, a uniform wall humidity boundary condition (ω) is
considered (ωw=const). In other words,
u*=0, θ=0 on the wall of the duct (5.31)

Inlet Condition
θ=1, at z*=0 (5.32)
In heat transfer, Nusselt number under uniform temperature conditions is denoted as NuT.
Another boundary condition often considered is the constant heat flux (H) boundary
condition. For a heat exchanger with highly conductive materials (e.g., copper, aluminum),
the H condition may apply. In practice, it may be difficult to achieve this boundary condition
for noncircular ducts, as discussed in the work of Shah and London [4]. It is already known
that NuH is higher than NuT for all duct geometries. For sinusoidal ducts, NuH is
approximately 30% higher than NuT. In mass transfer, Sherwood number under uniform
humidity boundary condition is denoted as Shω. The Sherwood number under uniform mass
flux boundary condition is denoted as ShH.
5.3. BOUNDARY FITTED COORDINATES

The difficulty with the complex nature of the duct shape may be circumvented by a
numerically generated coordinate system. The basic idea of the boundary fitted coordinate
system is to have a coordinate system such that the body contour coincides with the
coordinate lines. One of the methods often used to accomplish this goal was suggested by
Thompson et al [13] and Thomas and Middlecoff [14]. The transformation between the
physical coordinates (x, y) and the boundary fitted coordinates (ξ, η), which is usually a
square domain, is achieved by solving two Poisson equations on (x, y) domain, namely
∂ 2ξ ∂ 2ξ
+ = P (ξ ,η ) (5.33)
∂x 2 ∂y 2
∂ 2η ∂ 2η
+ = Q (ξ ,η ) (5.34)
∂x 2 ∂y 2
where P(ξ, η) and Q(ξ, η) are grid distribution inhomogeneous functions in the computational
domain. These two equations may be more easily solved on the computational plane.
Therefore, Thompson et al [13] inverted Eqs.(5.33) and (5.34) into the transformed domain
(ξ, η), where the boundary is easy to specify. At the same time, using the method proposed by
Thomas and Middlecoff [14] for selecting P, Q, Eqs. (5.33) and (5.34) are inverted into
⎡ ∂2x ∂x ⎤ ∂2 x ⎡ ∂2x ∂x ⎤
α ⎢ 2 + φ ⎥ − 2β + γ ⎢ 2 +ψ ⎥=0 (5.35)
⎣ ∂ξ ∂ξ ⎦ ∂ξ∂η ⎣ ∂η ∂η ⎦
84 Li-Zhi Zhang
⎡∂2 y ∂y ⎤ ∂2 y ⎡ ∂2 y ∂y ⎤
α ⎢ 2 + φ ⎥ − 2β + γ ⎢ 2 +ψ ⎥=0 (5.36)
⎣ ∂ξ ∂ξ ⎦ ∂ξ∂η ⎣ ∂η ∂η ⎦
where
2 2
⎡ ∂x ⎤ ⎡ ∂y ⎤
α =⎢ ⎥ +⎢ ⎥ (5.37)
⎣ ∂η ⎦ ⎣ ∂η ⎦
∂x ∂x ∂y ∂y
β= + (5.38)
∂ξ ∂η ∂ξ ∂η
2 2
⎡ ∂x ⎤ ⎡ ∂y ⎤
γ =⎢ ⎥ +⎢ ⎥ (5.39)
⎣ ∂ξ ⎦ ⎣ ∂ξ ⎦
⎡ ∂x ∂ 2 x ∂y ∂ 2 y ⎤
⎢ + ⎥
⎣ ∂ξ ∂ξ ∂ξ ∂ξ 2 ⎦
2
φ =− 2 2
(5.40)
⎡ ∂x ⎤ ⎡ ∂y ⎤
⎢ ∂ξ ⎥ + ⎢ ∂ξ ⎥
⎣ ⎦ ⎣ ⎦
⎡ ∂y ∂ 2 y ∂x ∂ 2 x ⎤
⎢ + ⎥
⎣ ∂η ∂η ∂η ∂η 2 ⎦
2
ψ =− 2 2
(5.41)
⎡ ∂x ⎤ ⎡ ∂y ⎤
⎢ ∂η ⎥ + ⎢ ∂η ⎥
⎣ ⎦ ⎣ ⎦
The numerical value of φ at each grid point along horizontal boundary η=ηb in terms of
boundary values x, y, is computed once the differential operators are replaced by central-
difference operators in Eq.(5.40). The values of φ at internal mesh points are computed by
linear interpolation along the vertical mesh lines ξ=const. Similarly, the numerical values of
ψ at mesh points along vertical boundary ξ=ξb are computed through central difference of
Eq.(5.41). Linear interpolations along horizontal lines η=const are performed to obtain the
values of ψ at internal grids. The procedure for evaluating the parameters φ and ψ insures that
the grid throughout the interior of the computational domain will be governed by the grid
distribution that is assigned on the boundaries, and that the transverse grid lines will be
locally orthogonal to the boundaries.
Once the values of φ and ψ are calculated, the numerical solution of Eqs.(5.35), (5.36) by
standard successive line over-relaxation (SLOR) [15] on a uniform, rectangular grid Δξ, Δη
results in a grid point distribution throughout the physical domain that is controlled entirely
by the distribution of grid points on the boundaries. The resulting grid constructions for the
corrugated ducts are shown in Figure 5.3 (a) for e>0 and (b) for e<0 respectively. The
corresponding computational domain is shown in Figure 5.3 (c). The number of nodes in the
figure is 21×21. The distribution of nodes on duct boundaries is pre-arranged to ensure the
dynamics in the corners of the duct are well reflected. As a result, in most of the cases, the
nodes on boundaries are not evenly distributed. With this technique, the unevenly distributed
nodes on boundaries can be reflected on the internal nodes.
(a)
(b)
1
η
0.5
0
0 0.5 1
ξ
(c)
Figure 5.3. Grid configurations, (a) the physical plane for e>0; (b) the physical plane for e<0; (c) the
computational domain.
86 Li-Zhi Zhang
After the set up of boundary fitted coordinates, Eqs. (5.10) and (5.11) can be transformed
to computational domain as following
∂ ⎡ 1 ⎛ ∂u * ∂u * ⎞⎤ ∂ ⎡ 1 ⎛ b ⎞ ⎛ ∂u * ∂u * ⎞⎤
2
4b 2
⎢ ⎜ α − β ⎟ ⎥ + ⎢ ⎜ ⎟ ⎜ γ − β ⎟ ⎥ + J = 0 (5.42)
∂ξ ⎣ J ⎜⎝ ∂ξ ∂η ⎟⎠⎦ ∂η ⎢⎣ J ⎝ a ⎠ ⎜⎝ ∂η ∂ξ ⎟⎠⎥⎦ Dh2
∂θ ∂ ⎡ 1 ⎛ ∂θ ∂θ ⎞⎤ ∂ ⎡ 1 ⎛ b ⎞ ⎛ ∂θ ∂θ ⎞⎤
2
JU * = ⎢ ⎜⎜α −β ⎟⎟⎥ + ⎢ ⎜ ⎟ ⎜⎜ γ −β ⎟⎥
∂ξ ⎟⎠⎥⎦
(5.43)
∂z ∂ξ ⎣ J ⎝ ∂ξ ∂η ⎠⎦ ∂η ⎢⎣ J ⎝ a ⎠ ⎝ ∂η
where α, β, γ have the same definition as in Eqs.(5.37)-(5.41), J is the Jacobian

transformation operator, which is defined as
∂ ( x, y ) ∂x ∂y ∂x ∂y
J= = − (5.44)
∂ (ξ ,η ) ∂ξ ∂η ∂η ∂ξ
Equations (5.42) and (5.43) were then discretized based on a control volume shown in
Figure 5.3. After these transformations, the governing differential equations were reduced to a
set of algebraic equation systems (see Appendix), which can be solved by ADI techniques
[15]. It is clear that iteration is needed to obtain the solution. In each direction, a tri-diagonal
matrix is solved, by treating the non-linear cross-derivative, ∂2/∂ξ∂η, as a source term and
using its values of last iteration. Although momentum equation in the finite difference form is
solved only in two directions for the determination of the velocity distribution, the energy
equation must be solved at every step change in the flow direction to determine the humidity
distribution. The proposed finite difference schemes are implicit numerical schemes with
second order accuracy. They are unconditionally stable. The following convergence criterion
was chosen for the study.
Fi ,mj ,k − Fi ,mj−,k1
∀ i, j, k,
m
≤ 10 −5 (5.45)
F i , j ,k
where F refers to the dependent variable u* or θ, respectively, m stands for the mth iteration.
Finite Difference Equations
Consider a control volume represented by node (i, j, k), Eq. (5.42) can be discretized to
⎛α ⎞ (
u i*+1, j − u i*, j ⎛ α ⎞
−⎜ ⎟
) u i*, j − u i*−1, j ( )
⎜ ⎟
⎝ J ⎠ i+ 1 , j Δξ 2 ⎝ J ⎠ i− 1 , j Δξ 2
2 2
⎛β⎞ u i*+1, j +1 + u i*, j +1 − u i*+1, j −1 − u i*, j −1

−⎜ ⎟
⎝ J ⎠i+ 1 , j 4 ΔξΔη
2
⎛β⎞ u i*−1, j +1 + u i*, j +1 − u i*−1, j −1 − u i*, j −1

+⎜ ⎟
⎝ J ⎠ i− 1 , j 4 Δξ Δη
2
⎛γ ⎞
+⎜ ⎟
* 2
(
⎛ b ⎞ u i , j +1 − u i , j ⎛ γ ⎞
*
−⎜ ⎟
)
⎛ b ⎞ u i , j − u i , j −1
2 * *
( )
⎜ ⎟ ⎜ ⎟
⎝ J ⎠ i, j + 1 ⎝ a ⎠ Δη 2 ⎝ J ⎠ i, j− 1 ⎝ a ⎠ Δη 2
2 2
⎛β⎞ ⎛ b ⎞ u i +1, j +1 + u i +1, j − u i −1, j +1 − u i −1, j

2 * * * *
−⎜ ⎟ ⎜ ⎟
⎝ J ⎠ i, j+ 1 ⎝ a ⎠ 4 Δξ Δη
2
⎛β ⎞ ⎛ b ⎞ u i +1, j −1 + u i +1, j − u i −1, j −1 − u i −1, j 4b

* 2 * * * 2
+⎜ ⎟ ⎜ ⎟ + 2 J i, j = 0
⎝ J ⎠ i, j − 1 ⎝ a ⎠ 4 ΔξΔη Dh (5.46)
2
Equation (5.43) can be discretized to
θi, j ,k −θ i, j ,k −1 ⎛ α ⎞ (θ − θ i , j ,k ) ⎛ α ⎞ (θi, j,k −θi−1, j,k )

(JU )i, j =⎜ ⎟
i +1, j ,k
−⎜ ⎟
Δz *
⎝ J ⎠i + 1 , j Δξ 2
⎝ J ⎠i − 1 , j Δξ 2
2 2
⎛β⎞ θ i+1, j+1,k +θ i, j +1,k −θ i+1, j −1,k −θi, j −1,k

−⎜ ⎟
⎝ J ⎠i + 1 , j 4ΔξΔη
2
⎛β⎞ θ i−1, j +1,k +θ i, j+1,k −θ i−1, j −1,k −θi, j −1,k

+⎜ ⎟
⎝ J ⎠i − 1 , j 4ΔξΔη
2
⎛ γ ⎞ ⎛ b ⎞ (θ i, j +1,k −θ i, j ,k ) ⎛ γ ⎞ ⎛ b ⎞ (θi, j ,k −θi, j −1,k )

2 2
+⎜ ⎟ ⎜ ⎟ −⎜ ⎟ ⎜ ⎟
⎝ J ⎠i , j + 1 ⎝ a ⎠ Δη 2 ⎝ J ⎠i , j − 1 ⎝ a ⎠ Δη 2
2 2
⎛ β ⎞ ⎛ b ⎞ θi+1, j +1,k +θ i+1, j ,k −θ i−1, j +1,k −θ i−1, j ,k

2
−⎜ ⎟ ⎜ ⎟
⎝ J ⎠i , j + 1 ⎝ a ⎠ 4ΔξΔη
2
⎛ β ⎞ ⎛ b ⎞ θi+1, j −1,k + θ i+1, j ,k −θ i−1, j−1,k −θ i−1, j ,k

2
+⎜ ⎟ ⎜ ⎟
⎝ J ⎠i , j − 1 ⎝ a ⎠ 4ΔξΔη (5.47)
2
88 Li-Zhi Zhang
where
2 2
⎛ xi , j +1 − xi , j −1 ⎞ ⎛ yi , j +1 − yi , j −1 ⎞
α i, j = ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟ (5.48)
⎝ 2Δη ⎠ ⎝ 2Δη ⎠
⎛ xi+1, j − xi−1, j ⎞ ⎛ xi, j +1 − xi, j −1 ⎞ ⎛ yi+1, j − yi−1, j ⎞ ⎛ yi, j +1 − yi, j −1 ⎞

β i, j = ⎜⎜ ⎟⎟ ⋅ ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟ ⋅ ⎜⎜ ⎟⎟ (5.49)
⎝ 2Δξ ⎠ ⎝ 2Δη ⎠ ⎝ 2Δξ ⎠ ⎝ 2Δη ⎠
2 2
⎛ xi +1, j − xi −1, j ⎞ ⎛ yi +1, j − yi −1, j ⎞
γ i, j = ⎜⎜ ⎟⎟ + ⎜⎜ ⎟⎟ (5.50)
⎝ 2Δξ ⎠ ⎝ 2Δξ ⎠
⎛ xi +1, j − xi −1, j ⎞ ⎛ yi , j +1 − yi , j −1 ⎞ ⎛ xi , j +1 − xi , j −1 ⎞ ⎛ yi +1, j − yi −1, j ⎞

J i , j = ⎜⎜ ⎟⎟ ⋅ ⎜⎜ ⎟⎟ − ⎜⎜ ⎟⎟ ⋅ ⎜⎜ ⎟⎟
⎝ 2Δξ ⎠ ⎝ 2Δη ⎠ ⎝ 2Δη ⎠ ⎝ 2Δξ ⎠
(5.51)
i-1, j+1 i, j+1 i+1, j+1
i, j i+1, j
i-1, j
ξ
i-1, j-1 i, j-1 i+1, j-1

(a)
i-1, j-1 i, j-1 i+1, j-1
i, j i+1, j
i-1, j
x*
i-1, j+1 i, j+1 i+1, j+1
y*
(b)
Figure 5.4. Grid numbering. (a) computational domain; (b) physical domain.
In equations (5.46) to (5.51), the subscripts (i, j, k) represent the ξ, η and z* directions in
the computational domain. The grid nodes (i, j) in the computational domain (ξ, η) and in the
physical domain are presented in Figure 5.4. The values at the boundaries of a control volume
are obtained by interpolation of values at nodes, for example, Ji, j+1/2 are calculated from Ji, j
and Ji, j+1.
5.4. FRICTION AND MASS TRANSFER COEFFICIENTS
Validation of the Procedure
To assure the accuracy of the results presented, numerical tests were performed for the
duct to determine the effects of the grid size. It indicates that 21×21 grids on cross section and
Δz*=0.00035 axially are adequate (less than 0.1% difference compared with 31×31 grids and
Δz*=0.00025). For hydrodynamically fully developed laminar flow in ducts, (fRe) is a
constant and in the mass entry region, the local Sherwood number will decrease and approach
asymptotically to a lower limiting value Shω with the marching of flow.
Mass transfer coefficients for various cross sectional shapes are scarce. To validate the
procedure, (fRe) and NuT for some ducts are calculated and compared with the results found
in references. It should be known that the dimensionless mass conservation equation is the
same form with heat conservation equation. Therefore heat transfer equation is first used to
validate the model. The comparisons are listed in Table 5.1.
From this table, it can be concluded that maximum errors are less than 0.8% for (fRe) and
less than 0.9% for NuT.
Table 5.1. Comparisons of (f⋅Re) and NuT of fully developed laminar flow for some ducts
from present case and those from literature
Shape τ (f⋅Re) NuT
Present Refs[4,5] Error Present Refs[4,5] Error

case (%) case (%)
Circular 16.151 16 0.51 3.692 3.657 0.41
Square 14.225 14.227 0.78 3.061 2.976 0.84
Elliptic 0.5 16.931 16.823 0.70 3.676 3.742 0.64
Equilateral Triangular 13.326 13.321 0.11 2.496 2.46 0.65
Isosceles Triangular 0.5 13.147 13.153 0.18 2.331 2.34 0.90
Sine 2.0 14.476 14.553 0.16 2.658 Unavailable
Sine 1.5 13.964 14.022 0.63 2.614 2.6 0.54
Sine 1.0 12.912 13.023 0.77 2.463 2.45 0.53
Sine 0.75 12.326 12.234 0.75 2.317 2.33 0.56
Sine 0.50 11.163 11.207 0.30 2.155 2.12 0.71
Notes: τ, aspect ratio (duct height/duct width); for the Sine ducts listed: e=0 (flat Sinusoidal).
90 Li-Zhi Zhang
Effects of Bending Ratios
For e=0, the corrugated ducts reduce to sine ducts with flat lower boundaries whose
results are listed in Table 5.1. The values of fRe and NuT are in excellent agreement with the
published data. The code can be used to calculate mass transfer coefficients. In practical
desiccant wheels, e would be greater or less than 0, especially when the ducts are in zones of
small diameters. For these ducts, the friction and mass transfer coefficients are affected by
bending ratio e. The values of (f⋅Re) and Shω for various bending ratios and aspect ratios are
listed in Table 5.2.
Table 5.2 shows that the greater the e, the smaller the (f⋅Re) and the Shω. This character
in return discloses that the higher the friction coefficients, the higher the Sherwood numbers.
Generally speaking, e>0 has positive effects on friction coefficients, but negative effects on
Sherwood numbers. On the other hand, e<0 would increase friction coefficients and
Sherwood numbers simultaneously, compared to ducts with flat lower boundaries. These
phenomena are attributed to the fact that the bigger the e, the larger the dead spaces in the
corners of ducts. The larger the dead zones, the more inefficient of the transfer area, which
results in decreased friction coefficients and Sherwood numbers.
Table 5.2. Values of (f⋅Re) and Shω with various aspect and bending ratios
τ=2.0
e -0.5 -0.25 0 0.25 0.5
(f⋅Re) 15.19 15.146 14.576 13.308 11.635
Shω 2.715 2.675 2.558 2.289 2.002
τ=1.0
e -0.5 -0.25 0 0.25 0.5
(f⋅Re) 14.738 14.209 12.922 11.281 8.298
Shω 2.966 2.735 2.363 1.895 1.228
τ=0.75
e -0.5 -0.25 0 0.25 0.5
(f⋅Re) 15.027 13.945 12.326 10.064 7.09
Shω 2.981 2.652 2.217 1.562 0.867
τ=0.65
e -0.5 -0.25 0 0.25 0.45
(f⋅Re) 15.366 13.975 12.117 9.701 6.931
Shω 3.008 2.605 2.071 1.4 0.778
τ=0.5
e -0.5 -0.25 0 0.25 0.35
(f⋅Re) 16.803 14.477 11.173 9.206 7.485
Shω 3.105 2.522 2.035 1077 0.757
τ=0.4
e -0.5 -0.25 0 0.2
(f⋅Re) 15.717 12.819 10.221 8.291
Shω 3.232 2.473 1.783 1.03
The effects of e on (fRe) are shown in Figure 5.5. and the effects on Shω are shown in
Figure 5.6. Figure 5.5 shows the variations of (fRe)/(fRe)0 (where subscript 0 means e=0) with
various e for different aspect ratios. It is seen that the greater the e, the smaller the (fRe). The
ratio can be as low as 0.5 when e=0.5 and as high as 1.5 when e=-0.5.
τ=2.0
1.5 τ=1.0
(f Re)/(f Re)0
τ=0.75
τ=0.4
1
0.5
-0.5 -0.25 0 0.25 0.5
e
Figure 5.5. Effects of duct bending ratios on friction coefficients.
τ=2.0
1.5 τ=1.0
τ=0.75
T0
/Shω0
τ=0.4
ShTω/Nu
1
Nu
0.5
0
-0.5 -0.25 0 0.25 0.5
ee
Figure 5.6. Effects of duct bending ratios on Sherwood numbers.
92 Li-Zhi Zhang
1. .25 5 5
1 0.7 0.2
0.75
75
2.25
0.02.751.25
5
y* 0.5
1.75
0 .2 5
1.2
5
0.25
0.25
0.75
5
0 .2
0
0 0.25 0.5 0.75 1
x*
(a) τ=0.5, e=0.14.
1
0.
25
0.75 75 5
0. 1.2
5
75
0.2
0.5
1.
2. 0.7 1.25
25
y*
5
0.2
1.75
0.25
5
5
0.2
1.
25
0
0.75
0.25
-0.25
0 0.25 0.5 0.75 1
x*
(b) τ=0.5, e=-0.14.
Figure 5.7. Fully developed velocity profile. The isoclines are lines of constant u*/u*m.
Figure 5.6 shows the variations of Shω/ Shω0 (ratio of the Sherwood number for bending
duct to that for e=0 with the same aspect ratio) for different aspect ratios. The smaller the
bending ratio, the higher the Sherwood number. The comparisons of Figures 5.5 and 5.6 also
disclose that the higher the friction coefficients, the higher the Sherwood numbers. Generally
speaking, e>0 has positive effects on friction coefficients, but negative effects on Sherwood
numbers. On the other hand, e<0 would increase friction coefficients and Sherwood numbers
simultaneously, compared to ducts with flat lower boundaries. To know the reason why, let’s
plot velocity fields in Figure 5.7 and humidity fields in Figure 5.8. The comparisons of
Figures 5.7(a), 5.7(b) and 5.8(a), 5.8(b) show that the greater the e, the larger the dead spaces
in the corners. The larger the dead zones, the more inefficient of the transfer area, which
results in decreased friction coefficients and Sherwood numbers. It is interesting to note that
the shapes of isotherms and iso-velocities are very similar to those of triangular ducts,
however, the maximum velocity and humidity occur closer to the center than those of the
triangular ducts [4] for same aspect ratios (duct height to width ratio). The shapes of isolines
change from triangle with round corners near the boundary to circles at the center gradually.
Furthermore, both the velocity gradients and the humidity gradients have their highest values
near straight boundaries, while minima in the corners. These phenomena are also in
agreements with the holographic interferometric observations of humidity fields in such ducts
[16].
0.3
0
0.1
0.75 0.5
0
0
0
0.1
y*
0
0.5
0.30
0.7
0.50
30
0.
0
0.1
0.25
0
0 0.25 0.5 0.75 1
x*
(a) e=0.14
1
0.75
0.
0.
30
10
0
0 .1
0.5
y*
0.50
0.25
0.30
0
0.5
0.10
0.30 10
0 0.
-0.25
0 0.25 0.5 0.75 1
x*
(b) e=-0.14
Figure 5.8. Equal concentration curves for τ=0.5, at z*=0.1. The values are dimensionless humidity θ.
94 Li-Zhi Zhang
Local and Mean Sherwood Numbers
The axial variations of bulk humidity are shown in Figure 5.9 for τ=1.0. Inspection of the
curves in this figure reveals that the bulk humidity is strongly variant with e at the entrance
region of the duct. However, as the air passes through the duct, the bulk humidity is
dependent on the value of e. Positive e results in a bigger bulk humidity than a negative e at
the same z* position, suggesting a decreased mass transfer rate. Besides, a bigger bulk
humidity would finally lead to a longer thermal entry length.
Local Sherwood numbers against z* for τ=1.0 are presented in Figure 5.10. It is seen that
ShL decreases from a high value near the entrance to the fully developed value Shω at a
greater axial distance. This figure also illustrates that the local Sherwood number decreases
dramatically at the entrance region of the duct as the bending ratio increases. However, this
variation decreases gradually until it reaches the asymptotic limiting value. Generally
speaking, positive e has greater impacts than negative e do.
Figure 5.11 shows the variations of mean Sherwood numbers along the flow. The mean
Sherwood numbers are higher than the local values at the same position. However, as z*
increases, the mean values will decrease gradually to the limit values of Shω. The effects of
bending ratio e on the mean Sherwood numbers are very similar to those on the local values.
The variations of bulk humidity, the local and mean Sherwood numbers along the tube
length for other aspect ratios are similar in shape to those for τ=1.0. The differences lie in the
length of the thermal entry region. The sharper the corner of a duct, the longer is the tube
from the entrance to the fully developed point. Most importantly, duct shape mainly
influences the cross-sectional velocity and humidity profiles in the duct, as shown in Figures
5.7 and 5.8. It is disclosed that the sharper the corner, the larger the dead space, the smaller
the Sherwood number.
0.8
0.6
0.4
e=-0.25
0.2 e=0.0
e=0.25
0 -4 -3 -2 -1 0
10 10 10 10 10
z*
Figure 5.9. Axial variations of bulk humidity, τ=1.0.

20
e=-0.25
e=0.0
15
e=0.25
NuL
ShL 10
0 -4
10 10-3 10-2 10-1
z*
Figure 5.10. Axial variations of local Sherwood numbers, τ=1.0.
50
e=-0.25
40 e=0.0
e=0.25
30
Num
Shm
20
10
0 -4
10 10-3 10-2 10-1
z*
Figure 5.11. Axial variations of mean Sherwood numbers, τ=1.0.

96 Li-Zhi Zhang
Besides bended sinusoidal cross sectional ducts, there are ducts of other cross sections.
The friction coefficients and Sherwood numbers of developed flow in other cross sectional
shapes, under uniform humidity conditions and uniform mass flux boundary conditions cab
be calculated.
5.5. CONCLUSION
Convective mass transfer and fluid flow in corrugated ducts confined by sinusoidal and
arc curves are analyzed numerically for various combinations of aspect and bending ratios
with uniform wall humidity conditions. The boundary-fitted coordinate is used to solve the
difficulty induced by the complex physical domain. The velocity and humidity fields are
calculated and graphically illustrated to investigate the effects of sharp corners in the ducts. It
is found that bending ratio, e, has a great influence on both the friction coefficients and the
Sherwood numbers. The product of friction coefficient and Reynolds number (fRe) could
drop 50% when e=0.5 and rise 50% when e=-0.5, compared to ducts with flat lower
boundaries. Positive e would decrease the Sherwood number and negative e would increase
the Sherwood number significantly. Besides, bending ratios other than zero could also affects
the thermal entry length, local Sherwood numbers, and humidity/velocity profile shapes. All
these are due to the fact that the greater the bending ratios, the larger the dead spaces, for both
the fluid flow and the mass transfer. The results can be used to analyze the pressure drop and
mass transfer properties in rotary energy wheels.
REFERENCES
[1] Jin, W.; Kodama, A.; Goto, M.; Hirose, T. An adsorptive desiccant cooling using
honeycomb rotor dehumidifier. Journal of Chemical Engineering of Japan, 1998, 31,
706-713.
[2] Zheng, W.; Worek, W.M. and Novosel, D. Performance optimization of rotary
dehumidifiers. ASME Journal of Solar Energy Engineering, 1995, 117, 40-44.
[3] Kodama, A.; Goto, M.; Hirose, T.; Kuma, T. Experimental study of optimal operation
for a honeycomb adsorber operated with thermal swing. Journal of Chemical
Engineering of Japan, 1993, 26, 530-535.
[4] Shah, R.K.; London, A.L. Laminar flow forced convection in ducts. New York:
Academic Press Inc.; 1978.
[5] Sherony, D.F.; Solbrig, C.W. Analytical investigation of heat or mass transfer and
friction factors in a corrugated duct heat or mass exchanger. International Journal of
Heat Mass Transfer, 1970, 13, 145-159.
[6] Fischer, L.; Martin, H. Friction factors for fully developed laminar flow in ducts
confined by corrugated parallel walls. International Journal of Heat Mass Transfer,
1997, 40, 635-639.
[7] Ebadian, M.A.; Zhang, H.Y. Fluid flow and heat transfer in the crescent-shaped lumen
catheter. ASME Journal of Applied Mechanics, 1993, 60, 721-727.
[8] Dong, Z.F.; Ebadian, M.A. A numerical analysis of thermally developing flow in elliptic
ducts with internal fins. International Journal of Heat and Fluid Flow, 1991, 12, 166-
172.
[9] Niu, J.L.; Zhang, L.Z. Heat transfer and friction coefficients in corrugated ducts
confined by sinusoidal and arc curves. International Journal of Heat Mass Transfer,
2002, 45, 571-578.
[10] Zhang, L.Z.; Niu, J.L. A numerical study of laminar forced convection in sinusoidal
ducts with arc lower boundaries under uniform wall temperature. Numerical Heat
Transfer, Part A: Applications, 2001, 40, 55-72.
[11] Shah, R.K. Laminar flow friction and forced convection heat transfer in ducts of
arbitrary geometry. International Journal of Heat Mass Transfer, 1975, 18, 849-862.
[12] Kays, W.M. and Crawford, M.E. Convective Heat and Mass Transfer. New York:
McGraw-Hill, Inc.; 1993.
[13] Thompson, J.F.; Thames, F.; Martin, C. Automatic numerical generation of body-filled
curvilinear coordinate system for field containing any number of arbitrary two-
dimensional bodies. Journal of Computational Physics, 1974, 24, 299-319.
[14] Thomas, P. D.; Middlecoff, J.F. Direct control of grid point distribution in meshes
generated by elliptic equations. AIAA Journal, 1982, 18, 652-656.
[15] Samarskii, A.A.; Vabishchevich, P.N. Computational Heat Transfer. New York: John
Wiley & Sons Inc.; 1995.
[16] Tauscher, R.; Dinglreiter, U.; Durst, B.; and Mayinger, F. Transport processes in
narrow channels with application to rotary exchangers. Heat and Mass Transfer, 1999,
35, 123-131.
Chapter 6
CONVECTIVE HEAT MASS TRANSFER

IN PLATE-FIN CHANNELS
ABSTRACT
Plate-fin structure is the most common structure for stationary total heat recovery.
Common traditional air-to-air sensible only heat exchangers use well conductive metal
foils as the heat transfer media. For total heat exchanger, novel water-permeable
materials are used to simultaneously permeate heat and moisture. They are less-
conductive materials both for heat transfer and mass transfer. In this chapter, the
influences of finite fin conductance on heat and mass transfer in plate-fin channels were
investigated.
NOMENCLATURE
a Half duct height (m)
Ac Cross-sectional area (m2)
At Transfer area (m2)
b Half duct width (m)
cP Specific heat (kJkg-1K-1)
Dh Hydraulic diameter (m)
Dva Vapor Diffusivity in air (m2/s)
Dwf Vater Diffusivity in fin material (m2/s)
j Chilton-Colburn j factor
k Mass transfer coefficient (m/s)
kp Partition coefficient
L Length (m)
Le Lewis number
Nu Nusselt number
NuT Nusselt number for thermally fully developed laminar flow with T condition
100 Li-Zhi Zhang
P Pressure (Pa)
Pf Perimeter of duct (m)
Pr Prandtl number
q Heat flux (kWm-2)
Q Total heat transfer (kW)
Re Reynolds number
RH Relative humidity
s Tangent coordinate for fin (m)
Sc Schmidt number
Sh Sherwood number
Shω Fully developed Sherwood number under uniform concentration condition
St Stanton number
T Temperature (K)
u Velocity (ms-1)
U Velocity coefficient
W Water uptake (kg moisture/kg material)
x, y Dimensional transversal coordinates (m)
yf Perpendicular coordinate for fin (m)
z Axial coordinate (m)
Greek Symbols
δ Fin thickness (m)

ρ Density (kgm-3)
θ Dimensionless temperature
μ Dynamic viscosity (kgm-1s-1)
Ω Conductance parameter
ψ Correction factor of temperature or humidity difference for cross flow
λ Heat conductivity (kWm-1K-1)
ηfin Fin efficiency
ξ Dimensionless humidity
ω Humidity ratio (kg vapor/kg dry air)
τ Aspect ratio
α Half apex angle
Superscripts
• Dimensionless
Subscripts
a Air
b Bulk
e Exhaust air
f Fin, fresh air
Convective Heat Mass Transfer in Plate-fin Channels 101
i Inlet
L Local, lower surface, moisture (latent heat)
m Mean
s Sensible heat
tot Total
u Upper surface
w Wall
6.1. INTRODUCTION
Energy wheels have the ability to recover both sensible heat and moisture from
ventilation air. However, there are some inherent shortcomings with desiccant wheels. They
have moving parts. The wheel rotates alternately between the fresh air and the exhaust air,
which leads to crossovers between the fresh air and exhaust air. Therefore it’s inevitable that
the fresh air will be polluted by the exhaust air. That fails to meet the requirements for
ventilation. In addition, desiccant wheels are very expensive and their maintenance is
difficult. The long-term reliability is questionable because of the cyclic nature of the system.
The wheel material’s ability to adsorb moisture deteriorates steadily with time. These factors
have restricted their developments in real applications.
In contrast, the stationary total heat exchangers are superior in that they are used in
steady-state adsorption and permeation status. They have no moving parts and the reliability
is quite high. The duct sealing is easy and the crossovers can be prevented. The difference
between a total heat exchanger and a common sensible-only heat exchanger is that water-
permeable plates are used in stead of metal plates. Currently there are two types of vapor-
permeable materials: paper and hydrophilic polymer membranes. Paper is cheap, but its
moisture efficiency is lower than 0.4. Membranes have higher latent effectiveness above 0.65.
The shortcomings are that they are expensive than paper. Therefore there are many researches
now to develop new membranes from cheap materials. Whether it’s paper or other vapor-
permeable material, heat and moisture transfer in the enthalpy exchangers are of interest.
As other air-to-air heat exchangers, plate-fin channels are the most popular structure for
total heat exchangers. Figure 6.1 shows a real photo of two plate-fin heat exchangers. Figure
6.2 shows the schematic of the plate-fin heat exchanger. In the figure, the cross-section for a
single channel is sinusoidal. Other popular geometries include triangular, rectangular, etc.
Usually a cross flow between fresh air and exhaust air is used, due to the convenience in duct
sealing.
The reason why plate-fin ducts are selected as the basic structure lies in several facts: (1)
They are easy to construct. Especially the corrugated fins are easy to manufacture in large
scale by machines; (2) The mechanical strength is very high even with very thin plates; (3)
Heat transfer intensification is needed on both sides of a plate, therefore equal structure and
area are realized on both sides; (4) Duct sealing are easy to realize because of the multi-points
contact between a fin and a neighboring plate. (5) Packing density is rather high because the
channel height can be very small (1-2mm) as a result of plates separating by a corrugated fin.
102 Li-Zhi Zhang
(a) Sinusoidal duct.
(b) Rectangular duct.
Figure 6.1. A photo of two plate-fin heat exchangers.

Figure 6.2. A schematic of a sinusoidal plate-fin heat exchanger.
6.2. SINUSOIDAL DUCTS OF FINITE FIN CONDUCTANCE

Heat and mass transfer in plate-fin total heat exchangers can be separated into three steps:
(1) convective heat mass transfer from fresh air to plates and fins; (2) conductive heat and
mass transfer in plates; (3) convective heat mass transfer from plates and fins to exhaust air.
Of the heat mass transfer parameters, convective heat mass transfer coefficients in plate-fin
duct are the key parameter to analyze performance and product design.
Convective Plate-fin ducts with traditional well-conductive metal walls have been studied
by many authors. A comprehensive review of the theoretical and experimental studies on
fully developed forced convection and heat transfer in ducts of various cross sections up to
the 1970s had been conducted by Shah and London [1], and documented in several well
known references [1-3]. The results have been regarded as the basic data for heat exchanger
design for these years. Strictly speaking, plate-fin ducts are different from common ducts of
uniform wall temperature. In a heat exchanger, plates contact with air streams directly, but
fins don’t. Heat is transferred by conduction from plates to fins first, then it is transferred
from fins to air stream by convection. It should be noted that real fins have limited
conductance, which may comprise fin efficiencies. The effects of finite fin conductance on
heat transfer have been investigated by several authors [4]. However, nowadays, with the
application of many new heat transfer materials like paper and polymers to total heat
recovery, the effects of finite fin conductance are very large. The fin conductance of such
materials is very low.
The problem aforementioned will be addressed in this chapter. The difference between
this chapter and previous studies is that beside heat transfer, mass transfer in fins will be
considered either. To overcome the non-rectangular nature of the duct cross section, as
before, a boundary-fitted coordinate technique will be used to transform the physical domain
to a computational domain. Another benefit with this technique is that the program written
can be easily modified to study ducts of other cross sectional shapes. Sinusoidal plate-fin duct
will first be considered.
104 Li-Zhi Zhang
y Flow
D
z
2a y1f s L
A C x
2b
Figure 6.3. Geometry of a plate-fin sinusoidal duct.
Governing Equations
The problem considered here is that of a representing duct shown in Figure 6.3. The
geometries of the sinusoidal duct are also depicted in the figure: height 2a; width 2b; duct
length, L. Fin curve can be expressed as a sinusoidal function
⎡ ⎛ π ⎞⎤
y = a ⎢1 − cos⎜ x ⎟⎥ (6.1)
⎣ ⎝ b ⎠⎦
Aspect ratio of duct
2a
τ= (6.2)
2b
Duct Area
2b
Ac = ∫ ydx (6.3)
0
Length of each fin, AD or DC
b
Lf = ∫ (dx) 2 + (dy ) 2 (6.4)
x =0
Perimeter of duct
Pf = 2b + 2 Lf (6.5)
Hydraulic diameter
4 Ac
Dh = (6.6)
Pf
where Ac is the cross section area of the duct (m2), Pf is the perimeter of the duct (m). The
flow in the duct is considered to be laminar and hydrodynamically fully developed, but
thermally developing in the entrance region of the duct. The fluid is Newtonian with constant
thermal properties. Additionally, a uniform wall temperature boundary condition is
considered for the plate (AB and BC). The Nusselt numbers under uniform heat flux
boundary conditions are usually 20-30% higher than those under uniform temperature
conditions [1,2].
The governing equations are summarized as following [5,6]. Momentum
⎛ ∂ 2 u ∂ 2 u ⎞ dP
μ⎜⎜ 2 + 2 ⎟⎟ = (6.7)
⎝ ∂x ∂y ⎠ dz
where μ is dynamic viscosity (Pa.s), u is fluid velocity (m/s), P is the pressure (Pa), z is the
axial coordinate (m), P is pressure (Pa).
Energy conservation
∂T ⎛ ∂ 2T ∂ 2T ⎞
ρc P u = λ a ⎜⎜ 2 + 2 ⎟⎟ (6.8)
∂z ⎝ ∂x ∂y ⎠
where T is fluid temperature (K), λa is thermal conductivity of air (kWm-1K-1), ρ is density

(kg/m3), cp is specific heat (kJkg-1K-1).
The above two governing equations can be normalized to
∂ 2u * ⎛b⎞ ∂ u
22 *
4b 2
+⎜ ⎟ + =0 (6.9)
⎝ a ⎠ ∂y *
2 2
∂x * Dh2
∂θ ∂ 2θ ⎛ b ⎞ ∂ 2θ
2
U * = +⎜ ⎟ (6.10)
∂z ∂x *
2
⎝ a ⎠ ∂y *
2
with a dimensionless velocity
μu
u* = − (6.11)
(dP / dz ) Dh2
and a dimensionless temperature

106 Li-Zhi Zhang
T − Tw
θ= (6.12)
Ti − Tw
where in the equations, Ti is the inlet temperature of the fluid, and Tw is the wall temperature.
Dimensionless coordinates are defined by
x
x* = (6.13)
2b
y
y* = (6.14)
2a
z
z* = (6.15)
Dh RePr
where Re is Reynolds number and Pr is Prandtl number.

In Eq.(6.10), velocity coefficient U is defined by
u * 4b 2
U= (6.16)
u m* Dh2
∫∫ u dA
*
u *
m = (6.17)
Ac
The characteristics of fluid flow in the duct can be represented by the product of the
friction coefficient and the Reynolds number as
⎛ dP ⎞
⎜ Dh ⎟
( fRe) = ⎜ − dz ⎟⎛⎜ ρu m Dh ⎞ 1
⎟⎟ = *
⎜ 2ρu m2 ⎟⎜⎝ μ
(6.18)
⎜ ⎟ ⎠ 2u m
⎝ ⎠
Dimensionless bulk temperature
∫∫ u θdA
*
θb = (6.19)
∫∫ u dA
*
Nusselt number
hDh
Nu = (6.20)
λa
where h is convective heat transfer coefficient (kWm-2K-1) between fluid and wall.
An energy balance in a control volume in the duct [5] will give the equation for
estimation of the local Nusselt number as
1 dθ b
Nu L = − (6.21)
4θ b dz *
and the mean Nusselt number from z*=0 to z* by
1
Nu m = − ln θ b (6.22)
4z *
Fins
The coordinate system for two fins (AD and DC) is s, y1, and shown in Figure 6.4. Axis s
is tangent to fin surface and y1 is normal to fin surface. The directions vary from point to
point. At any location along the fin, there is a balance between the net conduction along the
fin and the heat transfer from the surface of the fin to the fluid. Heat transfer in fin is
governed by the following one-dimensional model [4,6]
d 2Tf
λfδ = qu + qL (6.23)
ds 2
⎛ ∂T ⎞
q u = − λ a ⎜⎜ ⎟⎟ (6.24)
⎝ ∂y f ⎠u
⎛ ∂T ⎞
q L = λ a ⎜⎜ ⎟⎟ (6.25)
⎝ ∂y f ⎠L
where (∂T / ∂y f ) is the normal gradient of fluid temperature on the lower or upper surface of
fin. The heat flux at the lower surface and the upper surface are skew symmetric, as
schematically depicted in Figure 6.4. The relation is mathematically expressed by
qu s
= qL Lf − s
(6.26)
where Lf is the length of one curved fin and it is calculated numerically by Eq.(6.4).
108 Li-Zhi Zhang
qu D
Center
qL
A
Fin
Figure 6.4. Skew-symmetric heat flux distributions on the upper and lower fin fluid interfaces.
Equations (6.23)-(6.26) can be normalized to
d 2θ f ⎛ ∂θ ⎞ ⎛ ∂θ ⎞
Ωs = ⎜⎜ * ⎟⎟ + ⎜⎜ * ⎟⎟ (6.27)
⎝ ∂y f ⎠ s* ⎝ ∂y f ⎠ L*f − s*
2
ds
where Ωs is a dimensionless parameter named fin conductance parameter for sensible heat
transfer. It is defined by
λfδ
Ωs = (6.28)
λ a (2a )
where dimensionless fin temperature
Tf − TW
θf = (6.29)
Ti − TW
s
s* = (6.30)
2a
yf
y f* = (6.31)
2a
Lf
L*f = (6.32)
2a
Table 6.1 lists the fin conductance parameters for the commonly used materials, for a
passage of duct height 2mm, fin thickness 0.1mm, and air as working fluid, calculated by
Eq.(6.28). Heat conductivity data are taken from ref.[2]. As seen, for well-conductive metal
materials like copper, bronze, iron, steel, and aluminium, the fin conductance parameters are
larger than 100; while for low conductive non-metal materials like plywood, clay, glass,
papers, and polymers, the fin conductance parameters are generally less than 5, in some cases
even lower than 1.0. In summary, large differences exist in fin conductance parameters for
various materials, from 0.2 for polymer to 760 for pure copper.
Boundary Conditions
The boundary conditions for fluid are
u*=0, on the 3 walls of the duct (6.33)
θ=0, at y*=0 (6.34)
Inlet condition
θ=1, at z*=0 (6.35)
The boundary conditions for fins, TA=TC=TD=Tw, or
θf =0, at s*=0, Lf* (6.36)
Fin-fluid coupling:
θ f = θ , at fin-fluid interfaces (6.37)
Table 6.1. Values of fin heat conductance parameter for some fin materials
Fin materials λf (Wm-1K-1) Ωs

Pure copper 401 762
Aluminium 237 450.3
Iron 80.2 152.4
Steel 60.5 115
Bronze 52 98.8
Carbon 1.6 3.04
Glass 1.4 2.66
Clay 1.3 2.47
Teflon 0.35 0.67
Paper 0.18 0.36
Wood 0.16 0.30
Polymer Membrane 0.14 0.25
Plywood 0.12 0.23
110 Li-Zhi Zhang
Numerical Method
Boundary-fitted Coordinates
Commonly, duct of rectangular cross section is easy to solve under the x-y coordinate
system. However, the cross section in this case is a sinusoidal one. To ease the solution, a
boundary fitted coordinate transformation technique is used to transfer the sinusoidal domain
to a square domain. Another benefit with this methodology is that the program can be easily
modified to calculate other ducts of arbitrary cross sectional shapes, as long as the grids
points on boundaries are specified. This will provide a broad basis for program validation.
The basic idea of the boundary fitted coordinate system is to have a coordinate system
such that the body contour coincides with the coordinate lines. The transformation between
the physical coordinates (x, y) and the boundary fitted coordinates (ξ, η), which is usually a
square domain, is achieved by solving two Poisson equations on (x, y) domain [7,8],
∂ 2ξ ∂ 2ξ
+ = P(ξ, η) (6.38)
∂x 2 ∂y 2
∂ 2η ∂ 2η
+ = Q(ξ, η) (6.39)
∂x 2 ∂y 2
where P(ξ, η) and Q(ξ, η) are grid distribution inhomogeneous functions in the computational
domain. These two equations may be more easily solved on the computational plane.
Therefore, Thompson et al [7] inverted Eqs.(6.38) and (6.39) into the transformed domain (ξ,
η), where the boundary is easy to specify. At the same time, using the method proposed by
Thomas and Middlecoff [8] for selecting P, Q, Eqs. (6.38) and (6.39) are inverted into
⎡∂2 x ∂x ⎤ ∂2x ⎡∂2x ∂x ⎤

α ⎢ 2 + φ ⎥ − 2β + γ⎢ 2 + ψ ⎥ = 0 (6.40)
⎣ ∂ξ ∂ξ ⎦ ∂ξ∂η ⎣ ∂η ∂η ⎦
⎡∂2 y ∂y ⎤ ∂2 y ⎡∂2 y ∂y ⎤
α ⎢ 2 + φ ⎥ − 2β + γ⎢ 2 + ψ ⎥ = 0 (6.41)
⎣ ∂ξ ∂ξ ⎦ ∂ξ∂η ⎣ ∂η ∂η ⎦
where
2 2
⎡ ∂x ⎤ ⎡ ∂y ⎤
α=⎢ ⎥ +⎢ ⎥ (6.42)
⎣ ∂η ⎦ ⎣ ∂η ⎦
∂x ∂x ∂y ∂y
β= + (6.43)
∂ξ ∂η ∂ξ ∂η
2 2
⎡ ∂x ⎤ ⎡ ∂y ⎤
γ=⎢ ⎥ +⎢ ⎥ (6.44)
⎣ ∂ξ ⎦ ⎣ ∂ξ ⎦
⎡ ∂x ∂ 2 x ∂y ∂ 2 y ⎤
⎢ + ⎥
⎣ ∂ξ ∂ξ 2 ∂ξ ∂ξ 2 ⎦
φ=− 2 2
(6.45)
⎡ ∂x ⎤ ⎡ ∂y ⎤
⎢ ∂ξ ⎥ + ⎢ ∂ξ ⎥
⎣ ⎦ ⎣ ⎦
⎡ ∂y ∂ 2 y ∂x ∂ 2 x ⎤
⎢ + ⎥
⎣ ∂η ∂η 2 ∂η ∂η 2 ⎦
ψ=− 2 2
(6.46)
⎡ ∂x ⎤ ⎡ ∂y ⎤
⎢ ∂η ⎥ + ⎢ ∂η ⎥
⎣ ⎦ ⎣ ⎦
The numerical value of φ at each grid point along horizontal boundary η=ηb in terms of
boundary values x, y, is computed once the differential operators are replaced by central-
difference operators in Eq.(6.45). The values of φ at internal mesh points are computed by
linear interpolation along the vertical mesh lines ξ=const. Similarly, the numerical values of
ψ at mesh points along vertical boundary ξ=ξb are computed through central difference of
Eq.(6.46). Linear interpolations along horizontal lines η=const are performed to obtain the
values of ψ at internal grids. The procedure for evaluating the parameters φ and ψ insures that
the grid throughout the interior of the computational domain will be governed by the grid
distribution that is assigned on the boundaries, and that the transverse grid lines will be
locally orthogonal to the boundaries.
Once the values of φ and ψ are calculated, the numerical solution of Eqs.(6.40), (6.41) by
standard successive line less-relaxation on a uniform, rectangular grid Δξ, Δη results in a grid
point distribution throughout the physical domain that is controlled entirely by the
distribution of grid points on the boundaries.
The resulting grid constructions for the sinusoidal duct and the corresponding
computational domain are shown in Figure 6.5.
After the set up of boundary fitted coordinates, Eqs. (6.9) and (6.10) can be transformed
to computational domain as following
∂ ⎡ 1 ⎛ ∂u * ∂u * ⎞⎤ ∂ ⎡ 1 ⎛ b ⎞ ⎛ ∂u * ∂u * ⎞⎤
2
4b 2
⎢ ⎜⎜ α −β ⎟⎟⎥ + ⎢ ⎜ ⎟ ⎜⎜ γ −β ⎟⎟⎥ + J 2 = 0 (6.47)
∂ξ ⎣ J ⎝ ∂ξ ∂η ⎠⎦ ∂η ⎣⎢ J ⎝ a ⎠ ⎝ ∂η ∂ξ ⎠⎦⎥ Dh
∂θ ∂ ⎡ 1 ⎛ ∂θ ∂θ ⎞⎤ ∂ ⎡ 1 ⎛ b ⎞ ⎛ ∂θ ∂θ ⎞⎤
2
JU * = ⎢ ⎜ α − β ⎟ ⎥ + ⎢ ⎜ ⎟ ⎜ γ − β ⎟⎥
∂ξ ⎣ J ⎜⎝ ∂ξ ∂η ⎟⎠⎦ ∂η ⎣⎢ J ⎝ a ⎠ ⎜⎝ ∂η ∂ξ ⎟⎠⎦⎥
(6.48)
∂z
112 Li-Zhi Zhang
where α, β, γ have the same definition as in Eqs.(6.42)-(6.44), J is the Jacobian

transformation operator, which is defined as
∂ ( x, y ) ∂x ∂y ∂x ∂y
J= = − (6.49)
∂ (ξ, η) ∂ξ ∂η ∂η ∂ξ
D
•
A• • •C
B
(a) The physical plane.
D
1 C
0.75
0.5
η
0.25
0 B
A 0 0.25 0.5 0.75 1
ξ
(b) The computational plane.
Figure 6.5. Grid configurations for the duct cross section and the corresponding apexes. (a) physical
plane; (b) the computational domain.
Differential area on the physical cross section is
dA = Jdξdη (6.50)
RESULTS AND DISCUSSION

The meshes in fluid are three dimensional. To account for the inlet influences, grids are
denser near inlet, while relatively sparser and evenly distributed after inlet. The meshes on
two fins are two dimensional and they are the same structure in z direction with the meshes of
fluid.
The velocity and momentum equations (6.47) and (6.48) are further discretized. The
diffusion term is discretized by central-difference scheme and the convective term is by
upwind scheme. The proposed finite difference schemes are implicit numerical schemes with
second order accuracy. They are unconditionally stable. The problem is a typical conjugate
one. After the solution of fluid velocity and temperature, temperatures on one fin are solved
on the generated grids, taking the current fluid temperature as the default boundary
conditions. The values on the other fin are obtained by symmetry. The whole calculating
procedure can be summarized as the following:
a. Grid generation for both the fluid and the fins.

b. Solve momentum equation Eq.(6.47). Get the velocity fields and resistance data for
the duct.
c. Assume initial temperature fields in the fluid.
d. Taking current fluid temperature as the boundary conditions for fins. Get the
temperature fields on one fin, by the solution of Eq.(6.27). Get the temperature on the
other fin because it is in axis symmetry with the first fin.
e. Taking the current values of temperature on two fins as the default values, get the
temperature profiles in the fluid by solving Eq.(6.47).
f. Go to (d), until the old values and the newly calculated values of temperature at all
calculating nodes are converged.
Directions of coordinate s are dictated by neighboring grid points along ξ=0 for fin AD
and along η=1.0 for fin DC respectively. Directions of coordinate y1 are determined from
neighboring grid points normal to ξ=0 for fin AD and normal to η=1.0 for fin DC
respectively.
To assure the accuracy of the results presented, a grid independence test was performed
for the duct to determine the effects of the grid size. It indicates that 21×21 grids on duct
cross section and Δz*=0.001 axially are adequate (less than 0.1% difference compared with
31×31 grids and Δz*=0.0005).
To further validate the numerical program, ordinary ducts of various cross sections are
calculated under uniform temperature conditions for all walls. Namely, ducts with no fins are
considered first. For hydrodynamically fully developed laminar flow in ducts, (fRe) is a
constant. The local Nusselt numbers in the duct will decrease along the flow and reach stable
values when the flow is thermally fully developed. The fully developed Nu values under
114 Li-Zhi Zhang
uniform temperature conditions are denoted as NuT. The calculated values of (fRe) and NuT
for various cross sections and aspect ratios are listed in Table 6.2. Comparisons are made with
the values form well-known references [1-3]. As seen, the current study predicts the flow
well. They are in accordance with the published data. Maximum uncertainty is less than 5%.
Velocity and Temperature Profiles

After the program is validated, sinusoidal ducts with one plate of uniform temperature
and two fins shown in Figure 6.1 are modeled. Velocity and temperature fields in the fluid are
obtained. Figure 6.6 shows the velocity profile distribution on the duct cross section for
τ=1.0. The values in the figure are dimensionless velocity u*. As seen, in the center, the
contours are near-circular, but in places near the boundaries, they are near-triangles. In the
corners, there are dead spaces with little fluid flowing which will affect heat transfer finally.
The flow shows a same pattern as that of a common sinusoidal duct without fins.
The dimensionless temperature profiles in the fluid are shown in Figure 6.7. The results
also found that at higher fin conductance parameters, the temperature contours are similar to
those in ducts with three walls of uniform temperature. At lower fin conductance parameters,
temperature contours show in majority a pattern of straight lines parallel to the plate, which is
similar to temperature profiles in parallel-plates ducts. Besides, in the center, isotherms are in
elliptical shapes. When Ω increases, the contours transform from parallel horizontal lines to
triangular shapes, implicating increased heat transfer from fins to fluid.
Table 6.2. Fully developed (fRe) and Nusselt numbers for ducts of various cross sections
(fRe) NuT
Cross sections Error Error
Refs This study Refs This study
(%) (%)
Circular 16.0 15.88 0.8 3.657 3.66 0.6
0.125 20.5 20.34 0.9 5.60 5.73 2.3
0.25 18.25 18.16 0.6 4.44 4.55 2.5
Rectangular 2a/2b
0.5 15.50 15.32 1.3 3.39 3.43 1.8
1.0 14.227 14.11 1.5 2.976 3.06 2.8
0.289 13.243 12.77 2.8 2.301 2.262 1.7
Isosceles 0.5 13.301 12.94 2.9 2.359 2.451 3.9
2a/2b
Triangular 0.866 13.321 13.41 0.5 2.500 2.594 3.7
1.866 13.09 12.96 1.0 2.284 2.391 4.7
0.5 11.207 11.170 0.4 2.12 2.181 2.8
0.75 12.234 12.212 0.2 2.33 2.374 1.7
Sine 2a/2b 1.0 13.003 12.954 0.3 2.45 2.521 2.9
1.5 14.002 14.115 0.8 2.6 2.573 1.1
2.0 14.553 14.647 0.7 ⎯ 2.886 ⎯
2
01
0.0
0.
06
9
0.02
0.0
23
18
0.75
0.0
6
41
0.
0.0 0.00
0.0
03
5
18
0.0 3
0.0
0.
0.03
05
12
2
y*
4 7
47
12
0.0
35
0.5
0.
0.0
0.0
0.0
58
07
0.0 0.041
0.00
0
29
0.06
4
0.068
76
0.041 0.029
6 0.018
0.05
0.0
2
3
4
3
0.08
0.05
0.02
0.03
0.010.006
5
8
0.25
0.01
0.076
0.047
0
0.0 .070
3
64
0.023
2
7
05
2
04
0.01
0.058
0.
0.
0.0
0.01
29
0.041
8
0.023
0.035 0.029
6
0.006 0.012 0.018 0.00
0
0 0.25 0.5 0.75 1
x*
Figure 6.6. Dimensionless velocity profiles (u*) on duct cross section, τ=1.0.
1
0.099
0.149
0.198
0.75 48
0.2 0.297
0. 34
7
0.4 0.
46 3 96
95
y*
0.4
6
0.5
0.39
0.347
45
0. 5
0.396
0.545
47
95
0. 3
0. 4
0.25 0. 3 4
0.4 6
6
97
0.2 47 0.39 0. 2
97
0.248
0.198
0.099 0.149
0.050 0.099
0.050
0
0 0.25 0.5 0.75 1
x*
Figure 6.7. Dimensionless temperature contours on duct cross section at z*=0.1 for τ=1.0 and Ωs=1.0.
116 Li-Zhi Zhang
1.5
0.018
0.038 0.0510.077 0.018
0.038 0.0510.077
0.257 0.154 0.103 0.103
1 00.56 0.206 0.154
.6 6 0.309 0.257 0.206
69 0.360
0.514 0.463 0.411 0.257
0.309
s*
0.61
7
0.36
0. 72
0
0
0.5
71
0.7
6
0 . 56 0.411
0.514 0.463 0.309
9 0.360 0.257
66 0.309 0.206
0. 0.411 0.257 0.154
0.617
0 0.206
0.3066 0.154 0.103 0.077
0.20.154 0.103 0.051
0.038 0 018 0.077 0.051
0.038 0.018
0
0 0.05 0.1 0.15
z*
Figure 6.8. Dimensionless temperature contours on fins for ducts of τ=1.0 and Ωs=1.0.
Dimensionless temperature profiles on the fins are show in Figure 6.8. Temperatures at
upper and lower boundaries are equal to the plate temperature. Temperatures at other
locations vary with conductance parameters. The higher the fin conductance parameters, the
more homogeneous the temperature distributions on fins are. Under infinitely large fin
conductance parameters, temperatures on fins will be equal to plate temperature. In this
situation, the plate-fin duct becomes a common duct with three walls of uniform temperature.
6 1
θb
5 NuL
Num 0.8
4
Nu
3 0.6
θb
2
0.4
1
0 0.2
0 0.05 0.1 0.15
z*
Figure 6.9. Variations of bulk temperature, NuL and Num along flow axis, τ=1.0, Ωs=0.1.
Nusselt Numbers
Once the temperature fields in fluid has been determined, the dimensionless bulk
temperature, the local Nusselt numbers, and the mean Nusselt numbers can be calculated.
Figure 6.9 shows the axial variations of bulk temperature, local and mean Nusselt numbers
for a duct. The higher the fin conductance parameters, the more rapid for the fluid bulk
temperature to reach the plate temperature, indicating much more heat is exchanged. The
thermal entrance length zth* is defined as the axial distance required to achieve a value of the
local Nusselt number NuL, which is 1.05 times the fully developed Nusselt value NuT. The
calculated thermal entry length for various aspect ratios and fin conductance parameters are
around 0.09.
After the thermal entry length, the local Nusselt number will come to a stable value NuT.
This fully developed value varies with aspect ratios and fin conductance parameters. Table
6.3 lists the calculated NuT for various aspect ratios and fin conductance parameters.
Table 6.3. Fully developed Nusselt numbers and fin efficiencies for plate-fin sinusoidal
ducts
τ Ωs NuT ηfin τ Ωs NuT ηfin

∞ 1.541 1.0 ∞ 2.181 1.0
25 1.525 0.982 25 2.135 0.978
10 1.211 0.774 10 2.109 0.966
5 1.043 0.666 5 1.906 0.875
0.2 2 0.873 0.556 0.5 2 1.522 0.699
1 0.794 0.510 1 1.262 0.578
0.5 0.738 0.476 0.5 1.053 0.484
0.1 0.693 0.447 0.1 0.821 0.378
0 0.643 0.413 0 0.735 0.338
∞ 2.521 1.0 ∞ 2.886 1.0
25 2.494 0.988 25 2.712 0.940
10 2.367 0.940 10 2.653 0.918
5 2.166 0.860 5 1.922 0.666
1.0 2 1.723 0.683 2.0 2 1.513 0.523
1 1.376 0.547 1 1.193 0.413
0.5 1.083 0.429 0.5 0.924 0.321
0.1 0.752 0.298 0.1 0.631 0.218
0 0.578 0.228 0 0.356 0.124
∞ 2.584 1.0 1 1.015 0.392
25 2.391 0.924 0.5 0.831 0.322
5.0 10 2.171 0.839 5.0 0.1 0.662 0.256
5 1.657 0.640 0 0.172 0.067
2 1.276 0.493
118 Li-Zhi Zhang
Fin efficiency is defined as the ratio of NuT at a certain fin conductance to that at Ωs=∞,
or
Nu T, Ω s
η fin = (6.51)
Nu T, Ωs =∞
The fin efficiency for various cases is also listed in Table 6.3.
6.3. RECTANGULAR DUCT

A schematic of a plate-fin duct of rectangular cross section is shown in Figure 6.10 [9]. It
is also one of the most commonly encountered total heat exchanger structure. The duct height
is 2a, and the duct width is 2b. The definition of fin heat conductance parameter is similar to a
sinusoidal duct. The calculated values of (fRe) for various cross sectional aspect ratios are
listed in Table 6.4.
λfδ
Ωs = (6.52)
λ a (2a )
Figure 6.10. Schematic of a compact heat exchanger comprised of rectangular plate-fin passages.
Table 6.4. Fully developed (fRe) for rectangular ducts of various aspect ratios
τ=2a/2b (fRe)
0.125 20.51
0.25 18.25
0.5 15.51
1.0 14.227
3.0 17.252
Table 6.5 lists the calculated NuT for various aspect ratios and fin heat conductance
parameters [9]. Also listed are values of thermal entry length. The thermal entrance length zth*
is defined as the axial distance required to achieve a value of the local Nusselt number NuL,
which is 1.05 times the fully developed Nusselt value NuT [1]. Generally, the thermal entry
length is in the range from 0.01 to 0.1. The larger the aspect ratios are, the longer the thermal
entry length is. As listed in Table 6.5, the larger the fin conductance parameters, the higher
the fully developed NuT. In heat exchanger design, fully developed NuT can be estimated
from this table for different aspect ratios and fin conductance parameters.
Table 6.5. Fully developed Nusselt numbers and thermal entry length for rectangular
plate-fin ducts
2a/2b Ωs NuT zth* 2a/2b Ωs NuT zth*

∞ 5.746 0.014 ∞ 4.537 0.013
25 5.683 0.016 25 4.476 0.029
10 5.676 0.016 10 4.463 0.030
5 5.574 0.021 5 4.375 0.041
0.125 2 5.563 0.020 0.25 2 4.364 0.037
1 5.542 0.019 1 4.345 0.038
0.5 5.536 0.018 0.5 4.3431 0.033
0.1 5.524 0.014 0.1 4.305 0.026
0 5.504 0.014 0 4.294 0.021
∞ 3.501 0.041 ∞ 3.021 0.044
25 3.485 0.043 25 3.013 0.045
10 3.483 0.045 10 3.012 0.046
5 3.452 0.043 5 3.012 0.049
0.5 2 3.443 0.039 1.0 2 2.684 0.043
1 3.353 0.043 1 2.485 0.043
0.5 3.294 0.039 0.5 2.271 0.042
0.1 3.153 0.033 0.1 1.942 0.040
0 3.122 0.026 0 1.815 0.039
∞ 3.500 0.041 ∞ 4.527 0.025
25 3.471 0.046 25 4.464 0.019
10 3.465 0.048 10 3.396 0.018
5 3.352 0.021 5 3.033 0.048
2.0 2 2.256 0.051 4.0 2 2.044 0.065
1 1.896 0.056 1 1.393 0.082
0.5 1.491 0.062 0.5 0.938 0.098
0.1 1.014 0.074 0.1 0.485 0.125
0 0.875 0.070 0 0.399 0.094
The results in the table show that, besides aspect ratios, the fin heat conductance
parameters have equal tremendous impact on heat transfer phenomenon in the plate-fin
narrow passages. When fin conductance parameters increase from 0 to infinitely large, the
120 Li-Zhi Zhang
duct behaves from like a duct with two well-conductive wall and two adiabatic walls to a duct
with four well-conductive walls. The consequent fully developed Nusselt numbers increase
accordingly. For ducts with certain limited conductance parameters, the Nusselt numbers lie
between these two limiting values. For fins made with low conductivity materials like paper,
polymer, glass, carbon, plywood, etc, the nature of limited heat transfer from fins to fluid
should be taken into account, for a better estimation of the exchanger performance. The
Nusselt numbers under uniform heat flux boundary conditions are usually 20-30% higher than
those under uniform temperature conditions.
Parallel-plates Channels
Parallel-plates channels are a special type of rectangular ducts. The aspect ratio is
infinitely small. In this case, the finite fin conductance has no influence on convective heat
transfer coefficients. The fully developed Nusselt number under uniform wall temperature is
7.54 [2]. The friction coefficients and the fully developed Nusselt numbers for rectangular
ducts of other cross sectional shapes are also found in [2].
6.4. TRIANGULAR DUCT

A schematic of a plate-fin duct of triangular cross section is shown in Figure 6.11 [10]. It
is one of the most commonly encountered total heat exchanger structure. The cross section is
usually isosceles triangle. Sometimes, it is approximated by a sinusoidal cross section, or vice
versa. Anyway, this chapter gives its exact numerical results for heat mass transfer. As ducts
of other geometries, the duct height is 2a, and the duct width is 2b. For triangular duct, it’s
more convenient to use the apex angle 2α to define the geometry.
2α, Apex angle,
Figure 6.11. Schematic of a cross-flow plate-fin heat exchanger with triangular duct geometry.
Table 6.6. Fully developed (fRe) for triangular ducts of various half apex angles
α (°) (fRe)
5 12.511
15 13.08
30 13.322
45 13.302
60 13.244
75 13.141
The fin heat conductance parameter is still defined as
λfδ
Ωs = (6.53)
λ a (2a )
The calculated values of (fRe) for various half apex angles are listed in Table 6.6 for
triangular ducts [10].
Table 6.7 lists the calculated NuT for various half apex angles and fin conductance
parameters. The larger the fin conductance parameters, the higher the fully developed NuT.
Table 6.7. Fully developed Nusselt numbers for plate-fin triangular ducts
α Ωs NuT α Ωs NuT
∞ 2.391 ∞ 2.596
25 2.285 25 2.513
10 2.194 10 2.444
5 2.063 5 2.301
15° 2 1.597 30° 2 1.938
1 1.285 1 1.605
0.5 1.021 0.5 1.277
0.1 0.722 0.1 0.843
0 0.653 0 0.653
∞ 2.451 ∞ 2.262
25 2.395 25 2.103
10 2.296 10 2.027
5 2.183 5 1.927
45° 2 2.024 60° 2 1.794
1 1.741 1 1.619
0.5 1.441 0.5 1.415
0.1 0.975 0.1 1.065
0 0.656 0 0.593
122 Li-Zhi Zhang
6.5. CONVECTIVE MASS TRANSFER COEFFICIENTS

Convective heat transfer in plate-fin ducts with finite fin conductance has been
summarized. Duct cross sections include sinusoidal, rectangular, and triangular. The fins
include common metals to non-metal materials. Besides heat transfer coefficients, to properly
evaluate the heat mass exchanger performance, mass transfer coefficients are equally
important. Traditionally, mass transfer coefficients in a common pipe are obtained from heat
transfer coefficients by heat mass transfer analogies, of which the most frequently cited one is
the Chilton-Colburn Analogy [6]. According to this methodology, following equations can be
used to estimate convective mass transfer coefficients in a pipe:
js = j L (6.54)
The Chilton-Colburn j factor for heat transfer [6]
js = St s Pr 2 / 3 (6.55)
where Pr is the Prandtl number of air. Stanton number for heat transfer
h
St s = (6.56)
ρ a cp u
where h is convective heat transfer coefficients (kWm-2K-1), u is air stream bulk velocity
(m/s), cp is specific heat (kJkg-1K-1), ρa is dry air density (kgm-3).
The convective heat transfer coefficient is also represented by a Nusselt number
hd h
Nu = (6.57)
λa
where dh is the hydrodynamic diameter of the channel (m), λ is thermal conductivity (kWm-
1
K-1).
The Chilton-Colburn j factor for mass transfer
j L = St L Sc 2 / 3 (6.58)
where Sc is the Schmidt number of moisture air. Stanton number for mass transfer
k
St L = (6.59)
ρau
where k is convective mass transfer coefficient (m/s).

Mass transfer in boundary layers is also described by a Sherwood number
kd h
Sh = (6.60)
Dva
where Dva is vapor diffusivity in air (m2/s). Substituting Eqs.(6.55) through (6.60) into (6.54),
a relation can be obtained
Sh = Nu ⋅ Le −1 / 3 (6.61)
Pr
Le = (6.62)
Sc
where Le is commonly called the Lewis number. For ventilation air and vapor mixture, which
is always near atmospheric states, the Lewis number varies in the range of 1.19 to 1.22 [6],
therefore it is usually approximate that Sh=Nu.
If above analogy still holds in our case, then the estimation of convective mass transfer
coefficients would be simple, from the previously obtained data. However, in plate-fin ducts
used for non-metal total heat exchanger, the fin conductance parameters for heat and mass
transfer are so different that it’s questionable that such an analogy still exists. It’s therefore
imperative to obtain the convective mass transfer coefficients and find the relations between
the heat and mass transfer in such plate-fin ducts.
Governing Equations
Air travels in the duct while exchanging moisture with duct walls. Mass conservation in
the air stream can be expressed by
∂ω ⎛ ∂ 2ω ∂ 2ω ⎞
u = Dva ⎜⎜ 2 + 2 ⎟⎟ (6.63)
∂z ⎝ ∂x ∂y ⎠
where ω is humidity ratio of air (kg vapor/kg dry air).

The above equation can be normalized to
∂ξ ∂ 2ξ ⎛ b ⎞ ∂ 2ξ
2
U * = +⎜ ⎟ (6.64)
∂z ∂x * ⎝ a ⎠ ∂y *
2 2
where dimensionless humidity ratio

124 Li-Zhi Zhang
ω − ωw
ξ= (6.65)
ωi − ω w
where in the equations, ωi is the inlet air humidity, and ωw is the humidity on wall surface.
Dimensionless bulk humidity
∫∫ u ξdA
*
ξb = (6.66)
∫∫ u dA
*
Similarly, an mass balance in a control volume in the duct will give the equation for
estimation of the local Sherwood number as
1 dξ b
ShL = − (6.67)
4ξ b dz *
and the mean Sherwood number from z*=0 to z* by
1
Shm = − ln ξ b (6.68)
4z*
Similar to heat conduction in the fin in Figure 6.4, at any location, there is a mass balance
between the net water diffusion along the fin and the mass transfer from the surface of the fin
to the fluid. The phenomenon is expressed by
d 2W
Dwf ρ f δ 2 = mu + mL (6.69)
ds
⎛ ∂ω ⎞
mu = − Dva ρ a ⎜⎜ ⎟⎟ (6.70)
⎝ ∂y f ⎠u
⎛ ∂ω ⎞
mL = Dva ρ a ⎜⎜ ⎟⎟ (6.71)
⎝ ∂y f ⎠L
where Dwf is diffusivity of water in fin materials (m2/s); ρf is density of fin materials (kg/m3);
W is water content in fin materials (kg water/kg dry material); mu is moisture flux from upper
surface of fin, and mL is moisture flux from lower surface. (∂ω / ∂y f ) is the normal gradient
of vapor concentration on the lower or upper surface of fin. Due to symmetry, the mass flux
at the lower surface and the upper surface are also skew symmetric. The relation is
mathematically expressed by
mu s
= −mL Lf − s
(6.72)
Water uptake in a hygroscopic material is a function of air relative humidity. Figure 6.12
shows the measured sorption isotherm of the material used for the investigated total heat
exchanger. The discrete dots are the measured data with a thermo-hygrostat. A polynomial
curve can be regressed to represent this isotherm from 0% to 100% RH. The relation between
RH and humidity ratio is
RH = 10 −6 e 5294 / T ω (6.73)
The temperature differences between two surfaces of a fin are quite small due to the
small fin thickness. For air conditioning industry, heat mass exchanger always works between
40%RH and 80%RH. Considering these factors, the relative humidity can be a linear
expression of humidity ratio, and the sorption isotherm for the material can also be
approximated by a linear equation as
W = kpω (6.74)
where kp is defined as the partition coefficient.

Substituting Eqs.(6.74) in (6.69) to (6.72), following dimensionless equation can be
found
0.5
0.4
W (kg/kg)
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Relative humidity RH
Figure 6.12. Sorption isotherm for a material in total heat exchanger.

126 Li-Zhi Zhang
d 2ξ f ⎛ ∂ξ ⎞ ⎛ ∂ξ ⎞
ΩL = −⎜⎜ * ⎟⎟ − ⎜⎜ * ⎟⎟ (6.75)
⎝ ∂y f ⎠ s* ⎝ ∂y f ⎠ L*f − s*
*2
ds
where ΩL is defined as the dimensionless fin mass conductance parameter. It is calculated by
δρ f Dwf k p
ΩL = (6.76)
ρ a Dva (2a )
where dimensionless humidity ratio in fins
ωf − ω W
ξf = (6.77)
ωi − ω W
Boundary Conditions
The boundary conditions of humidity for fluid
ξ=0, at y*=0 (6.78)
Inlet condition
ξ=1, at z*=0 (6.79)
The boundary conditions for fins
ξf=0 at s*=0, Lf* (6.80)
Fin-fluid coupling:
and ξ f = ξ at fin-fluid interfaces (6.81)
It can be found that Eqs.(6.64) and (6.75) are in the same forms as Eqs.(6.10) and (6.27)
respectively. Therefore, the solution of dimensionless humidity ratio can be an analogy to the
solution of dimensionless temperature, if the fin mass conductance parameters are the same
values as the fin heat conductance parameters. In other words, the solution of dimensionless
humidity is the same as the dimensionless temperature, if assuming same values of heat and
mass fin conductance parameters. Consequently, it is only necessary to solve either the heat
transfer equations or the mass transfer equations. For mass transfer properties, if assuming ΩL
is the same value as Ωs, ShL will equal to NuL. Correspondingly, the fully developed
Sherwood number under uniform concentration boundary conditions Shω is equal to NuT.
In a summary, convective mass transfer coefficients can be estimated in the following

steps: (1) Calculate the value ΩL by Eq.(6.76); (2) Use ΩL to replace Ωs in Table 6.3, 6.5 or
6.7, depending on which shape the cross section is; (3) the value of Shω is equal to NuT in the
table.
6.6. COMPARISONS OF HEAT AND MASS TRANSFER

As defined by Eq.(6.28), fin heat conductance parameters are determined by duct
geometry and heat conductivity of fin materials. Table 6.3 lists the values of fin heat
conductance parameters for some frequently encountered materials, including metal and non-
metals, for a sinusoidal duct of height 2mm, aspect ratio 0.5, and fin thickness 0.1mm, as
mentioned above. As seen, for almost all the metals, the fin heat conductance parameters are
larger than 100, and the resulting fin heat efficiencies can be as high as 0.90-0.98. For such
traditional metal compact heat exchangers, the influences of finite fin heat conductance on
heat transfer are negligible. Direct utilization of heat transfer properties for a common duct is
acceptable. For the total heat exchangers which use non-metals to simultaneously transfer
heat and moisture, the fin heat conductance parameters are usually less than 1.0, and the
resulting fin efficiencies for heat transfer can be as low as 0.40. Though it’s true that the fins
still participate in the heat transfer enhancement, at least partially, at this stage, the effects of
finite fin heat conductance on heat transfer will be substantial.
Similarly, as defined by Eq.(6.76), fin mass conductance parameters are determined by
duct geometry and mass diffusivity of moisture in fin materials. Moisture diffusivity in
materials is not accumulated as detail as heat conductivity. There is only sporadic data
available. Moisture diffusivity in air is around 2.5×10-5m2/s [11]. A literature review found
that moisture diffusivities in most non-metal hygroscopic materials are in or less than the
order of 10-10 m2/s. For example, water diffusivities are from 0.5×10-12 to 1.5×10-12m2/s in a
PSS-Na/Al2O3 composite silicate [12]; from 8.0 ×10-10m2/s to 5.0 ×10-12m2/s in a Nafion
polymer membrane [13]; and from 1.0 ×10-13m2/s to 1.0 ×10-12m2/s in a methylcellulose film
[14]. The densities of these materials are in the order of 1000kg/m3. The partition coefficients
are in the order of 10, or 15-20% water uptake under 60% RH. Based on these properties, the
values of fin mass conductance parameters are in the order of 10-3. With such low mass
conductance parameters, the fin efficiencies for mass transfer will be below 0.1 to 0.2. Under
such circumstances, nearly all the mass transfer between the two air streams will be
accomplished by the plate, rather than by the fins. The fins seem to behave only like
supporting materials, if excluding their role in partial participation in heat transfer. In
engineering, sometimes supporting materials or spacers are necessary to separate the two
streams because the plates are thin and soft.
6.7. CONCLUSION
Laminar flow and heat mass transfer in plate-fin ducts with various geometries such as
sinusoidal, rectangular, triangular, parallel-plates, are investigated by considering finite fin
128 Li-Zhi Zhang
conductance in heat and mass transfer. Finite fin conductance parameters in heat and mass
have determining effects in convective heat and mass transfer in the ducts.
In designing novel heat mass exchangers, it’s simple to emulate a common compact heat
exchanger with such a plate-fin configuration. However, the effectiveness of fins on heat
transfer will be strongly compromised. Even worse, it has little use in enhancing mass
transfer. Nevertheless, this structure is still popular sue to its high mechanical strength and
compactness.
REFERENCES
[2] Incropera, F.P.; Dewitt, D.P. Introduction to Heat Transfer. 3rd edn. New York: John
Wiley & Sons; 1996. Chapter 8, pp. 416.
[3] Shah, R.K.; Bhatti, M.S. Laminar Convection Heat Transfer In Ducts. in: Handbook of
Single-Phase Convective Heat Transfer, (Ed. S. Kakac, R.K. Shah, W. Aung). New
York: Wiley; 1987.
[4] Baliga, B.R.; Azrak, R.R. Laminar fully developed flow and heat transfer in triangular
plate-fin ducts. ASME Journal of Heat Transfer, 1986, 108, 24-32.
confined by sinusoidal and arc curves. International Journal of Heat and Mass
Transfer, 2002, 45, 571-578.
[6] Zhang, L.Z. Heat and mass transfer in plate-fin sinusoidal passages with vapor-
permeable wall materials. International Journal of Heat Mass Transfer, 2008, 51, 618-
629.
[7] Thompson, J.F.; Thames, F.; Martin, C. Automatic numerical generation of body-filled
curvilinear coordinate system for field containing any number of arbitrary two-
dimensional bodies. J. Comput. Physics, 1974, 24, 299-319.
[8] Thomas, P.D.; Middlecoff, J.F. Direct control of grid point distribution in meshes
generated by elliptic equations. AIAA J., 1982, 18, 652-656.
[9] Zhang, L.Z. Thermally developing forced convection and heat transfer in rectangular
plate-fin passages under uniform plate temperature. Numerical Heat Transfer, Part A-
Applications, 2007, 52, 549-564.
[10] Zhang, L.Z. Laminar flow and heat transfer in plate-fin triangular ducts in thermally
developing entry region. International Journal of Heat Mass Transfer, 2007, 50, 1637-
1640.
[11] Kays, W.M.; Crawford, M.E. Convective heat and mass transfer, 3rd edn. New York:
[12] Aranda, P.; Chen, W.J. Martin, C.R. Water transport across polystyrenesulfonate/
alumina composite membranes. Journal of Membrane Science, 1995, 99, 185-195.
[13] Ye, X.H. LeVan, M.D. Water transport properties of Nafion membranes Part I. Single
161.
[14] Debeaufort, F.; Voilley, A.; Meares, P. Water vapor permeability and diffusivity
through methylcellulose edible films. Journal of Membrane Science, 1994, 91, 125-
133.
Chapter 7
EFFECTIVENESS CORRELATIONS
OF TOTAL HEAT EXCHANGERS
ABSTRACT
How to estimate the sensible and latent effectiveness of a total heat exchanger is a
key issue in engineering design and optimization. Effectiveness correlations are a simple
way to calculate performance. Contrary to sensible-only heat exchangers, total heat
exchangers are influenced not only by geometries, but also by material properties. In this
chapter, effectiveness correlations are deduced for total heat exchanger, both parallel
types and plate-fin types. They can be conveniently used in engineering design.
NOMENCLATURE
A Transfer area (m2)
C Constant in sorption curve
cp Specific heat of air (kJkg-1K-1)
Dwp Diffusivity of water in plate (m2/s)
Dwp* Equivalent diffusivity of moisture in plate (m2/s)
H Specific enthalpy (kJ/kg)
H* Ratio of latent to sensible energy differences between the inlets of two air
streams
k Convective mass transfer coefficient (m/s)
m Mass flow rate of air streams (kg/s)
m W Mass flow rate of moisture flow (kgm-2s-1)
NTU Number of transfer units for sensible heat
NTUL Number of transfer units for latent heat
Nu Nusselt number
n Number of channels
R Ratio for heat/mass capacity
132 Li-Zhi Zhang
Sh Sherwood Number
T Temperature (°C)
U Total heat transfer coefficient (kWm-2K-1)
UL Total mass transfer coefficient (m/s)
wmax Maximum water uptake of desiccant (kgkg-1)
x, y Coordinates (m)
x*, y* non-dimensional coordinates
xF Length of supply channel (m)
yF Length of exhaust channel (m)
Greek Letters
ψ Coefficient of moisture diffusive resistance in membrane (CMDR)

λ Thermal conductivity of membrane (kWm-1K-1)
θ Moisture uptake in membrane (kgkg-1)
ε Effectiveness
r Resistance (m2K/kW for thermal and s/m for moisture)
δ Thickness of membrane (m)
ω Moisture content (kg moisture/kg dry air)
α Ratio of diffusive to convective moisture resistance for membrane
β Ratio of total number of transfer units for moisture to that for sensible heat
Subscripts
a Air
c Convective
e Exhaust
f Fresh
i Inlet
L Latent, moisture
p plate
o Outlet
tot Total
w Water
7.1. INTRODUCTION
Stationary total heat exchangers are similar to air-to-air sensible heat exchangers, either
plate type or plate-fin type. The difference is that water vapor-permeable materials are used
instead of metal foils. For common heat exchangers, ε-NTU (effectiveness-NTU) method is a
simple way to predict performance and to design a heat exchanger [1,2]. How to emulate this
methodology to total heat exchanger is of interest.
Effectiveness Correlations of Total Heat Exchangers 133
Heat mass transfer in a total heat exchanger is more complicated than a sensible-only heat
exchanger. Detailed modeling of the system requires finite difference computations, which
are hard and time consuming for common engineers. Therefore, ε-NTU method for the total
heat exchanger provides an efficient yet simple tool for exchanger optimization and design. In
this chapter, simple correlations that could predict the sensible and latent effectiveness will be
introduced, by summarizing the couplings between the performance and the sorption
characteristics of plate materials, and the operating conditions. The deduction of this chapter
is more general than previous studies.
7.2. EFFECTIVENESS-NTU CORRELATIONS

Consider a simple parallel-plates total heat exchanger. The plates are plain. It is a cross-
flow total heat exchanger with paper or membranes as the plates. The plate thickness is δ.
This structure is common in practice. The device is shown in Figure 7.1. Two air streams- the
fresh and the exhaust flow in thin, parallel, alternating membrane or paper layers, in order to
transfer heat and moisture from one air stream to the other. In air conditioning, the fresh air is
usually the outside air and the exhaust is the room air that needs to be discharged to the
outside. The governing dimensionless equations for simultaneous heat and moisture transfer
in the total heat exchangers, based on the assumptions listed in Table 7.1, are as follows [3,4]:
Fresh air
∂Tf
= 2 NTU f (Tpf − Tf ) (7.1)
∂x *
∂ω f
= 2 NTU Lf (ω pf − ω f ) (7.2)
∂x *
Exhaust air
∂Te
= 2 NTU e (Tpe − Te ) (7.3)
∂y *
∂ω e
= 2 NTU Le (ω pe − ω e ) (7.4)
∂y *
134 Li-Zhi Zhang
Exhaust in
Membrane plates
Fresh out
Fresh in
Duct Sealing
Exhaust out
Figure 7.1. Schematic of a cross-flow total heat exchanger.
Table 7.1. Assumptions used in governing equations
1. There is no lateral mixing of the two air streams.

2. Heat conduction and vapor diffusion in the two air streams are negligible compared
to energy transport and vapor convection by bulk flow.
3. Adsorption of water vapor and plate material is in equilibrium adsorption-state.
4. Both the heat conductivity and the water diffusivity in the plate are constants.
5. Heat and moisture transfer is one-dimensional in membrane. It is in thickness.
where
x y
x* = , y* = ,
xF yF
nf hf x F y F hf Atot nhx y h A
NTU f = = , NTU e = e e F F = e tot ,
m f c pf m f c pf m e cpe m e cpe
ρ a k f Atot ρ a k e Atot
NTU Lf = , NTU Le = ,
m f m e
where Tpf and Tpe are the temperature of plate in fresh side and exhaust side respectively (°C),
ω is humidity ratio in air streams (kg/kg), cp is specific heat of air (kJkg-1K-1); x and y are
coordinates (m), h is convective heat transfer coefficient (kWm-2K-1), k is convective mass
is mass flow rate of dry air (kg/s), xF and yF are lengths of flow
transfer coefficient (m/s), m
channels (m), n is the number of channels for each flow. The subscript “f”,”p”, and “e” mean
“fresh air”, “plate”, and “exhaust”, respectively. Previous studies found that the temperature
difference between the two sides of membrane is very small due to the small thickness of
membrane [5]. So it is reasonable to assume that Tmf = Tme =0.5(Tf+Te).
From assumptions (1) and (2), it is seen that heat and moisture transfer is one-
dimensional and along the flow direction. However, due to the cross-flow arrangements, the
temperature and humidity distributions in the air streams are two-dimensional.
The above dimensionless parameters NTU are the commonly recognized Number of
Transfer Units. They give an insight into the characteristics of heat and moisture exchange
between fluids and surfaces.
Moisture flow rate through the plate:
θ pf − θ pe
m w = ρ p Dwp (7.5)
δ
where θpf, θpe are moisture uptake in plate at two surfaces (kg.kg-1), δ is the plate thickness,
and Dwp is the water diffusivity in plate (m2s-1).
The equilibrium between the plate and moisture at its surface can be expressed with a
general sorption curve as
wmax
θ= (7.6)
1− C + C /φ
where wmax represents the maximum moisture content of the plate material (i.e., moisture
uptake when φ=100%) and C determines the shape of the curve and the type of sorption.
The parameters θ, φ, and ω can be correlated by ideal gas state equation and
psychrometric relations.
The convective heat transfer coefficients are obtained from Nusselt numbers for parallel
plates [1,2] and the mass transfer in boundary layers is often described by Sherwood
correlations. By using the well-known Chilton-Colburn Analogy [6]
Sh = Nu ⋅ Le −1 / 3 (7.7)
We have
h
k= Le −1 / 3 (7.8)
cp
For ventilation air and vapor mixture, which is always near atmospheric states, the Lewis
number, Le, varies in the range of 1.19 to 1.22, see Ref. [6].
For plate-fin ducts, the convective heat transfer coefficients and convective mass transfer
coefficients can be referenced form the calculated data in previous sections.
Analogous to the heat transfer effectiveness commonly used in heat exchanger analysis,
the concept of effectiveness can be applied to the heat and moisture transfer processes in a
136 Li-Zhi Zhang
membrane based enthalpy exchanger. For a constant specific heat and heat of phase change,
the effectiveness is defined as
Sensible effectiveness
(m c p ) f (Tfi − Tfo )

εs = (7.9)
(m c p ) min (Tfi − Tei )
Latent effectiveness
(m c p ) f (ω fi − ω fo )
εL = (7.10)
(m c p ) min (ω fi − ω ei )
Enthalpy transfer effectiveness, i.e., total energy transfer effectiveness
m f ( H fi − H fo )
ε tot = (7.11)
m min ( H fi − H ei )
where H is the specific enthalpy of air, and it is calculated by [4]
H = c pT + ω(2501 + 1.86T ) (7.12)
where T is in °C.
The third term in Eq.(7.12) usually has a less than 3% effect, thus it can be neglected.
Then the enthalpy effectiveness can be further simplified as
εs + εL H *
ε tot = (7.13)
1+ H *
where
2501(ω si − ω ei ) Δω
H* = ≈ 2501 (7.14)
c p (Tsi − Tei ) ΔT
where H* is essentially a ratio of latent to sensible energy differences between the inlets of
two air streams flowing through the total heat exchanger. H* can in theory vary from –∞ to
+∞, but varies typically from –6 to +6 for enthalpy recovery in HVAC applications. From
above equation, it is clear that the total enthalpy effectiveness is not a simple algebraic
average of sensible and latent effectiveness. When H*=1, εtot =(εs +εL)/2. As H* → ∞, εtot →
εL; as H* → 0, εtot → εs; as H* → -1, εtot → ±∞.
An overall number of transfer units is used to reflect the sensible heat transfer in an
exchanger. For the total heat exchanger that has equal area on both sides, the total number of
transfer units for sensible heat is
AtotU
NTU = (7.15)
(m c p ) min
where U is the total heat transfer coefficient. Its general form is
−1
⎡1 δ 1⎤
U =⎢ + + ⎥ (7.16)
⎣ hf λ he ⎦
The term in the middle is the thermal resistance of plate, which value is around 0.005
2
m K/kW. Other two terms are convective thermal resistance. Their values are in the order of
40 m2K/kW, or 8000 times larger than plate resistance. Therefore, plate resistance for heat
transfer can be neglected.
The sensible effectiveness is a function of two dimensionless parameters, NTU and R1,
the ratio of minimum to maximum heat capacity rate of two air streams.
R1 = (m cpa ) min /(m cpa ) max (7.17)
The moisture flow rate through the membrane can also be expressed as
m W = ρ a k f (ω f − ω ps ) = ρ a k e (ω pe − ω e ) (7.18)
∂θ ∂θ
θ pe = θ ps + Δφ = θ pf + (φ pe − φ pf ) (7.19)
∂φ pf
∂φ pf
Substituting Eq.(7.19) into Eq.(7.5), we have
ρ p Dwp ⎛ ∂θ ⎞
m W = ⎜⎜ ⎟⎟ (φ pf − φ pe ) (7.20)
δ ⎝ ∂φ ⎠ pf
Using Clapeyron equation to represent the saturation vapor pressure and assuming a
standard atmospheric pressure of 101325Pa gives the relation between humidity and relative
humidity as
φ e 5294 /(T + 273.15)

= − 1.61φ (7.21)
ω 10 6
138 Li-Zhi Zhang
where the second term on the right side of the equation will generally have less than a 5%
effect, thus it can be neglected. Thus the relation between φ and ω is expressed by
e 5294 /(T + 273.15)

φ= ω (7.22)
10 6
Substituting Eq.(7.22) into Eq.(7.20), we have
ρ p Dwp ⎛ ∂θ ⎞ e 5294 /(T + 273.15)

m W = ⎜⎜ ⎟⎟ (ω pf − ω pe ) (7.23)
δ ⎝ ∂φ ⎠ pf 10 6
The total moisture emission rate can be written as
m W = ρ aU L (ω f − ω e ) (7.24)
where
−1
⎛1 1⎞
U L = ⎜⎜ + rp + ⎟⎟ (7.25)
⎝ kf ke ⎠
where in the equation, the first and the third term are the convective mass transfer resistance
in air sides, and the middle term is plate mass transfer resistance, which is expressed by
ρaδ 10 6
rp = (7.26)
ρ p Dwp ⎛ ∂θ ⎞
e (5294 / T ) ⎜⎜ ⎟⎟
⎝ ∂φ ⎠ pf
Term UL can be called the overall mass transfer coefficient for the device. It is indicated
the overall moisture transfer coefficient, has an expression similar to overall heat transfer
coefficient.
The differentiation of Eq.(7.6) gives
∂θ wmax C
= (7.27)
∂φ (1 − C + C / φ ) 2 φ 2
The Eq. (7.26) can be further simplified as
ρaδ
rp = ψ (7.28)
ρ p Dwp
10 6 (1 − C + C / φ ) 2 φ 2
ψ= (7.29)
e ( 5294 / T ) wmax C pf
where the coefficient of diffusive resistance for plate, ψ, is co-determined by the operating
conditions and the slope of sorption curves of plate material.
Sometimes the sorption curve is represented by a simple linear equation
θ = kpω (7.30)
where kp is called partition coefficient. We can see that
1
kp = (7.31)
ψ
It means that partition coefficient is not a constant. However, it is often approximated by

a constant.
Similar to the definition of total number of transfer units for heat, the total number of
transfer units for moisture can be written as
ρ a AtotU L
NTU L = (7.32)
(m ) min
The comparison of total transfer units for moisture and sensible heat, assuming equal
specific heats for two air streams, gives
NTU L U L ρ a c p
β= = (7.33)
NTU U
It can be further simplified to [3]
1
β= (7.34)
1+α
where
rm
α= (7.35)
rc
2
rc = (7.36)
kf
140 Li-Zhi Zhang
where rc is the convective moisture transfer resistance, and α is the ratio of diffusive
resistance to convective resistance for plate. As can be seen, the total number of transfer units
for moisture can be estimated from the total number of transfer units for sensible heat, and
ratio of diffusive to convective moisture resistance. As α→∞, NTUL→0, no moisture can be
permeated through the plate. In this case, the “plate” is like a metal plate. On the other hand,
as α→0, NTUL→NTU, εL=εs. If α=1, NTUL=NTU/2. Under the common operating
conditions, the values of α vary from 2 to 10 [4], which implies that plate resistance for
moisture transfer cannot be neglected.
Deduction of Effectiveness Correlations
Considering a cross-flow membrane exchanger with only one flow channel. At any point
in the exchanger a heat and mass balance for an infinitely small volume dxdy can be written
from as
dq = U (Ts − Te )dxdy (7.37)
dm W = U L (ω s − ω e )dxdy (7.38)
Equation (7.37) is a basic heat transfer equation, and Eq.(7.38) has been widely employed
as a mass permeation model through a membrane in chemical industry.
Across the elements xF and yF units in length the energy and moisture balances yield
(m c )
pa s ∂Ts
dq = − dxdy (7.39)
yF ∂x
(m c )
pa e ∂Te
dq = dxdy (7.40)
xF ∂y
m s ∂ω s
dm W = − dxdy (7.41)
y F ∂x
m e ∂ω e
dm W = dxdy (7.42)
xF ∂y
Combining Eqs. (7.37) and (7.39) and then Eqs. (7.37) and (7.40) gives
Uy F ∂T
(Ts − Te ) = − s
(m cpa )s ∂x
(7.43)
UxF ∂T
(Ts − Te ) = e
(m cpa )e ∂y
(7.44)
Similarly, combining Eqs. (7.38) and (7.41) and then Eqs. (7.38) and (7.42) gives
U L yF ∂ω
(ω s − ω e ) = − s (7.45)
m s ∂x
U L xF
(ω s − ω e ) = ∂ω e (7.46)
m e ∂y
Differentiating Eqs. (7.43) and (7.44) with respect to y and x and taking their sum gives
UxF ∂ (Ts − Te ) Uy F ∂ (Ts − Te ) ∂ 2 (Ts − Te )

(m cpa )e ∂x + (m cpa )s ∂y = − ∂x∂y (7.47)
Similarly, differentiating Eqs. (7.45) and (7.46) with respect to y and x and taking their
sum gives
U L xF ∂ (ω s − ω e ) U L y F ∂ (ω s − ω e ) ∂ 2 (ω s − ω e )
+ =− (7.48)
m e ∂x m s ∂y ∂x∂y
Let dimensionless variables
Ts − Te ω − ωe x y
θ1 = , θ2 = s , x =
*
, y =
*
Tsi − Tei ω si − ω ei xF yF
and substituting in Eqs.(7.47), (7.48),
UxF y F ∂θ 1 UxF y F ∂θ 1 ∂ 2θ 1
+ = −
(m cpa )e ∂x* (m cpa )s ∂y * ∂x*∂y * (7.49)
U L xF y F ∂θ 2 U L xF y F ∂θ 2 ∂ 2θ 2
+ = − (7.50)
m e ∂x * m s ∂y * ∂x *∂y *
With
142 Li-Zhi Zhang
Ux F y F Ux F y F U x y
NTU a = , NTU b = , NTU La = L F F ,
(m c pa )e (m c pa )s m e
U L xF y F
NTU Lb =
m s
Eqs.(7.49) and (7.50) become
∂θ 1 ∂θ 1 ∂ 2θ 1
NTU a + NTU + =0 (7.51)
∂x * ∂y * ∂x *∂y *
b
∂θ 2 ∂θ 2 ∂ 2θ 2
NTU La + NTU Lb * + * * = 0 (7.52)
∂x * ∂y ∂x ∂y
Initial condition: θ 1 (0,0) = θ 2 (0,0) = 1

Mason [3] obtained a solution for Eq.(7.51) in the form of an infinite series by employing
the Laplace transformation as following:
n
∞ ⎡ ( NTU a )( NTU b ) x * y * ⎤
θ1 ( x , y ) = e
* * − ( NTU a ) y* + ( NTU b ) x*
∑⎢ (n!) 2
⎥ (7.53)
n =0 ⎣ ⎦
Since Eqs.(7.52) and (7.51) are the same form of differential equations. They are
identical if NTUa is replaced by NTULa and NTUb by NTULb. Therefore, the solution to
Eq.(7.52) can be written as
n
∞ ⎡ ( NTU La )( NTU Lb ) x * y * ⎤
θ 2 (x , y ) = e
* * − ( NTU La ) y* + ( NTU Lb ) x*
∑⎢ (n!) 2
⎥ (7.54)
n =0 ⎣ ⎦
The overall heat transferred in the exchanger is the integral of Eq.(7.53)
Q = UxF y F (Tsi − Tei )∫0 ∫0θ 1 ( x * , y * )dx * dy *

1 1
(7.55)
The overall moisture transferred in the exchanger is the integral of Eq.(7.54)
M w = U L xF y F (ω si − ω ei )∫0 ∫0θ 2 ( x * , y * )dx * dy *

1 1
(7.56)
UxF y F U L xF y F
NTU = , NTU L =
(m cpa )min (m )min
(m c ) m min
R1 = R2 =
pa min
(m c )
pa max
,
m max
1 1 1 1
Ω1 = ∫0 ∫0θ 1 ( x * , y * ) dx * dy * , Ω 2 = ∫ ∫ θ 2 ( x * , y * ) dx * dy *
0 0
Q MW
εs = , εL =
(m cpa )min (Tsi − Tei ) m min (ω si − ω ei )
Then
ε s = NTUΩ1 (7.57)
ε L = NTU L Ω 2 (7.58)
∞
1
Ω1 = ∑ f ( NTU a ) f ( NTU b )
( NTU a )( NTU b ) n=0
(7.59)
∞
1
Ω2 = ∑ f ( NTU La ) f ( NTU Lb )
( NTU La )( NTU Lb ) n=0
(7.60)
where
zmn
f ( z) = 1 − e ∑ −z
(7.61)
m =0 m!
There are two cases:
• If (m c pa ) e = (m cpa ) min , then NTUa=NTU, and NTUb= R1NTU

• If (m c pa ) e = (m cpa ) max , then NTUa= R1NTU, and NTUb= NTU
In both cases, Eq.(7.57) can be replaced by the relationship
1 ∞ ⎡ n
( NTU ) m ⎤ ⎧ n
[R1 ( NTU )] ⎫
m
εs = ∑ ⎢1 − e ∑ m! ⎥ ⎨1 − e
R1 ( NTU ) n=0 ⎣
− NTU − R1 ( NTU )
∑ ⎬
m =0 ⎦⎩ m =0 m! ⎭
(7.62)
144 Li-Zhi Zhang
Similarly, for moisture effectiveness, we have
1 ∞ ⎡ n
( NTUL ) m ⎤⎧ n
[R2 ( NTUL )] ⎫
m
εL = ∑ ⎢1 − e
R2 ( NTUL ) n=0 ⎣
− NTUL
∑ m! ⎥⎨1− e −R2 ( NTUL )
∑ ⎬
m=0 ⎦⎩ m=0 m! ⎭
(7.63)
We can see at this step that the moisture effectiveness has the same form of expression
with sensible effectiveness. The only differences are that NTU is replaced by NTUL and R1 is
replaced by R2.
For heat transfer, Eq.(7.62) is too complicated, since it has infinite series. Therefore,
Kays and London [1] used following empirical equation to represent the sensible
effectiveness as
⎡ exp(− R1 NTU 0.78 ) − 1⎤

ε s = 1 − exp ⎢ ⎥ (7.64)
⎣ R1 NTU −0.22 ⎦
Similar to the deduction of Eq.(7.64) for sensible heat transfer, the correlation for latent
effectiveness can be written as [3]
⎡ exp(− NTU L 0.78 R2 ) − 1⎤

ε L = 1 − exp ⎢ − 0.22 ⎥ (7.65)
⎣ NTU L R2 ⎦
NTU L = β ⋅ NTU (7.66)
R2 = m min / m max (7.67)
The value of relative humidity of plate in fresh air side, which is determined by latent
effectiveness (permeation rate), needs to be known before the calculation of diffusive
resistance for membrane rm and the diffusive to convective ratio α. Iterations are performed
to obtain a converged solution for φpf.
The latent effectiveness correlations for other flow arrangements, such as concurrent flow
and counter flow, can also be derived from those corresponding correlations for sensible
effectiveness, using the definition of Eqs.(7.66) and (7.67).
For concurrent flow,
1 − exp[ − NTU L (1 + R2 )]
εL = (7.68)
1 + R2
0.6
□ sensible
0.5 ○ Latent
Effectiveness
0.4
0.3
0.2
0.1
0
0 10 20 30 40
V (L/min)
Figure 7.2. Calculated and experimental sensible and latent effectiveness for a cross-flow membrane
enthalpy exchanger. Solid lines are calculated, discrete points are measured.
For counter flow,
1 − exp[− NTU L (1 − R2 )]
εL = (R2<1) (7.69)
1 − R2 exp[− NTU L (1 − R2 )]
NTU L
εL = (R2=1) (7.70)
1 + NTU L
The sensible effectiveness can also be calculated with above equation if NTUL is replaced
by NTU for every flow arrangement.
Validation
To demonstrate the suitability of the correlations in predicting the effectiveness, sensible

and latent effectiveness are calculated with the proposed correlations and compared with
experimental results in [5], as shown in Figure 7.2.
146 Li-Zhi Zhang
2.5
2
NTUL
1.5 C=0.1
C=1
1
C=10
0.5
0
0 0.2 0.4 0.6 0.8 1
◊
Figure 7.3. Variations of NTUL with relative humidity for different membranes, NTU=4.2.
From Figure 7.2, it is obvious that both the sensible and the latent effectiveness are
properly represented by the correlation from the present study. The largest discrepancies
between the predictions by the correlation and the experimental data result from the cases
with the smallest air flow rate, where the uncertainties of the experimental data are the
biggest. The average errors between the predicted and experimental results are 7.3% and
8.6% for sensible and latent effectiveness respectively.
For a given exchanger, the sensible effectiveness is a fixed value at the specified flow
rates. However the latent effectiveness will be affected by the two important dimensionless
factors proposed in this study: the ratio of diffusive to convective resistance (α), and the ratio
of total number of transfer units of moisture to that of sensible (β). The values of α and β are
in turn affected by the membrane material types and operating inlet conditions.
Figure 7.3 demonstrates the variations of NTUL, when NTU is kept constant, with
different inlet humidities. The value of NTUL decreases and increases for first-type (C<1) and
third-type (C>1) materials, with increasing inlet humidity respectively. The trends of resulting
latent effectiveness are the same as those of NTUL, which can be deduced from Eq.(7.30), see
Figure 7.4. For the linear type material (C=1), the NTUL and the latent effectiveness will not
change with the outside conditions. The number of transfer units for moisture would keep at
0.45 times of that for sensible heat for this material.
The above discussions suggest that an total heat exchanger with linear type membrane
cores always performs better than those with other membrane cores. For example, to obtain a
latent effectiveness of 0.6 under an inlet humidity of 50%, NTUL should be at least 2.0, which
means that the minimum values of NTU required for the exchanger are: 4.44 with linear type;
8 with third-type; and 11.1 with first-type membrane. A smaller NTU usually makes the total
heat exchanger more compact and cheaper.
0,6
C=0.1
Latent effectiveness ( πL )
0,5
C=1
0,4 C=10
0,3
0,2
0,1
0
0 0,2 0,4 0,6 0,8 1
Relative Humidity π
Figure 7.4. Latent effectiveness for three types of membranes, NTU=4.2.
It should be kept in mind that the calculation of plate moisture resistance by Eq.(7.26) is
rather complicated, though it discloses many material and operating conditions. In real
applications, to simplify analysis, moisture diffusivity is directly measured form permeation
test. In such cases, an equivalent mass transfer coefficient is obtained. Plate mass resistance is
written as
δ
rp = e
(7.71)
Dwp
where Dwpe is called the equivalent mass diffusivity (m2/s), which is calculated by humidity
ratio gradients between the two side of a plate, rather than by water uptake gradients. It has
already reflected properties like sorption, humidity, temperature, etc. The equivalent mass
transfer coefficient 1/rp is usually directly measured by permeation tests. Then the
performance of a total heat exchanger is estimated by ε–NTU correlations, by substituting the
obtained mass resistance rp in Eq.(7.25). Certainly accuracies will be comprised since the
operating conditions of permeation test are different from operating conditions of total heat
exchangers. However it’s simple.
Comparing Eqs(7.28), (7.29) and (7.71), we have
ρ p Dwp
e
Dwp = (7.72)
ρa ψ
148 Li-Zhi Zhang
7.3. CORRELATIONS FOR PLATE-FIN TOTAL HEAT EXCHANGER

For plate-fin total heat exchangers, heat transfer area on both sides of a plate is extended
by fins. Therefore the extended surface and changed heat transfer coefficients should be
considered. Heat conduction through the plate is in equilibrium with the convective heat
transfer on both sides. The equilibrium on a whole exchanger basis can be expressed by
Ap λ p
(hA)f (Tf − Tpf ) = (T pf − Tpe ) (7.73)
δ
Ap λ p
(hA)e (Te − Tpe ) = − (T pf − Tpe ) (7.74)
δ
where Ap is total transfer area (m2) of plates; λp is heat conductivity of plate (kWm-1K-1); δ is
thickness of plate (m); (hA)f is the fresh side convective heat transfer coefficient times total
fresh air side transfer area. The convective heat transfer coefficients in plate-fin ducts can be
obtained from previous sections. Total transfer area includes plate and fins area.
Moisture diffusion through the plate is in equilibrium with the convective mass transfer
on two surfaces. The equations can be expressed by
ρ p Ap Dwp
ρ a (kA)f (ω f − ω pf ) = (θ pf − θ pe ) (7.75)
δ
ρ p Ap Dwp
ρ a (kA)e (ω e − ω pe ) = − (θ pf − θ pe ) (7.76)
δ
where Dwp is water diffusivity in plate material (m2/s). If the equivalent moisture diffusivity
in plate Dwpe is used, then the above two equations can be re-written as
(kA)f (ω f − ω pf ) = (ω − ω pe )
Ap Dwp
pf (7.77)
δ
(kA)e (ω e − ω pe ) = − (ω − ω pe )
Ap Dwp
pf (7.78)
δ
The overall number of transfer units for sensible heat transfer and moisture transfer are
(UA)tot (UA)tot
NTU = =
(Gc )
p f (Gc ) p e
(7.79)
ρ a (U L A)tot ρ a (U L A)tot
NTU L = = (7.80)
(G )f (G )e
respectively, where
−1
⎛ Ap λ p ⎞
(UA) −1
= (hA)
−1
+ ⎜⎜ ⎟⎟ + (hA)e−1 (7.81)
⎝ δ
tot f
⎠
−1
⎛ Ap Dwp
e
⎞
(U L A) −1
tot
= (kA)
−1
f + ⎜
⎜ δ
⎟ + (kA)e−1
⎟
(7.82)
⎝ ⎠
Generally, the two streams have the same structure and exchanger area, therefore
Af = Ae = Ap + Afin (7.83)
The equivalent moisture diffusivity in plate Dwpe, is the same definition as Eq.(7.72). At
this step, the effectiveness can be calculated from ε–NTU correlations developed for parallel-
plates exchangers.
Table 7.2. Structural and physical parameters of the total heat exchanger
Name Symbol Unit Value

Number of channels for each flow n 115
Half duct height a mm 1
Half duct width b mm 2.5
Hydrodynamic diameter Dh mm 1.66
Exchanger length xF, yF mm 185
Plate thickness δ μm 55
Equivalent diffusivity, paper Dwpe m2/s 2.97e-7
Equivalent diffusivity, membrane Dwpe m2/s 8.92e-6
Heat conductivity, paper λp Wm-1K-1 0.38
Heat conductivity, membrane λp Wm-1K-1 0.44
Heat conductivity of air λa Wm-1K-1 0.0263
Density of air ρa kg/m3 1.18
Plate density, paper ρp kg/m3 860
Plate density, membrane ρp kg/m3 800
Moisture diffusivity in air Dva m2/s 2.82×10-5
Plate partition coefficient kP 0.58
Kinematic viscosity ν m2/s 15.89e-6
Volumetric air flow rates V m3/h 150
150 Li-Zhi Zhang
7.4. AN APPLICATION EXAMPLE
Problem
Considering two plate-fin total heat exchangers. The schematic is shown in Figure 4.2.
The geometries are the same and they are listed in Table 7.2. The first exchanger uses paper
as the plates, and the second one uses membranes as the plates. The measured equivalent
moisture diffusivity from permeation tests are: Dwpe=2.97e-7m2/s for paper and Dwpe=8.92e-
6m2/s for membrane. The task now is to estimate the sensible and latent effectiveness of the
two total heat exchangers.
Solution
Air mean velocity in the channels
V 150
ua = = = 0.98m / s
3600n(2a) x F 3600 * 115 * 0.002 * 0.185
Reynolds number
u a Dh 0.98 * 1.66e − 3
Re = = = 98.7 <<2300
ν 15.89e − 6
Therefore the flow is laminar flow. The problem is laminar flow in plate-fin ducts.
Duct aspect ratio
2a
τ= = 0.4
2b
Let’s first consider the paper exchanger. It uses paper as the plate and fin material. The
fin materials is the same as the plate material.
Fin heat conductance parameter
λf δ 0.38 * 55e − 6
Ωs = = = 0.40
λ a (2a ) 0.0263 * 0.002
Fin mass conductance parameter
δρ p Dwp k p δDwp
e
(55e − 6) * (2.97e − 7)
ΩL = = = = 2 .9 e − 4 ≈ 0
ρ a Dva (2a ) Dva (2a ) (2.82e − 5) * 0.002
As seen, the equivalent diffusivity considers the partition coefficient, density ratio, etc
inside.
ρ p Dwp k p ρ p Dwp
e
Dwp = =
ρa ρa ψ
Based on the calculated aspect ratio and fin conductance parameters, by referring to
Table 4.3, we have
Nu L = 0.95
ShL = 0.705
( f Re ) = 11.21
Nu L λ a 0.95 * 0.0263
h= = = 15.05Wm − 2 K −1
Dh 0.00166
ShL D va 0.705 * 2.82e − 5
k= = = 0.012m/s
Dh 0.00166
Friction coefficient
11.21
f = = 0.1136
98.7
Pressure drop
xF 0.185
ΔP = 2 f ρ a u a2 = 2 * 0.1136 * 1.18 * 0.98 2 = 28.7 Pa
Dh 0.00166
Each duct has two plates area on both sides, so the total plate area
Ap = 2nx F y F = 2 * 115 * 0.185 * 0.185 = 7.87 m 2
For the considered duct geometry, the fin to plate area ratio α = 1.29
Air side transfer area
Af = Ae = (1 + α ) * A p = 18.02m 2
Consequently we have
152 Li-Zhi Zhang
−1
⎞ ⎛ Ap λ p
(UA) = (hA)
−1
tot
⎟⎟ + (hA)e−1
−1
f + ⎜⎜
⎠ ⎝ δ
7.87 * 0.38e − 3 −1
= 2 × (0.01505 *18.02) −1 + ( ) = 7.39
55e − 6
(UA)tot = 0.135kWm −2
−1
⎛ Ap Dwp
e
⎞
(U L A) = (kA) + ⎜
−1
⎜ ⎟ + (kA)e−1
⎟
−1
⎝ δ ⎠
tot f
7.87 * 2.97e − 7 −1
== 2 × (0.012 * 18.02) −1 + ( ) = 32.78
55e − 6
(U L A)tot = 0.031m 3 /s
(UA)tot 0.135
NTU = = = 2.72
(Gc )
p f 1.18 * 150 / 3600 * 1.005
ρ a (U L A)tot 1.18 * 0.031
NTU L = = = 0.744
(G )f 1.18 * 150 / 3600
The calculated effectiveness with varying NTU are drawn in Figure 7.5 for a cross flow
heat exchanger, based on Eq.(7.37).
0.9
0.8
0.7
0.6
Effectiveness
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7
NTU
Figure 7.5. Effectiveness-NTU for a cross flow heat exchanger.

According to the figure, when NTU=2.72, the sensible effectiveness is
ε s = 0.67
When NTUL=0.744, the latent effectiveness is
ε L = 0.40
As seen, the latent effectiveness with paper is around 0.40.
Let’s consider the second exchanger. Based on the same procedure, when membrane is
used as the plate and fin materials, the results are
NTU=2.72, ε s = 0.67
NTUL=2.03, ε L = 0.62
Pressure drop is the same with the first exchanger.

As seen, the sensible effectiveness of membrane exchanger is the same as the paper
exchanger, but the membrane exchanger’s latent effectiveness is 55% higher than paper
exchanger. The main reason is that the moisture diffusivity in membranes is higher.
7.5. CONCLUSION
Heat and moisture transfer coefficients in total heat exchangers have been analyzed. The
ε–NTU method has been extended to total heat exchangers. Example shows that it is an
efficient yet simple way to estimate total heat exchanger performance. It also provides a
design tool for total heat exchangers, either in the form of parallel-plates or plate-fin ducts.
REFERENCES
[1] Kays, W.M.; Crawford, M.E. Convective heat and mass transfer. New York: McGraw-
Hill Inc.; 1993. pp.432-435.
[2] Incropera, F.P.; Dewitt, D.P. Fundamentals of heat and mass transfer, 3rd edn. New
York: Wiley; 1990. pp.416-420.
[3] Zhang, L.Z.; Niu, J.L. Effectiveness correlations for heat and moisture transfer
processes in an enthalpy exchanger with membrane cores. ASME Journal of Heat
Transfer, 2002, 122, 922-929.
[4] Niu, J.L.; Zhang,L.Z. Membrane-based enthalpy exchanger: material considerations
and clarification of moisture resistance. Journal of membrane science, 2001, 189, 179-
191.
154 Li-Zhi Zhang
[6] Taylor, R..; Krishna R.. Multicomponent mass transfer. New York: John Wiley & Sons, Inc.;
1993.
Chapter 8
NUMERICAL SIMULATION
OF TOTAL HEAT EXCHANGERS
ABSTRACT
Computer modeling has become en efficient tool to analyze total heat exchangers.
Correlations are simple, however, they cannot disclose the insight into the mechanisms of
heat and moisture transport. In total heat exchangers, heat and moisture are strongly
coupled, which makes performance analysis difficult. In this chapter, a detailed
mathematical modeling is provided both for the simple parallel-plates exchanger and for
the complicated plate-fin structure. Special efforts are spent in the methodology in
mathematical modeling.
NOMENCLATURE
a half duct height (m)
As cross section area (m2)
b half duct width (m)
C shape factor for the isotherm
cp specific heat (kJkg-1K-1)
D diffusivity (m2/s)
Dh Hydrodynamic diameter (m)
h convective heat transfer coefficient (kWm-2K-1)
K total transfer coefficient (kWm-2K-1 for heat and m/s for mass)
m v local emission rate (kgm-2s-1)
NTU Number of transfer units
Nu Nusselt number
P pressure (Pa)
156 Li-Zhi Zhang
Pe perimeter (m)
Re Reynolds number
RH air relative humidity
Sc Schmidt number
Sh Sherwood number
T temperature (K)
u velocity (m/s)
V volumetric flow rate (m3/s)
W water uptake (kg/kg)
Wmax maximum water uptake of membrane material (kg/kg)
x, y, z coordinates (m)
Greek letters
ε effectiveness
δ plate or fin thickness (m)
ρ density of dry air (kg/m3)
ξ dimensionless humidity
θ dimensionless temperature
μ dynamic viscosity (Pas)
λ thermal conductivity (kWm-1K-1)
Superscripts
* dimensionless
‘ exhaust air duct
Subscripts
1 fresh air side

2 exhaust air side
a air
b bulk
e exhaust
G geometric
f fresh
i inlet
L local, latent
m mass, mean, membrane
o outlet
p plate
s surface, sensible
tot total
v vapor
Numerical Simulation of Total Heat Exchangers 157
8.1. INTRODUCTION
Numerical simulation is an efficient tool to simulate heat mass transfer. In fact
computational modeling of membrane-related energy systems has been actively conducted
recently [1-9]. In previous chapter, the effectiveness-NTU correlations have been proposed
for total heat exchangers. Though it’s simple and convenient enough for engineers to evaluate
system performance, detailed analysis of the local heat and mass transfer in the exchanger
cannot be performed.
In this chapter, a detailed finite difference modeling of the coupled heat and moisture
transfer in total heat exchangers will be addressed. The effects of non-ideal boundary
conditions on the convective heat and mass transfer between the fluids and the solid surfaces
will be studied. The object of interest is a membrane-based cross-flow total heat exchanger, as
illustrated in Figure 8.1. As previously introduced [10], the device is like an air-to-air sensible
heat exchanger. But in place of traditional metal heat exchange plates, novel hydrophilic
membranes, which can exchange both heat and moisture simultaneously, are used as the heat
mass transfer media. The boundaries conditions for channels are neither uniform temperature
nor uniform concentration, but are naturally formed by heat mass coupling. Therefore the
established Nusselt and Sherwood numbers for parallel plates-channels are not used.
8.2. PARALLEL-PLATES EXCHANGER
Governing Equations
A schematic of the unit cell for model set up is shown in Figure 8.2. The fresh air and the
exhaust air flow through the passages in a cross flow arrangement. Geometries of ducts:
height 2a, width 2b, membrane thickness δ, fresh duct length xF, exhaust air duct length yF.
The fresh air is usually hot and humid and the exhaust air is usually cool and dry. Due to
symmetry and balanced flow, the equations governing the transport phenomena are the same
for the two passages. In the following, governing equations will be set up for the fresh air.
Then the different coordinate system for the exhaust flow will be given.
In most applications in air conditioning, Reynolds numbers are far below 2000, therefore
it can be assumed laminar flow. Other assumptions include: (1) Adsorption of water vapor
and membrane material is in equilibrium adsorption-state; (2) Both the heat conductivity and
the water diffusivity in the membrane are constants; (3) The heat of sorption is assumed
constant and equal to the heat of vaporization; (4) It is hydrodynamically fully developed, but
thermally and concentrationally developing; (5) The fluid is Newtonian with constant thermal
properties. As seen, this investigation uses heat mass coupling to find the boundary
conditions, rather than by assuming uniform temperature or concentration conditions.
158 Li-Zhi Zhang
Membrane Exhaust out
Fresh out
Fresh in
Exhaust in
Figure 8.1. A cross-flow membrane-based total heat exchanger.
Dimensionless equations governing momentum, energy and mass conservation in the

duct are written by [11,12]
∂ 2u * ⎛b⎞ ∂ u
2
2 *
4b 2
+⎜ ⎟ + =0 (8.1)
⎝ a ⎠ ∂y *
2 2
∂x * Dh2
∂θ ∂ 2θ ⎛ b ⎞ ∂ 2θ
2
U * = +⎜ ⎟ (8.2)
∂z h ∂x * 2 ⎝ a ⎠ ∂y * 2
∂ξ ∂ 2ξ ⎛ b ⎞ ∂ 2ξ
2
U * = +⎜ ⎟ (8.3)
∂z m ∂x * 2 ⎝ a ⎠ ∂y * 2
where the dimensionless velocity
μu
u* = − (8.4)
(dP / dz ) Dh2
Dimensionless temperature
T − Tei
θ= (8.5)
Tfi − Tei
Dimensionless humidity
ω − ω ei
ξ= (8.6)
ω fi − ω ei
Fresh air in
x Exhaust air in Membrane
Figure 8.2. The coordinate system of the unit showing one membrane and two neighboring flow
passages.
where T is temperature and ω is humidity. Subscripts “fi” and “ei” denotes fresh air in and
exhaust air in respectively.
x
x* = (8.7)
2b
y
y* = (8.8)
2a
z
z h* = (8.9)
Dh RePr
z
z m* = (8.10)
Dh ReSc
Hydraulic diameter
4 As
Dh = (8.11)
Pe
where As is the cross section area of the duct (m2)

160 Li-Zhi Zhang
, Pe is the perimeter of the duct (m). Pr is Prandtl number, and Sc is Schmidt number.
The dimensionless axis z* has a different definition for heat transfer and mass transfer.
In Eqs.(8.2) and (8.3), velocity coefficient U is defined by
u * 4b 2
U= (8.12)
u m* Dh2
∫∫ u dA
*
u *
m = (8.13)
As
The characteristics of fluid flow in the duct can be represented by the product of the
friction factor and the Reynolds number as
⎛ dP ⎞
⎜ Dh ⎟⎛ ρu D ⎞ 2
⎜
( fRe) = − dz ⎟⎜ m h ⎟⎟ = *
⎜ 1 2 ⎟⎜⎝ μ
(8.14)
⎜ ρu m ⎟ ⎠ um
⎝ 2 ⎠
Dimensionless bulk temperature
∫∫ u θdA
*
θb (z *
)= (8.15)
∫∫ u dA
h *
Dimensionless bulk humidity
∫∫ u ξdA
*
ξb (z *
)= (8.16)
∫∫ u dA
m *
Nusselt number
hDh
Nu = (8.17)
λ
Sherwood number
kDh
Sh = (8.18)
Dva
where h and k are convective heat transfer coefficient (kWm-2K-1) and convective mass
transfer coefficient (m/s) between fluid and wall, respectively. The local Nusselt and
Sherwood numbers may change from point to point on a duct surface. It is more practical to
evaluate the peripherally mean local Nusselt and Sherwood numbers along the duct.
An energy balance in a control volume of length Δz [12] helps to deduce the peripherally
mean local heat transfer coefficient
1 λ Δθ b
hL = − (8.19)
4(θ w − θ b ) Dh Δz h*
and the peripherally local Nusselt number
1 dθ b
Nu L = − (8.20)
4(θ w − θ b ) dz h*
Average Nusselt number from 0 to zh*
1 z h*
Nu m =
z h* ∫
0
Nu L dz h* (8.21)
Similarly, for mass balance we have
1 Dva Δξ b
kL = − (8.22)
4(ξ w − ξ b ) Dh Δz m*
1 dξ b
ShL = − (8.23)
4(ξ w − ξ b ) dz m*
1 *
zm
Shm =
z m* ∫
0
ShL dz m* (8.24)
As will be discussed later, the local Nusselt number and the local Sherwood number
decrease asymptotically from very high values near the entrance of a tube to certain fully
developed values at the end of thermal entry length. Under uniform temperature (uniform
mass concentration) boundary condition, the fully developed value is denoted as NuT (ShT for
mass). Under uniform heat flux (mass flux) boundary condition, it is denoted as NuH (ShH ).
In this study, under the real convective flow boundary conditions, the fully developed value is
denoted as NuC (ShC for mass transfer).
Inlet and boundary conditions. The governing equations for the exhaust air stream are in
the same form with the fresh air, however the coordinate systems are different. If we use x’*,
162 Li-Zhi Zhang
y’*, z’ to replace the coordinates x*, y*, z, in Eqs.(8.1)-(8.3), and (8.9)-(8.10), respectively,
then we get the equations for the exhaust stream. The relations of the two coordinates systems
are
⎧ x '* = z *
⎪ *
⎨ y' = y
*
(8.25)
⎪ z ' = 2b − x
⎩
Inlet conditions for fresh air
z h* = 0 , θ = 1 (8.26)
z m* = 0 , ξ = 1 (8.27)
Inlet conditions for exhaust air
z h* = 0 , θ = 0 (8.28)
z m* = 0 , ξ = 0 (8.29)
Boundary conditions of two air streams for velocity
u*=0, on all walls of the duct (8.30)
Adiabatic boundary conditions for fresh air
∂θ ∂ξ
x * = 0 or x * = 1 , = * =0 (8.31)
∂x *
∂x
and for exhaust air
∂θ ∂ξ
x'* = 0 or x'* = 1 , = * =0 (8.32)
∂x' *
∂x'
The boundary conditions on other two tube wall surfaces must be obtained numerically,
due to the interactions of temperature and humidity between the air streams and membrane
materials.
Temperature boundary conditions on membrane surfaces, for fresh air
( )
y * = 0 or y * = 1 , θ x * , z G* = θ m1 (8.33)
For exhaust air
( )
y '* = 0 or y '* = 1 , θ x'* , z '*G = θ m 2 (8.34)
( )
where term θ x , z G refers to air temperature adjacent to membrane surface at point (x*,
* *
zG*), subscript “m” refers to membrane, and “1” and “2” refer to fresh side and exhaust side
at the same point of membrane, respectively. Variable zG* denotes membrane geometric
position
z
. zG =
*
(8.35)
2b
Due to the small thickness in membrane (around 100μm, measured thermal conductivity
0.127Wm-1K-1), temperature differences between the two sides of a membrane are rather
small. Heat released on the adsorption side of the membrane (fresh air side) could be balanced
by the heat absorbed on the desorption side (exhaust side) of the membrane [13]. In fact,
previous investigations disclosed that the temperature differences are in the order of 10-4 °C
[14], meaning that Eqs.(8.33), (8.34) can be expressed by a single equation by
θ m1 = θ m2 = θ m (8.36)
The value of θm is not a fixed value. Rather, it is a result of local couplings between the
two streams on membrane surface. It should be also noted that in some cases when the solid
plate is thick (such as dehumidification process with a sorbent plate), temperature difference
between the two surfaces may be large. Under such conditions, heat transfer equations in the
solid should be set up to account for the thickness effect and the adsorption-desorption
thermal effect [15].
Mass boundary conditions on membrane surfaces, for fresh air
( )
y * = 0 or y * = 1 , q x * , z G* = m v (8.37)
For exhaust air
( )
y '* = 0 or y '* = 1 , q x'* , z '*G = m v (8.38)
v (kgm-2s-1) is the moisture emission rate through membrane at point (x*, zG*), and it
where m
is determined by diffusion equation in membrane as
Wm1 − Wm 2
m v = ρ m Dvm (8.39)
δ
164 Li-Zhi Zhang
where ρm and Dvm are density of membrane and moisture diffusivity in membrane,
respectively. Variable W is water uptake in membrane (kg moisture/kg dry membrane), and it
is expressed by a general sorption equation as
Wmax
W = (8.40)
1 − C + C / RH
where Wmax is the maximum water uptake of membrane material (kg/kg); C is a constant
named the shape factor for the material; RH is air relative humidity. This sorption curve
directly links water content to RH, a variable can be directly measured.
As seen from Eqs.(8.37), (8.38), mass transfer in the fresh air and in the exhaust air are
coupled together by moisture emissions through membrane.
The relative humidity is calculated by humidity ratio and temperature as [15]
RH e 5294 / T
= (8.41)
ω 10 6
where T is in K.
Moisture emission on membrane surface on fresh air or exhaust air side is also calculated
by
∂ω
m v = − ρ a Dva (8.42)
∂y y =0, 2 a
Emission rate q is not a constant value either. However, the values are attainable from
couplings of Eqs.(8.37)-(8.42), plus governing equations (8.1)-(8.3).
Numerical Methods
Numerical methods are always important for mathematical modeling. The objective of
the current numerical work is to find a solution for the model. To fulfill this task, two steps
should be implemented: (1) the solution of the governing equations for various components,
fresh air, exhaust air, and membrane; and (2) the couplings of different components to find
the values on membrane boundary surfaces.
The partial differential equations for momentum, energy, and mass transport, Eqs.(8.5)-
(8.7), are discretized by means of a finite volume method. According to this technique, during
the calculation values are stored only at the control volume (CV) centers (grid points). The
values at all CV boundaries must be expressed. For the convective terms, an upwind
differencing scheme is used, by replacing the boundary value by the values of the
neighbouring upstream CV. For the Peclet numbers involved in the present problem, this
approximation is close to reality. The diffusive fluxes at the CV boundary are approximated
by assuming a linear variation of variable between two grid points. With this procedure, an
algebraic tri-diagonal matrix will be formed to find the values at each node.
Table 8.1. Properties of the membrane used
Properties Unit Values

ρm kg/m3 856
Dvm m2/s 3.4×10-11
δ μm 105
Wmax kg/kg 0.28
C 0.85
There are totally three components: fresh air, exhaust air, and membrane, to be
calculated. When solving equations for one component, the values in other components are
assumed already-known values and will give assumed boundary conditions to this
component. The solution of three components is repeated until the solutions are converged,
i.e., differences of old and new temperature and humidity values at each grid node are within
a lower limit.
After these procedures, all the governing equations are satisfied simultaneously. The
temperature and humidity fields on membrane surfaces are the obtained.
Characteristics of the membrane used for the modeled total heat exchanger are listed in
Table 8.1. The nominal operating conditions: fresh air inlet 35°C and 0.025kg/kg; exhaust air
inlet 25°C and 0.010kg/kg.
After the inlet and outlet temperature and humidity are calculated, the sensible exchange
effectiveness and the latent exchange effectiveness can be calculated for a balanced flow
Tfi − Tfo
εs = (8.43)
Tfi − Tei
ω fi − ω fo
εL = (8.44)
ω fi − ω ei
The outlet temperature and humidity could also be calculated from the model. It is
obvious that the sensible effectiveness is equal to the dimensionless outlet temperature of
exhaust air, while the latent effectiveness is equal to the dimensionless outlet humidity of
exhaust air. It should be also noted that since there is only one membrane in the exchanger,
the boundary conditions at y*=1 should be modified to have zero heat and mass fluxes.
Experimental Work
An experimental set up has been built in the laboratory of SCUT to study the heat and
mass transport in a membrane based cross flow exchanger. This is a single-plate small scale
total heat exchanger. The whole set-up is shown in Figure 8.3. Ambient air is humidified and
166 Li-Zhi Zhang
is driven to a heating/cooling coil in a hot/cool water bath. The cooling coil can also act as
dehumidifier when necessary. After the temperature and humidity reach the set points, it is
then drawn through the exchanger for heat and moisture exchange. Another flow is driven
directly from ambient to the exchanger as the second flow. A Cellulose Acetate membrane is
sandwiched by two stainless steel shell. Two air passages on both sides of membrane are
formed, which is like a one-plate plate-and-shell heat exchanger. A schematic of the
exchanger is shown in Figure 8.4. Photograph of the real two half shells is shown in Figure
8.5. In the test, a 10mm thick insulation layer is placed on the inner surface of the shell to
prevent heat dissipation from the shell to the surroundings. After the exchanger and tubes are
installed, additional insulation is added on the outside surfaces to minimize heat losses from
the unit. The dimensions of the air passages formed by the shell inner insulation and the
membrane are: 2b=10cm, 2a=2mm.
Ambient
Stream 2 in
Stream 1 in Flow Meter
Ambient Stream 1 out
Hot/cold water bath Vacuum Pump

Ambient Valve Heat Mass Exchanger
Humidifier Stream 2 out

Temperature and
Humidity Sensor
Figure 8.3. Experimental set-up.
H a lf s h e ll
2a
A A
M em b ran e

2b
2b
A−A
Figure 8.4. Structures of the exchanger comprised by two symmetric half cells.
Figure 8.5. Photo of the two stainless steel half shells.
0.7
◇ Sensible, calculated
0.6 △ Sensible, experimental
□ Latent, calculated
0.5 ○ Latent, experimental
Effectiveness
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25 30
V (L/min)
Figure 8.6. Sensible and latent effectiveness of the exchanger, experimental and calculated.
168 Li-Zhi Zhang
Exhaust in
1 0.4
6 0 .3 1
0.39 0 .1 6 0 .0 9
0.2
4
0.3 0.1
0.
0.75 1 6
61
0. 0.4
0.76
54 6 0.3
9
0.
96
0.2
4
0.84
0.3
x*
0.5 1
0.91
0.
Fresh in 0. 0. 0.3
6
54 46 9
1
0. 7
0.
6
69
0.25
0.8
4
0.
0.
54
0.9
61
1
0
0 0.25 0.5 0.75 1
z*
zG*
Figure 8.7. Distribution of dimensionless surface temperature on membrane surfaces.
E x haust in
1 0 .3 9 0 .3 1
0 .5
0.7
4 0 .1 6
6
0.2
4
0.
46
0.3
0.84
1
0.
0.7 5 0.3
61
9
0.
96
0.
54
0. 7
0 .4
6
6
x*
0.5
F resh in
0.
61
0.8
0.
4
0.91
69
0.
54
0.2 5
0.
76
0.
6 1
0
0 0.25 0.5 0 .7 5 1
zz1G**
Figure 8.8. Distribution of dimensionless humidity on membrane surface on fresh air side.
Exhaust in
1 0.1 6 0. 09
0 .1 6
0.24
0.31
0.75
0.3
9 0.2
4
0. 5
0.3
1
4
0. 4
x*
0.5 6 0.3
9
Fresh in
0 .5
4
0.6
1 0 .4
0.25 6 0 .3
9
0. 6
9
0.76
0.
54
0
0 0.25 0.5 0.75 1
zzG1**
Figure 8.9. Distribution of dimensionless humidity on membrane surface on exhaust air side.
Exhaust in
1 7 .6
27 34.
8
2
.5
6.
.8 8 0
15
04 1
9
83.3
2
1.
59.
13
1
6
7.
10
0.75
Fresh in 8 3. 3
x*
0.5
1
59.
5 9 .1
0.25
34
.8
0
0 0.25 0.5 0.75 1
z G **
Figure 8.10. Distribution of heat flux on membrane surface (W/m2).

170 Li-Zhi Zhang
Exhaust in
1 -0 4
2 2E 04
E-
6.41
4 37 -04
E- 0 .
E-05
1
-04 1. 4
9 5E
E-0
61 E 1.2 04
E-
8.84
5
. 13
1
E-05
4
0.75 -0
0E
7.62
0
1.
5
E-0
8 . 84
05
E-
x*
0.5
1
E- 05
6.4
7. 62
Fresh in
6. 4
5
6.41E -0
1E
0.25
-05
5.19 E-05 5.19E-05

0
0 0.25 0.5 0.75 1
zGz**
G
Figure 8.11. Distribution of mass flux on membrane surface (kgm-2s-1).
14
13 NuL
V Num
12
Nusselt Numbers
V
11

10
V
9
V
V
8 V
7
6
0 0.05 0.1 0.15 0.2
zzh**
Figure 8.12. Local and mean Nusselt numbers along flow direction, b/a=50.
14
13 ShL
12 V Shm
Sherwood Numbers
11 V
10
9 V
V
8 V V
7
6
0 0.05 0.1 0.15 0.2 0.25
zzm**
Figure 8.13. Local and mean Sherwood numbers along flow direction, b/a=50.
During the experiment, air flow rate is changed, to have different Reynolds numbers.
Humidity, temperature, and volumetric flow rates are monitored at the inlet and outlet of the
exchanger. Equal air flow rates are kept for the two air streams. The uncertainties are:
temperature ±0.1ºC; humidity ±1%; volumetric flow rate ±0.5%.
The calculated sensible and latent effectiveness is plotted in Figure 8.6. They are
compared to the experimental data. The differences between the calculated and the tested are
less than 7%. The model is validated.
Membrane Surface Values
Contrary to ε–NTU correlations, with the model, besides the outlet variables, the
variables in the exchanger can also be calculated. The temperature, humidity, heat flux, and
moisture flux across the membrane are calculated and plotted in Figures 8.7-8.11. From these
figures, it can be shown that the variables on membrane surfaces are neither uniform values
nor uniform fluxes. The temperature, humidity, heat flux, and mass flux exhibit a 2
dimensional distributed nature. The contour values of heat and mass fluxes decrease along the
diagonal line like a ripple in a pool, whose center is at the point where the two flows intersect
and enter the exchanger. The contour lines of temperature and humidity are parallel to
diagonal line of the square membrane. The differences of humidity between the two surfaces
of the membrane are demonstrably large, indicating large internal moisture resistance.
Nusselt and Sherwood Numbers
Figures 8.12 and 8.13 show the local and mean Nusselt and Sherwood numbers along the
duct respectively. In the thermal and concentrational developing region, the local Nusselt and
Sherwood numbers decrease sharply, from very high values to 2 lower limiting values
172 Li-Zhi Zhang
denoted by NuC and ShC. The flow is called thermally or/and concentrationally developed
after that period. The calculated NuC and ShC for various aspect ratios are listed in Table 8.2.
It is found that ShC≈NuC for most aspect ratios, indicating that heat mass analogy is satisfied.
For comparison, the values of NuT and NuH from published references are also listed. It is
found that NuH is usually 20% larger than NuT, and NuC is generally less than NuH. It is
higher than NuT at larger aspect ratios, but 37% less than NuT at smaller aspect ratios. For
parallel plates that have large aspect ratios, NuC could be approximated by NuT. Further, for
hydrodynamically fully developed laminar flow in ducts, (fRe) is a constant. The calculated
values of (fRe) for various aspect ratios are also listed in Table 8.2. As seen, the current study
predicts the flow well.
It could be stressed that the moisture resistance through the membrane is larger than the
internal thermal resistance, consequently sensible effectiveness is 25% larger than the latent
effectiveness under current situations, though Nusselt numbers are almost equal to Sherwood
numbers in the passages.
Correlations for local and mean Nusselt numbers:
0.0668(Re Pr Dh / z )
Nu L = Nu C + (8.45)
1 + 0.045(Re Pr Dh / z )
2/3
The obtained boundary conditions are neither uniform flux nor uniform value boundary
conditions. In the thermal and concentrational developing region, the local Nusselt and
Sherwood numbers decrease sharply, from very high values to 2 lower limiting fully
developed values.
The fully developed Nusselt number NuC is generally less than NuH. It is higher than NuT
at larger aspect ratios, but 37% less than NuT at smaller aspect ratios. For parallel plates that
have large aspect ratios, NuC can be approximated by NuT.
Table 8.2. Fully developed (fRe) and Nusselt, Sherwood numbers for
various aspect ratios
b/a NuH NuT NuC ShC (fRe) (fRe)

Sources Ref [7,8] Ref [7,8] (This study) (This study) Ref [7,8] (this study)
1.0 3.61 2.98 1.89 1.89 57.0 56.51
1.43 3.73 3.08 2.51 2.52 59.0 58.03
2.0 4.12 3.39 3.09 3.10 62.0 61.82
3.0 4.79 3.96 4.08 4.00 69.0 68.05
4.0 5.33 4.44 4.65 4.52 73.0 72.42
8.0 6.49 5.60 6.08 6.03 82.0 81.63
50.0 7.74 7.81 92.37
100.0 8.03 8.05 93.96
∞ 8.23 7.54 96.0
The internal moisture diffusion resistance of the membrane is larger than the internal
thermal resistance, therefore sensible effectiveness is 25% larger than the latent effectiveness,
though Nusselt numbers are almost equal to Sherwood numbers in the passages.
8.3. PLATE-FIN EXCHANGER

From above simulations, it is found that the fully developed Nusselt numbers and
Sherwood numbers under naturally balanced boundary conditions from heat mass transfer can
be approximated by those under uniform temperature or concentration boundary conditions.
That will simplify modeling in future. In this part, we will extend the numerical simulations
to the more complicated plate-fin total heat exchangers. However, the boundary conditions
will be simpler. The convective heat mass transfer between the membrane and air streams
will use the fully developed Nusselt and Sherwood numbers developed for plate-fin ducts in
Section 4.
Figure 8.14. Schematic of a cross-flow plate-fin heat exchanger with sinusoidal passages.
2a
2b
Figure 8.15. Geometries of a single sinusoidal plate-fin duct.
The modeled structure is shown in Figure 8.14, which is a popular plate-fin sinusoidal
channels structure. Figure 8.15 shows the single duct geometry. As previously mentioned, it
174 Li-Zhi Zhang
is preferred because it has many virtues like it is stationary, compact, and easy to construct.
The differences between total heat exchangers and other air-to-air heat exchangers are that in
place of metal materials, current commercial total heat exchangers employ hygroscopic paper
as the material for fins and plates, to transfer both sensible heat and moisture simultaneously.
Heat and Mass Transfer in Air Streams
Actual heat mass transfer in the numerous channels is complicated. A meso-scopic model
is set up. Each channel cross section is represented by one temperature or humidity. The
temperature and humidity vary along flow directions (x for fresh air and y for exhaust air) and
the corresponding cross directions (y for fresh air and x for exhaust air) simultaneously. On
each channel cross section, though temperature and humidity are two-dimensionally different
locally, as studied for the parallel-plates channels, in this study for the whole exchanger, they
are represented by a lumped parameter for each channel cross section. It can be considered as
a semi-lumped parameter model. The two air streams, one hot and humid (fresh air), and the
other cool and dry (exhaust air), exchange both sensible heat and moisture simultaneously in
the exchanger in a cross flow arrangement. Two-dimensional heat mass transfer model can be
set up to govern the energy and mass conservations in the two air streams [15]:
∂Tf*
= NTU sf (Tmf* − Tf* ) (8.46)
∂x *
∂Te*
= NTU se (Tme
*
− Te* ) (8.47)
∂y *
∂ω *f
= NTU Lf (ω *mf − ω *f ) (8.48)
∂x *
∂ω *e
= NTU Le (ω *me − ω *e ) (8.49)
∂y *
where x is flow direction for fresh stream and y is flow direction for exhaust stream.
The dimensionless temperature and humidity are defined by
T − Tei
T* = (8.50)
Tfi − Tei
ω − ω ei
ω* = (8.51)
ω fi − ω ei
The dimensionless coordinates are defined by
x
x* = (8.52)
xF
y
y* = (8.53)
yF
where xF and yF are channel lengths for fresh air and exhaust air (m). Here xF=yF. The air side
number of transfer units for heat and moisture are defined by
(hA)f
NTU sf =
(Gc )
p f
(8.54)
(hA)e
NTU se =
(Gc )
p e
(8.55)
(ρ a kA)f
NTU Lf = (8.56)
(G )f
(ρ a kA)e
NTU Le = (8.57)
(G )e
where k and h are air side convective mass transfer coefficient (m/s) and convective heat
transfer coefficient (kWm-2s-1) respectively; G is air mass flow rate (kg/s); A is total transfer
area including plates and fins (m2) for each stream; cp is specific heat (kJkg-1K-1). Subscripts
“f” refers to fresh side and “e” refers to exhaust side; “s” refers to sensible and “L” refers to
latent; “mf” refers to membrane surface on fresh side, and “me” refers to membrane surface
on exhaust side.
The outlet temperature and humidity values are calculated by Eqs.(8.46) to (8.49). Then
it’s convenient to calculate sensible and latent effectiveness from Eqs.(8.43) and (8.44).
Humidity ratios can be converted to relative humidity from psychrometric chart.
The convective heat transfer coefficient and mass transfer coefficient can be calculated
by
hDh
Nu = (8.58)
λa
kDh
Sh = (8.59)
Dva
176 Li-Zhi Zhang
where Dva is vapor diffusivity in air (m2/s), Dh is the hydrodynamic diameter (m). For plate-
fin channels of finite fin conductance, the fully developed Nusselt and Sherwood numbers are
influenced by aspect ratios (a/b), and fin conductance parameters, as calculated in Chapter 6.
This is quite different from the simple classical data of a sensible-only heat exchanger with
infinite fin conductance [17,18].
Heat and Mass Transfer through Plates
Heat conduction through the plate is in equilibrium with the convective heat transfer on
both sides. A meso-scale approach is employed. Whether it’s paper or membrane, the
equilibrium can be expressed by
Ap λ p
(hA)f (Tf − Tpf ) = (Tpf − Tpe ) (8.60)
δ
Aλ
(hA)e (Te − Tpe ) = − p p (Tpf − Tpe ) (8.61)
δ
where Ap is total transfer area (m2) of plates; λp is heat conductivity of plate (kWm-1K-1); δ is
thickness of plate (m); (hA)f is the fresh side convective heat transfer coefficient times total
fresh air side transfer area. The convective heat transfer coefficients in plate-fin ducts can be
obtained from previous chapters. Total transfer area includes plate and fins area.
Moisture diffusion through the plate is in equilibrium with the convective mass transfer
on two surfaces. The equations can be expressed by
ρ p Ap Dvp
ρ a (kA)f (ω f − ω pf ) = (θ pf − θ pe ) (8.62)
δ
ρ p Ap Dvp
ρ a (kA)e (ω e − ω pe ) = − (θ pf − θ pe ) (8.63)
δ
where Dvp is moisture diffusivity in plate material (m2/s). If the equivalent moisture
diffusivity in plate Dvpe is used, then the above two equations can be re-written as
(kA)f (ω f − ω pf ) = (ω − ω pe )
Ap Dvp
pf (8.64)
δ
(kA)e (ω e − ω pe ) = − (ω − ω pe )
Ap Dvp
pf (8.65)
δ
Moisture emission rate through the plate from the fresh air to the exhaust air
ρ a Dvp
e
m v = (ω − ω pe ) (8.66)
δ
pf
In fact, the final sensible effectiveness and the latent effectiveness can be estimated from
the total number of transfer units with established correlations [7,8]. However, to know the
details of heat and moisture transfer in the exchanger, detailed equations should be solved.
Boundary Conditions
Fresh:
Tf* x* = 0
=1 (8.67)
ω *f x* = 0
=1 (8.68)
Exhaust:
Te* x* =1
=0 (8.69)
ω*e x* =1
=0 (8.70)
Table 8.3. Structural and physical parameters of the total heat exchanger
Symbol Unit Value Symbol Unit Value

n 116 A mm 1
xF, yF mm 185 B mm 2.4
δ μm 105 kP 0.59
Dvpe, paper m2/s 2.97e-7 Da 2
m /s 2.82e-5
Dvpe, membrane m2/s 8.92e-6 λa Wm-1K-1 0.0263
λp, paper Wm K -1 -1
0.44 λp, membrane Wm K -1 -1
0.42
ρa kg/m 3
1.18 ρp, paper kg/m 3
860
ρp, membrane kg/m 3
810 RHfi 0.56
Tfi °C 35 RHei 0.52
Tei °C 25
178 Li-Zhi Zhang
Simulation Case
Consider two total heat exchangers. One is paper-plate and paper fin, called unit 1.
Another one is paper fin and membrane plate, called unit 2. The structural and basic physical
parameters of the two total heat exchangers are listed in Table 8.3.
Solution Procedure
A finite difference technique is used to discrete the partial differential equations

developed for the air streams. The calculating domain is divided into a number of discrete
nodes. Each node represents a control volume. The number of calculating node is 50 in x
direction. An upstream differencing scheme is used for two air streams. The two air streams
and the asymmetric membrane are closely coupled. Heat transfer and mass transfer are also
related to each other. Therefore iterative techniques are needed to solve these equations. A
description of the iterative procedure is as following: (1) assume initial temperature and
humidity fields in the two streams. (2) Calculate the temperature and humidity values on
membrane surfaces by Eqs. (8.60) through (8.65). (3) Taking the current values of
temperature and humidity on membrane surfaces as the default values, get the temperature
Heaters & A
Total heat
Humidifiers exchanger
Straightener
Nozzles & straightener
Blowers C
Shells
Fresh air
Exhaust air
exchanger
Sensors
C B
A
Figure 8.16. Experimental set up for commercial scale total heat exchanger.
and humidity profiles in two air streams by solving Eqs.(8.46) through (8.49). (4) Go to (2),
until the old values and the newly calculated values of temperature and humidity at all
calculating nodes are converged.
After these procedures, all the governing equations are solved simultaneously. To assure
the accuracy of the results presented, numerical tests were performed for the duct to
determine the effects of the grid size. It indicates that 50 grids are adequate (less than 0.1%
difference compared with 80 grids). The final numerical uncertainty is 0.1%.
When the temperature and humidity fields in the exchanger are calculated, the sensible
and latent effectiveness are calculated using mean outlet values. These are numerically
obtained data.
Another experimental test rig has been set up in SCUT. This is a real scale test rig for the
measurement of sensible and latent effectiveness of commercial-scale total heat exchangers.
The test rig is shown in Figure 8.16.
The purpose of the experiment is to measure the steady state heat and moisture transfer
through the enthalpy exchanger cores, by the measurements of inlet and outlet temperature,
humidity and air flow rates. The sensible and latent effectiveness and pressure drop are the
performance indices.
Figure 8.17. The structure of the unit core which is inserted into the shell in the test rig in the test.
A schematic of the test-rig is shown in Figure 8.16. It shows the ducting work, air
controlling units, and instrumentation. Figure 8.17 shows the unit core structure. For the test
rig, two parallel air ducts with a 210×210mm cross section are assembled. Each duct is
comprised of a variable speed blower, a wind tunnel, a set of nozzles, wind straightenners,
180 Li-Zhi Zhang
electric heating coils, steam humidification tubes, temperature and humidity sensors. An
exchanger shell is designed to hold the core. The small converging wind tunnels produce
steady, homogeneous, fully developed air flow to the exchanger. The heating power and the
steam generation currents can be adjusted according to the set points temperature and
humidity. After the air temperature and humidity are adjusted to the set points, the two ducts
are connected to the two inlets of the exchanger shell respectively. The exchanger shell is
designed to station the exchanger core and separate the cross flow two air streams. The cores,
either all paper or paper-fin and membrane-plate, can be inserted into the quadrate cavity in
the center of the shell. The whole test rig is built in a constant temperature and constant
humidity room, so the inlet temperature and humidity can be controlled and maintained very
well even under very hot and humid ambient weather conditions. A 15mm thick plastic foam
insulation layer is pasted on the outer surfaces of the ducts and the shells to prevent heat
dissipation from the system to the surroundings. Moisture dissipation from air stream to the
surroundings is negligible since the duct and shell materials are highly hydrophobic and they
could adsorb little moisture. The heat loss from the system is below 0.6%, and moisture loss
is less than 0.5%.
Figure 8.18. Photo of the test rig for commercial scale total heat exchangers.
0.9
Sensible effectivenes
0.8
0.7
Unit 1
Unit 2
0.6
80 100 120 140 160 180 200 220
3
V (m /h)
Figure 8.19. Sensible effectiveness of the two plate-fin total heat exchangers under various air flow
rates.
The completed test rig is demonstrated in Figure 8.18. A software is designed to balance
the heat and moisture transfer in the whole test rig. The nominal operating conditions: fresh
air inlet 35°C and 0.021kg/kg; exhaust air inlet 27°C and 0.012kg/kg. The corresponding inlet
relative humidity (RH) is 59% and 54% for fresh air and exhaust air respectively. During the
experiment, equal air flow rates are kept for the two ducts. The design air flow rates are
150m3/h. In the test, they are changed by variable speed blowers, to have different air
velocities. Humidity, temperature, and volumetric flow rates are monitored at the inlet and
outlet of the exchanger. Before and after each test, temperature and humidity sensors are
calibrated with a Pt-100 temperature sensor and a chilled-mirror dew-point meter. Hot-wire
anemometers that are used to measure the wind speed before and after the exchanger are
compared with the air flow rates measured by nozzles. The offset is controlled to within 1%
limit. Volumetric air flow rates are varied from 80m3/h to 210m3/h, corresponding to frontal
air velocities from 0.27m/s to 0.78m3/h which are typical for commercial enthalpy
exchangers. Air flow under such conditions is laminar, with Reynolds numbers not exceeding
200. A digital pressure differential gauge is used to measure the pressure drop across the
tested core. The uncertainties are: temperature ±0.1ºC; humidity ±2%; volumetric flow rate
±1%, pressure drop ±1%. The final uncertainty is ±4.5% for sensible and latent effectiveness.
Six sensors are uniformly positioned on the inlet and outlet surfaces of the shell to have a
mean value of measurements. Due to the small channels in the core, outlet air from the core is
quite evenly distributed. In other words, the core itself has the effect of another ideal wind
straightener. Anyway, another 6 wind straighteners, which are made of plates with numerous
evenly distributed small holes drilled, are installed in the ducts before and after the shell and
nozzles to well distribute the wind. In addition, heat and mass balance between the fresh air
and the exhaust air are checked. Their differences are controlled to be less than 0.1%. From
these preparatory works, the test rig is considered to be reliable.
182 Li-Zhi Zhang
0.8
Unit 1
0.7 Unit 2
Latent effectiveness
0.6
0.5
0.4
0.3
0.2
80 100 120 140 160 180 200 220
3
V (m /h)
Figure 8.20. Latent effectiveness of the two plate-fin total heat exchangers under various air flow rates.
After the measurement of mean inlet and outlet temperature and humidity, the sensible
and latent effectiveness are calculated. They are the experimentally obtained data. Figure 8.19
and 8.20 show the calculated and tested sensible and latent effectiveness, respectively. The
membrane plate unit has a higher latent effectiveness than paper plate unit. The sensible
effectiveness is the same for the two exchangers. Membrane is superior to paper.
The model can disclose details inside the channel that the experiment cannot.
Temperature and humidity fields in the air streams are calculated for the nominal operating
conditions. The dimensionless temperature of the plate is shown in Figure 8.21. The unit is
paper-fin and membrane-plate.
As seen from these figures, the temperature of fresh air decreases along the flow in x
direction, while the temperature of the exhaust increases along the flow in y direction.
Because the two flows are in cross flow arrangement, the temperature profiles exhibit a two-
dimensional nature. The plate temperature is almost equal to the average temperature of the
fresh air and exhaust air, meaning little conductance resistance in plate.
The humidity on membrane surface in fresh air side is shown in Figure 8.22. To intensify
moisture transfer in enthalpy exchangers, it’s necessary to use plate materials that have high
moisture diffusivities.
The local moisture permeation rates through the membrane are shown in Figure 8.23 for
the paper unit and Figure 8.24 for the membrane unit. In both cases, the highest moisture
emission rates are located on the surfaces where the two air streams interconnect, since here
the driving force is the highest. It is also observed that the emission rates through the
membrane plate are far higher than those through the paper plate. This is just the reason that
the unit 2 has 60% higher latent effectiveness than unit 1.
1 9
20
59
39
0.
0.9
19
0. 8
7
0.
9
38
63
9
59
55
0
0.4 8
0.
0. 96
0.
0. 7
99
9 356
9
8
67
0.75 0.3 0.
0.7
51
0. .478
0.
39 8 0
0
0.8 0.8
9
59
0
0.
19
2
0.9
0.7
8
9
y*
9 43
63
0.5 55 0. 98 .356 10
59 0.6 9
0. 18
0.
5
0.3 0 0.3
0.7
79
5
0. 478
99 0. 68
0.7
0.2
9
23
0
88
0. 0.2 87
0.
0. 0
0. 39 19
9 76
0.25 83 0. 1
7
38 56 0
0 . 6 0.
0..4398 0.30.31 32
9
55
8 0 68 0.1
51 0.2
79 04. 78
0.6 0. 23 0.08
7
0.26
0. 1 7 0.041
0
0 0.25 0.5 0.75 1
x*
Figure 8.21. Dimensionless temperature of membrane temperature.
1
9
67
99
0.
9
55
0.7
0
0.
5
0.92
39
0.7
19
0. 8
9
0.7
63
8
47
0.
0
0.
8
0.75 8
0.8
51
0. 38
9
0. 4
55 59
0. 0.
79
0.6
56
0 799
0. 3
0.7 .759
0.
y*
39
98
19
0.5 8
9
47 0.3
63
0. 8
0.
0.
8
51
0. 38
0. 4
9 99
10
55 . 5
8
6 0.3 0.26
0. 0
5
0.3
79
0. 6
0.25 98
78 0.3 23 76
.4 0. 2 0.1
39
8 0 32
0. 6
51 0. 1
0. .438 10 8
0
56 0.3 0.26
0 0.3
0 0.25 0.5 0.75 1
x*
Figure 8.22. Dimensionless humidity of membrane surface on fresh air side.
184 Li-Zhi Zhang
1 7.598E
-06
7.971E
- 06
8.343E-
06
0.75 8.715E-06
9.088E-06
9.460E
-06
9.833E-06
y*
0.5
1.020
1.058E-05 E-05
1.1
0.25 1.207E-05 3 2E
-05
1.318E-05
1.430
E-05
0
0 0.25 0.5 0.75 1
x*
Figure 8.23. Emission rate on plate surface for core 1 of paper plate (kgm-2s-1).
1
-05
. 124E
1
0.75 1.390E-05
1.656E-05
y*
0.5
2.187E-05
5
0E-0
0.25 2.187E-05
1.39
4. 0
2.7
47
E-
18
05
E-
05
2.984E-05
0
0 0.25 0.5 0.75 1
x*
Figure 8.24. Emission rate on plate surface for core 2 of membrane plate.
8.4. CONCLUSION
Detailed mathematical models have been set up for analysis of coupled heat and moisture
transfer in total heat exchangers. Two structures are considered: parallel-plates and plate-fin
ducts. Due to the couplings between heat and mass transfer, iterative numerical schemes are
required. Though the procedures are complicated, they provide a tool for the detailed analysis
of the inside heat mass transfer in the exchanger.
Both Temperature and humidity exhibit a two dimensional nature. The higher the
moisture diffusivity is, the higher the latent effectiveness is. Simulated results consider all the
material and operating conditions. They are more accurate.
REFERENCES
[1] Al-Sharqawi, H.; Lior, N. Conjugate computation of transient flow and heat and mass
transfer between humid air and desiccant plates and channels. Numerical Heat Transfer
Part A-Applications, 2004, 46, 525-548.
[2] Saman, W.Y.; Alizadeh, S. Modelling and performance analysis of a cross-flow type
plate heat exchanger for dehumidification/cooling. Solar Energy, 2001, 70, 361-372.
[3] Yeh, H.M.; Chen, Y.K. Membrane extraction through cross-flow rectangular modules.
Journal of Membrane Science, 2000, 170, 235-242.
[4] Dindore, V.Y.; Brilman, D.W.F.; Versteeg, G.F. Modelling of cross-flow membrane
contactors: physical mass transfer processes. Journal of Membrane Science, 2005, 251,
209-222.
[5] Hagg, M.B. Membranes in chemical processing - A review of applications and novel
developments. Separation and purification methods, 1998, 27, 51-168.
[6] Incropera, F.P.; Dewitt, D.P. Introduction to Heat Transfer, 3rd edn. New York: John
Wiley & Sons; 1996. Chapter 8, pp. 416.
[7] Shah, R. K.; London, A.L. Laminar flow forced convection in ducts. New York:
[8] Kays, W.M.; Crawford, M.E. Convective Heat and Mass Transfer. New York:
[9] Mengual, J.I.; Khayet, M.; Godino, M.P. Heat and mass transfer in vacuum membrane
distillation. International Journal of Heat and Mass Transfer, 2004, 47, 865-875.
[10] Zhang, L.Z.; Niu, J.L. Energy requirements for conditioning fresh air and the long-term
2001, 26, 119-135.
[11] Ebadian, M.A.; Zhang, H.Y. Fluid flow and heat transfer in the crescent-shaped lumen
catheter. ASME Journal of Applied Mechanics, 1993, 60, 721-727.
confined by sinusoidal and arc curves. International Journal of Heat Mass Transfer,
2001, 45, 571-578.
[13] Favre, E. Temperature polarization in pervaporation. Desalination, 2003, 154, 129-138.
[14] Zhang, L.Z.; Jiang, Y.; Zhang, Y.P. Membrane-based Humidity Pump: performance
and limitations. Journal of membrane science, 2000, 171, 207-216.
186 Li-Zhi Zhang
[15] Niu, J.L.; Zhang, L.Z. Membrane-based enthalpy exchanger: material considerations
and clarification of moisture resistance. Journal of Membrane Science, 2001, 189, 179-
191.
[16] Zhang, L.Z. Heat and mass transfer in a cross flow membrane-based enthalpy
exchanger under naturally formed boundary conditions. International Journal of Heat
Mass Transfer, 2007, 50, 151-162.
[17] Zhang, L.Z. Thermally developing forced convection and heat transfer in rectangular
plate-fin passages under uniform plate temperature. Numerical Heat Transfer, Part A-
Applications, 2007, 52, 549-564.
[18] Zhang, L.Z. Laminar flow and heat transfer in plate-fin triangular ducts in thermally
developing entry region. International Journal of Heat Mass Transfer, 2007, 50, 1637-
1640.
Chapter 9
NOVEL MEMBRANES
FOR TOTAL HEAT EXCHANGER
ABSTRACT
Stationary total heat exchangers use novel vapor-permeable membranes for
simultaneous heat and moisture transfer. To improve performance, air side intensification
and material side intensification should be taken simultaneously. Novel membranes that
have high vapor permeability are the key factor to the success of commercial total heat
exchangers. In this chapter, two novel membranes are developed for total heat
exchangers. They are composite supported liquid membrane and hudrophobic-hydriphilic
composite membrane. Their performances and characterization are conducted.
NOMENCLATURE
A area (m2)
c volumetric concentration of CO2 (m3 CO2/m3 air)
C Concentration (KG/M3)
D diffusivity (m2/s)
Dh hydrodynamic diameter (m)
dp pore diameter (m)
F gas flow through a single pore (kg/s)
G flow coefficient [kgm/(sPa)]
Hd duct height at inlet (m)
J emission rate (kgm-2s-1)
kB Boltzmann constant (1.38×10-23 J/K)
Kn Knudsen number
kp Henry constant (kgm-3Pa-1)
K total transfer coefficient (m/s)
L height of air gap (m)
m molality of electrolyte (mol LiCl /kg water)
188 Li-Zhi Zhang
M molecule weight (kg/mol)

N number of pores per unit area (m-1)
NTU Number of Transfer Units
p partial pressure (Pa)
p0 Saturation pressure (Pa)
P total pressure (Pa)
Pe permeability (kgm-1s-1)/(kg/kg)
q heat flux (kWm-2)
R gas constant, 8.314 J/(mol K)
r radius coordinate (m); resistance (m2s/kg, or s/m)
r0 cell radius (m)
Sh Sherwood number
T temperature (K)
ua air bulk velocity along radius (m/s)
V air flow rate (L/min)
v molecular diffusion volume
x mass fraction of solute (kg LiCl/kg solution)
Greek Letters
φ relative humidity
τ pore tortuosity
ε porosity
δ thickness (μm)
λ heat conductivity (kWm-1K-1), mean free path (m)
η viscosity (Pas)
ψ Effectiveness
σi molecular collision diameter (m)
σp geometric standard deviation
μp mean pore diameter (m)
γ resistance (m2s/kg)
ρ density (kg/m3)
α selectivity
Superscripts
* dimensionless
Subscripts
a air
D air stream
e effective
i inlet
Novel Membranes for Total Heat Exchanger 189
k Knudsen
l liquid
L lower chamber
m mean
o outlet, ordinary
p Poisseuille
s solid
sol solution
v vapor
w liquid water
9.1. INTRODUCTION
Air dehumidification is a major task in air conditioning in hot humid regions [1,2]. With
the developments in membrane technology, polymer membranes have been used in air
dehumidification. Hydrophilic polymer membranes that are permeable to vapor, but
impermeable to air, have been considered. Various materials have been tested: for instance
Nafion [3,4], regenerated cellulose [5], Cellulose triacetate [6], sulfonated poly(phenylene
oxide) [7], polyether-polyurethane [8] siloxane-amide copolymer [9], polystyrene-sulfonate
[10], polyvinylidene fluoride and polyethersulfone [11], and cellophane [12]. They are dense
solid membranes.
Due to the similarity in vapor permeation mechanisms, recently, membranes that were
used for air dehumidification, are used for total heat exchangers [13,14]. However, air
dehumidification is driven by trans-membrane pressure gradients higher than several bars,
while the trans-membrane partial pressure difference in a total heat exchanger is less than
2kPa. In other words, pressure driving force in a total heat exchanger is only 1/50 of the
traditional air dehumidification. Consequently, membrane-based total heat exchangers with
common dehumidification membranes have very low vapor permeability, and a low latent
effectiveness. Besides, they have one shortcoming: they are expensive.
Membrane
PP Net
Figure 9.1. Composite hydrophobic-hydrophilic membrane structure.
Vapor diffusion in common dense membranes is rather small. Recently there have been
great developments in membranes for total heat exchanger. Among the various novel
inventions, composite hydrophobic-hydrophilic membrane [15] and composite supported
liquid membrane [16] are two categories.
190 Li-Zhi Zhang
9.2. HYDROPHOBIC-HYDROPHILIC COMPOSITE MEMBRANE

Vapor diffusion in dense layers is rather small. To increase vapor permeation rates,
composite membranes have been used. According to this scheme, a thin active layer is cast
onto a thick porous PP (polypropylene) support layer or other materials. The porous support
layer provided the necessary mechanical strength while the thin active layer provided the
permselective separating effect. The permeation rates can be greatly improved due to the
reduction in resistance.
It is accepted that vapor permeations through dense membranes are based on solution-
diffusion mechanism. The more hydrophilic the material is, the more moisture it can adsorb,
and more moisture can permeate through membrane. According to this theory, materials that
have large quantities of hydrophilic groups such as -SO3H, -NH2, -COOH, -OH are required
to have a strong hydrophilicity.
Cellulose triacetate (CTA) is a material that is very hydrophilic. It has good membrane-
forming properties, good chemical and thermal stability. Furthermore, most importantly it is
cheap, which is the prerequisite condition for commercial applications. On the other hand,
LiCl salts are very hydrophilic since they can build hydrogen bonds with water molecules.
When dispersed in the CTA membrane, they can off-set the effects of CTA cross-linking and
make the membrane very hydrophilic and thus facilitate the transport of water vapor.
Membrane Preparation
The composite membrane is formed by coating CTA casting solution onto the PP support
membrane. The fabrication process is comprised of the following three steps:
Formulation of casting solution. CTA powder is weighted and placed into a vessel with
Acetone at about 90 . The solution is heated and stirred until it is completely dissolved. It
took about 2 hours.
A certain amount of cross-linking agent (eg. L-malic acid), catalyst (eg. glacial acetic
acid) and additive (LiCl) is added to the solution. The solution is continuously stirred at 70
until these different compositions are completely dissolved to form a homogeneous solution.
The solution is cool down and placed still for de-bubbling. It took 2 hrs.
The polymer solution is coated on the PP support membrane. The thickness of the active
layer is controlled by a casting knife. The thickness of the final active layer can be from
several microns to a dozen microns, depending on the gap between the knife edge and the
support layer. Then the asymmetric membrane is placed into a vacuum drying oven for cross-
linking at 100°C for 1 hour. The membrane is further dried at 60°C for 2 hours.
The composition of the casting solution is listed in Table 9.1. L-malic acid is the cross-
linker. Glacial acetic acid is catalyst. LiCl is used as additive to increase hydrophilicity.
Acetone is used as the solvent for CTA. For a comparison, same thickness of CTA layer is
ensured for different membranes.
Table 9.1. The composition of the casting solution for active layer
Composition of CTA casting solution

Membrane
Acetone L-malic glacial acetic LiCl
Symbol CTA (g)
(g) acid (g) acid (drops) (g)
CMB 8 92 0 0 0
CM0 8 92 4 1 0
CM1 8 92 4 1 0.5
CM2 8 92 4 1 1
CM3 8 92 4 1 1.5
CM4 8 92 4 1 2
CM5 8 92 4 1 2.5
Ambient
Exhaust in
Valve
Fresh in Flow Meter
Ambient Fresh out
Hot/cold water bath Vacuum Pump

Ambient Total heat exchanger
Humidifier Exhaust out

Temperature and
Humidity Sensor
Figure 9.2. Experimental set-up for membrane vapor permeation.
Membrane Half shell
Figure 9.3. Structure of the single plate total heat exchanger comprised by two symmetric half cells.
192 Li-Zhi Zhang
Figure 9.4. Photo of the two stainless steel half shells.
Vapor Permeation Measurements
A test rig has been set up to measure the vapor permeability through the fabricated
membranes. The whole test set-up is shown in Figure 9.2. Two air streams, one humid fresh
air and one dry exhsust, flow through a membrane exchanger to exchange moisture. For the
humid strip, ambient air is humidified and is driven to a heating/cooling coil in a hot/cool
water bath. After the temperature and humidity reach test conditions, the air is then drawn
through the exchanger for moisture exchange. For the dry stream, it is driven directly from
ambient to the exchanger. The two inlet temperatures are set to the same values. The
composite membrane is sandwiched by two stainless steel half shells. Two parallel air
passages on both sides of membrane are formed, which is like a counter-flow one-plate plate-
and-shell heat exchanger. A schematic of the single plate total heat exchanger is shown in
Figure 9.3. The real photo of the two shells is shown in Figure 9.4. The flow channel height is
2mm, and both the width and length are 10cm. The effective membrane area is 100cm2.
To have a balanced flow, equal air flow rates are kept for the two air streams. The
uncertainties are: temperature ±0.1ºC; humidity ±1%; volumetric flow rate ±1%. The final
uncertainty is ±4.5%.
The moisture transfer in the exchanger is governed by three resistances: the boundary
layer resistance on humid air side, the membrane resistance, and the boundary layer resistance
on dry air side. For convenience, a total mass transfer coefficient k is used to summarize the
moisture transfer through the membrane. It summarizes the three resistance simultaneously.
For different membranes, the boundary layer resistance is the same if the working conditions
are the same, i.e., with the same air bulk velocities in the channels. The resistance from the
porous support layer is the same. The thickness of the active layer is the same. As a result, the
total moisture transfer coefficient is an index of vapor permeability of the fabricated CTA
layer. After the inlet and outlet humidity are measured, the total mass transfer coefficients are
calculated by:
u a Ac (ω1i − ω1 o )
K= (9.1)
At Δω lm
where Ac is the cross section area of air duct (m2), At is the transfer area of membrane in the
cell (m2), Δωlm is the logarithmic mean humidity difference between the solution surface and
air stream, and it is calculated by
(ω1i − ω 2o ) - (ω1o − ω 2i )
Δω lm =
(ω − ω 2o )
(9.2)
ln 1i
(ω1o − ω 2i )
where ω represent temperature humidity ratio (kg/kg); subscripts 1 and 2 represent air stream
1 and air stream 2 respectively; subscripts i and o represent inlet and outlet respectively.
The vapor permeation rate Pe (kg/(m2·s)) is calculated by the following equation
u a ρ a Ac (ω1i − ω1 o )
Pe = (9.3)
At
where ρa is air density (kg/m3), At is the membrane transfer area (m2).
6.0-5
6.0x10
CM0
5.5x10-5 5.5 CM1
CM2
5.0-5
5.0x10 CM3
CM4
CM5
4.5-5
4.5x10
Pe (10-5 kgm-2s-1)
Pe (kgm-2s-1)
4.0-5
4.0x10
3.5-5
3.5x10
3.0-5
3.0x10
2.5-5
2.5x10
2.0-5
2.0x10
0.6 0.8 1.0 1.2 1.4
ua (m/s)
Figure 9.5. Effects of LiCl content on vapor permeation rates.

194 Li-Zhi Zhang
Characterization of the Membrane
The fabricated membranes are characterized by Contact angle measurements (OCA20,

Dataphysics, Germany), vapor sorption experiments (Hydrosorb-1000, Quantachrome, USA),
Scanning Electron Microscope (SEM) (LEO 1530VP, Germany) to investigate their physical
properties. The characterization may be helpful to disclose why the LiCl additive is useful to
facilitate moisture transfer.
Vapor Permeation Tests
The vapor permeations between the dry and the humid air streams are measured in the
test rig shown in Figure 9.2. The moisture flux and the total mass transfer rate are shown in
Figures 9.5 and 9.6 respectively. As seen, the higher the air flow rates are, the higher the
moisture permeation rates are. The reason behind is that the higher the air stream velocities,
the less the convective moisture transfer resistance in the two boundaries layers adjacent to
membrane. Furthermore, both the moisture permeation rate and the total mass transfer
coefficient increase with higher LiCl content in membranes. This proves that LiCl salts
indeed can improve moisture permeation substantially. CM5 that has 2.3% LiCl has the
highest moisture permeation rate. Actually, under air velocity of 1.0m/s, its vapor permeation
rate is 70% higher than CM0 that has no LiCl additives. Further increase of LiCl content in
casting solution is not suggested since the membrane will be difficult to get dry.
1.4x10-2
CM0
1.2-2
1.3x10
CM1
1.2x10-2 CM2
CM3
1.0-2
1.1x10
CM4
1.0x10-2 CM5
K k (10-2 m/s)
0.8-3
9.0x10
8.0x10-3
0.6-3
7.0x10
6.0x10-3
0.4-3
5.0x10
4.0x10-3
0.2-3
3.0x10
0.6 0.8 1.0 1.2 1.4 1.6 1.8

ua (m/s)
Figure 9.6. Effects of LiCl content on total mass transfer coefficient.

Table 9.2. Contact angles between the membrane surface and water droplet
Mambrane symbol CM0 CM1 CM2 CM3 CM4 CM5

LiCl content (g) 0 0.5 1 1.5 2 2.5
Contact angles (°) 85.9 82.6 79.3 65.3 57.9 54.3
SEM Studies
SEM images of the surface and the cross-section of the developed composite membranes
are shown in Figure 9.7, in which, the surface of the support PP layer is shown in (a), and the
surface and the cross-section of the composite membrane CM3 are shown in (b) and (c)
respectively. It can be seen that the PP layer is porous and the CTA layer is dense. The porous
layer provides the mechanical support and the CTA active layer provides the selective
permeation of moisture. CO2 permeation tests are also conducted with one air stream of
higher CO2 ratio (20%) and other air stream of low CO2 (0%). Air flow rates are kept at
1.0m/s. No CO2 permeation through the composite membranes is observed. The precision of
CO2 sensor is 0.1 vol%. The calculated H2O/CO2 permeation is beyond 1000. The membranes
made are assumed defects free.
(a)
Figure 9.7. (Continued on next page.)

196 Li-Zhi Zhang
(b)
PP
CTA
(c)
Figure 9.7. The SEM graphs of (a) surface of PP, (b) surface of CM3, and (c) cross-section of CM3.
Contact Angles
The contact angles between the surface of the composite membrane and water droplets
are measured. Totally 5 locations on CTA side are measured for each membrane. The
measured values are then averaged to represent the final contact angle between the membrane
and water droplets. Table 9.2 lists the contact angles for membranes with LiCl content varied
from 0 to 2.5g. The compositions of other components are shown in Table 9.1. As shown in
Table 9.2, the contact angle decreases as the content of LiCl increases, indicating increased
hydrophilicity. The membrane without LiCl (CM0) shows the largest contact angle,
implicating the least hydrophilicity. The reason behind is that when the content of LiCl is
increased, the high water sorption potential of LiCl can increase hydrophilicity.
Sorption Tests
The sorption isotherms between the membrane material and water vapor are measured in
a sorption analyzer Hydrosorb-1000. The adsorption (in solid line) and desorption (in dashed
line) isotherms of three membranes CM0, CM2 and CM4, at 25°C are shown in Figure 9.8.
The relative humidity changes from 0 to 1.0. As seen, membranes can adsorb more water
vapor with increased LiCl content. LiCl is a strong hygroscopic salt. Addition of LiCl in the
CTA membrane can increase its potential for moisture adsorption. At 60% relative humidity,
cross-linked CTA membrane CM0 can only adsorb 0.07g/g moisture. After adding 2% LiCl,
membrane CM4 can adsorb 0.4 g/g moisture. The moisture adsorption potential increases by
4.7 times. Considering the solution-diffusion mechanisms of vapor permeation in membranes,
that’s no wonder vapor permeability was greatly improved. When RH approaches 1.0, the
adsorbed moisture increases sharply due to multilayer, chemical adsorptions. This on the
other hand implicates that the membranes are nonporous. This is in agreement with the SEM
results.
It is also observed from Figure 9.8 that the higher the relative humidity is, the steeper the
sorption isotherms become. Considering the driving force for vapor permeation is the
moisture gradients across the membrane, therefore the higher the relative humidity, the
greater the vapor permeation rates are.
The results of the present work suggest that the addition of LiCl changes the CTA
membrane hydrophilicity substantially. The composite membranes fabricated with LiCl
additives exhibit improved moisture permeation properties, as well as good mechanical and
physical properties. They provide promising choices for total heat exchanger industry. The
increased hydrophilicity is mainly due to LiCl addition.
Membrane that has 2.3% LiCl has the highest moisture permeation rate. Its vapor
permeation rate is 70% higher than membrane that has no LiCl additives. The reason is that
with addition of LiCl, the hydrophilicity of the membrane is improved greatly. For instance,
at 60% relative humidity, addition of 2% LiCl in cross-linked CTA membrane can increase its
moisture adsorption potential by 4.7 times.
198 Li-Zhi Zhang
1.4
▲ CM0 adsorption curve
CM0
1.2 CM0 desorption curve
■ CM2
CM2 adsorption curve
1.0 ♦ CM4
CM2 desorption curve
CM4 adsorption curve
CM4 desorption curve
0.8
x (g/g)
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0

φ
Figure 9.8. Water vapor sorption isotherms of membranes at 298K. The solid lines represent adsorption,
the dashed lines represent desorption.
Protective layer
Figure 9.9. Concept of the composite supported liquid membrane (CSLM).

9.3. COMPOSITE SUPPORTED LIQUID MEMBRANE

Moisture diffusivity in solid polymer membranes is usually very low, in the order of 10-
12
~10-13m2/s. Total heat exchangers only have limited trans-membrane vapor partial pressure
difference, therefore performances are quite limited with common homogeneous solid
membranes.
In contrast to solid membranes, moisture diffusion in liquid membranes (~10-9m2/s
[17,18], diffusivity) is several orders higher than that in solid membranes. Due to this reason
and the inherent high selectivity, in recent years, there has been much effort in progressing
the researches of supported liquid membranes (SLM) in various fields: air dehumidification
[19], SO2/CO2 separation [20], H2S/CH4 separation [21], wastewater treatment [22], metal
ions concentration (uphill transport) [23], separation of isomeric amines between two organic
phases [24], to name but a few.
To improve the performances of total heat exchangers, a novel membrane, a composite
supported liquid membrane (CSLM), which employs LiCl liquid solution immobilized in a
porous support membrane to facilitate the transport of moisture, is prepared [25,26]. To
protect the SLM, two hydrophobic PVDF (Polyvinylidene Fluoride) layers are formed on
both surfaces of the SLM. The concept is shown in Figure 9.9. The sweep represents exhaust
air. In this chapter, a different skin layer fabricated in our laboratory is used. In stead of
PVDF layers, PES (Polyether sulphone) layers are used as the skin layers. As before, the CA
(Cellulose acetate) is stilled used as the middle support layer to absorb liquid LiCl solution.
PVDF are highly hydrophobic, so the adhesion between PVDF and the CA liquid layer is
difficult. However, PES are easier to be glued to the CA layer.
Figure 9.10. SEM graph of the surface of PES membrane, 3000 times magnified.
200 Li-Zhi Zhang
Figure 9.11. SEM graph of the surface of the Cellulose Acetate membrane, 5000 times magnified.
Preparation of the Supported Liquid Membranes
Three types of commercial membrane were obtained from a supplier. Very hydrophilic
Cellulose Acetate (CA) membranes with nominal pore diameter 0.45μm a thickness 80μm are
used as the support media to immobilize LiCl solution. Two hydrophobic PES membranes
(equal nominal pore diameter 0.15μm, thickness 45μm) are used as the protective layer.
Crystals of LiCl⋅H2O with laboratory class purity is used as the solute.
PES layer
CA layer
PES layer
Figure 9.12. SEM graph of the cross section of the composite membrane without LiCl solution
immobolized, 1000 times magnified.
PES
layer
CA
layer
PES
layer
Figure 9.13. SEM graph of cross section of the composite membrane with LiCl solution immobilized,
1000 times magnified.
Before the preparation of composite membrane, each membrane is experimented and

observed for their basic micro structures. Figures 9.10 and 9.11 show the SEM (Scanning
Electron Photomicrograph) graphs of the CA membrane and PES membrane, respectively.
Under room temperature, well-stirred LiCl solution with 40% mass fraction is first
prepared in a closed glass container. Vacuum degassing is applied for 2 hours for the three
membranes, after which, the CA membrane is dipped into the LiCl solution. After 24 hours,
the CA membrane is moved from the solution and placed onto a clean glass plate which is
cleaned by alcohol. Surplus LiCl solution on surfaces of CA membrane is blotted off with
paper tissue. To be sure that no ionic liquid is removed from the membrane pores, the
cleaning procedure is very gentle. At this stage, PVC glue is brushed on one surface of the
two PES membranes, and at the same time on both surfaces of the CA membrane. After a few
seconds, the two PES membranes are glued to the CA membrane and are pressed together
gentlely for a few seconds. The prepared composite membrane is placed in a constant-
temperature-constant-humidity chamber for another 24 hours, before experiment is
performed.
For comparison, a composite membrane with no LiCl solution immobilized in the CA
membrane is also made with the same procedure. The cross section SEM views of the two
composite membranes are shown in Figure 9.12 and Figure 9.13 respectively. To prevent the
microstructure being destroyed by knife crushing when preparing cross section samples, the
membranes are first frozen in liquid nitrogen before they are broken off to see the cross
sections. Before observations, they are gilded with gold.
As seen from Figure 9.12, there are some gaps between different layers. Some big
cavities in the support layers are also observed, which are presumed to be imperfections in
202 Li-Zhi Zhang
membrane fabrications. However, they have no adverse effects for this study because during
operation, they will be filled with liquid solution. In the preparation process, some thickness
of the CA membrane is dissolved by the glue, resulting in a lesser support layer thickness
than raw material. Figure 9.13 with LiCl solution shows that the CA layer and PES layer
connect to each other very closely and have a dense and continuum interface. There are more
big cavities in the support layer. The reason behind this may be that with LiCl solution
soaked, the wetted molecular chains in CA membrane structure become more flexible and
they will swell and expand to two sides. The boundaries between different layers are pressed
together and linked to each other closely. The final CA layer thickness is 52 μm.
Moisture Transport Measurement
Here another permeation cell is used. The cell is very sensitive. The membrane module is
a circular cell having an exchange area of 176.7cm2. It is composed of two parts: the lower
chamber and the cap, as shown in Figure 9.14. When testing, the flat sheet membrane is
placed on the lower chamber inside which distilled water is contained. The cap is then
covered on the membrane surface and form a sandwiched structure. The membrane and the
inner surface of the cap form a cone-shaped cavity. The air is supplied through the air slits in
the cap. It is introduced through two diametrically positioned inlets (symmetrically placed)
into a circular-shaped channel at the perimeter, from where the air is distributed over the
membrane surface through the circular air slit. The air flows inward radially, until it exits the
cap outlet in the center. The cap is designed that a constant axial air velocity is realized.
When flowing across the membrane, the air stream exchanges moisture with the distilled
water through the composite CSLM, and is humidified.
Inlet Outlet Inlet

Cap
CSLM
Membrane
Lower chamber
Distilled water
Figure 9.14. Schematic of the test cell for CSLM.

Flow Meter and

Pressure Meter Controller/Pump
Valve
By pass
Distilled
Water
Flow Meter and

Compressed air Bubbler Humidity Sensor Controller/Pump
Humidity Sensor
Cell
Figure 9.15. The set-up of the test apparatus.
The whole experimental set up is shown in Figure 9.15. The cell is supplied with clean
and humidified air from an air supply unit. The supply air flows from a compressed air bottle
and is divided into two streams. One of them is humidified through a bubbler immersed in a
bottle of distilled water, and then re-mixed with the other dry air stream. The humidity of the
mixed air stream is controlled by adjusting the proportions of air mixing. The airflow rates are
controlled by two air pumps/controllers at the inlet and outlet of the cell. The humidities and
temperatures to and from the cell are measured by the built-in RH and temperature sensors,
which are installed in the pumps/controllers. A detailed description of the test procedure is
given in [27].
In the test, the vapor evaporation is slow, and the cell is well conductive. Therefore only
moisture transfer is considered, by neglecting thermal influences.
Analysis of Transfer Resistance
The schematic of the moisture transport in the cell is represented in Figure 9.16. The
variations of air humidity are depicted in Figure 9.17. The representations are: 1-2, from
lower chamber solution surface to membrane lower surface; 2-3, from first layer’s (PES)
lower surface to its upper surface; 3-4, from the second layer’s (liquid layer) lower surface to
its upper surface; 4-5, from the third layer’s (PES) lower surface to its upper surface; 5-6,
from membrane upper surface to air stream.
Moisture conservation in air stream is represented by a one-dimensional transient
equation:
∂ω 6 ∂ω 6 1 ∂ ⎛ ∂ω ⎞ J
+ ua = ⎜ Dva r 6 ⎟ + (9.4)
∂t ∂r r ∂r ⎝ ∂r ⎠ H d ρ a
204 Li-Zhi Zhang
ua Air Duct Hd
δ3
Composite SLM δ2
δ1
O Air gap L
r
Solution in cell
Figure 9.16. Schematic of the mass transfer model in the cell.
2 3
L1 4
5
L2 L3 6
O
Z
Figure 9.17. Air humidity variations through the composite membrane.
where ω6 is humidity in air stream (kg/kg), t is time (s), r is radius (m), Dva is vapor
diffusivity in air (m2/s), ua is air velocity (m/s) in radial direction, J is the local moisture
emission rate from the membrane to air (kgm-2s-1), Hd is height of air stream at inlet (m), ρa is
density of dry air (kg/m3). The upper cavity of the cell is specially designed that the radial air
velocity ua keeps constant in the flow. Air duct height Hd changes with flow.
On the membrane upper surface, the moisture emission rate
J = kρ a (ω 5 − ω 6 ) (9.5)
where k is the local convective mass transfer coefficient (m/s) between air stream and
membrane. Convective mass transport in the channel can be represented by [27]
−0.834
⎛r −r⎞
Sh = 0.3359 Re Sc⎜⎜ 0 ⎟⎟ (9.6)
⎝ 2H d ⎠
where Sh, Re, and Sc are Sherwood number, Reynolds number and Schmidt number,
respectively. They are defined as
2kH d
Sh = (9.7)
Dva
2u a H d
Re = (9.8)
ν
ν
Sc = (9.9)
Dva
where ν is the kinematic viscosity of air (m2/s), r is radial coordinate (m) and r0 is the radius
of the cell (m).
Initial conditions:
t=0, ω=ωL (9.10)
where ωL is the humidity ratio determined by the saturated NaCl solution and temperature
(kg/kg);
Boundary conditions:
r=r0, ω=ωi (9.11)
r=0, stream outlet (9.12)
Resistance in this component
1
rD = (9.13)
ρa k
This resistance is the inverse of convective mass transfer coefficient in air flow side,
divided by the density of dry air. It is also called the convective resistance.
Moisture Transfer through the Composite Membrane
Moisture diffusion through the composite membrane is expressed by
ω 2 - ω5
J = De ρ a (9.14)
δ1 + δ 2 + δ 3
206 Li-Zhi Zhang
where De is the effective moisture diffusion coefficient in the composite membrane (m2/s), it
is calculated by
δ1 + δ 2 + δ 3
De = (9.15)
δ1 δ δ
+ 2 + 3
De1 De2 De3
where De1, De2, De3 are the effective diffusivity in the first layer, second layer and the third
layer of the composite membrane.
Gas Transport in Porous Media
Transport of gas through porous media has been extensively studied and theoretical
models have been developed based on the kinetic theory of gases. Various types of
mechanisms have been proposed for transport of gases or vapors through microporous
membranes: Knudesn model, viscous model, ordinary molecule diffusion model, often
summarized by the dusty gas model. The pore size is important for elucidating the physical
nature of the mass transport through the membrane. Most of the literature reports have used
the average pore size to calculate the mass flux. Nevertheless, because of pore size
distribution of the membranes, more than one mechanism of mass transport can
simultaneously occur. In this study, pore size distribution will be considered to classify
various mechanisms.
Pore Size Distribution
The pore size distribution can be expressed by the probability density function (that is,
log-normal distribution) described by the following equation
1.00
▽
□ △
◇
0.75 ▽
△
□ ◇
▽
f (dp)
0.50
△
□ ◇ △ M1
0.25 ▽ ▽ M2
△ ◇ M3
□
◇ □ M4
▽
0.00
10 100 1000 10000
d p (nm)
Figure 9.18. Pore size distribution of four porous membranes.

df (d p ) 1 ⎡ (ln d p − ln μ p )2 ⎤
= exp ⎢− ⎥
2(ln σ p )
(9.16)
d (d p ) d p ln σ p (2π)1 / 2 ⎢⎣
2
⎥⎦
where dp is the pore size, μp is the mean pore size, and σp is the geometric standard deviation,
f(dp) is fraction of number of pores with diameters not greater than dp. Figure 9.18 shows the
pore size distributions of four PES porous membranes.
The surface porosity εs, defined as the ratio between the area of the pores to the total
membrane surface area, can be calculated from following Equation
Nπ n
εs = ∑
4 j =1
f j d j2 (9.17)
where N is the number of pores per unit area, known as pore density, and fj is the fraction of
the number of pores with size dj, j=1,…, n, is the jth class of pore sizes.
The surface porosity is different from the void volume, which is determined by
ε = ε s τ (9.18)
where τ is the pore tortuosity.

Thus, the number of pores per unit area can be calculated from the following equation
ε/τ
N= (9.19)
π n
∑ f j d j2
4 j =1
if the effective membrane porosity, which takes into account the tortuosity of the membrane
pores ε/τ, is known.
Mass Transfer of Vapor through a Single Membrane Pore
The established theory considers three mechanisms for mass transfer in a pore as depicted
in Figure 9.19: Poiseuille flow, ordinary molecular diffusion and Knudsen diffusion, or a
combination of them.
The governing quantity that provides a guideline in determining which mechanism is
operative in a given pore under given operating conditions is the ratio of the pore size to the
mean free path λ, which is calculated for a species i using the following expression [17],
k BT
λi = (9.20)
2 πσ i2 Pm
208 Li-Zhi Zhang
where σi is the molecular collision diameter (m), 2.641Å and 3.711 Å for water vapor and air,
respectively; kB is the Boltzmann constant, 1.38×10-23 J/K, Pm is the mean total pressure
within the membrane pores (Pa), and T is the absolute temperature (K).
Poiseuille flow
Knudsen diffusion
Molecular diffusion
Figure 9.19. Diffusion mechanisms for gases in pores.
For gaseous mixtures of two components, the mean free path and the collision diameters
are different from the corresponding quantities for the pure component. The following
relationship can be applied for vapor-air mixtures
σa +σv
σ va = (9.21)
2
Under room temperature and atmospheric pressure, calculated λ for air is 0.07μm; while
under vacuum conditions, mean free path for air may be several microns to several meters.
Obviously, operating conditions have a great influence on diffusion mechanisms.
Knudsen number
λ
Kn = (9.22)
dp
Poisseuille Flow
When Kn<0.01, i.e., the pore size is large in relation with the mean free path of gas
molecules, the molecule-molecule collisions between gas molecules themselves will dominate
and viscous Poisseuille flow will occur. Under this mechanism, the gas flow through a single
pore FP (kg/s) is
πd P4 M v p m 1
FP = Δp v (9.23)
128η v RT τδ
where Mv is the molecule weight of vapor (kg/mol); R is gas constant, 8.314 J/(mol K); ηv is
the vapor viscosity (Pas); τ is the tortuosity of membrane. pm is the mean partial pressure of
water vapor (Pa), Δpv is the transmembrane vapor partial pressure difference (Pa).
Equation (9.23) can be re-written in the following form as
1
FP = GP Δp v (9.24)
τδ
where the flow coefficient Gp [kgm/(sPa)] is
πd P4 M v p m
GP = (9.25)
128η v RT
Knudsen Diffusion
If the mean free path of the gas molecules is large in relation with the pore size, the
molecule-pore wall collisions are dominant over the molecule-molecule collisions and the gas
transport takes place via Knudsen flow. In this case, one obtains the following relationship for
the gas mass flow in a single pore,
πd p3 M v 8RT 1
FK = Δp v (9.26)
12 RT πM v τδ
For a relatively small pore size, Kn≥10, Knudsen flow is assumed predominant.
Vapor density (kg/m3)
pv M v
ρv = (9.27)
RT
Equation (9.26) can be written as

210 Li-Zhi Zhang
1
FK = GK Δp v (9.28)
τδ
πd p3 M v 8 RT
GK = (9.29)
12 RT πM v
Molecule Diffusion
Unless steps are taken to remove dissolved air from the feed and permeate prior to
processing, the dissolved air acts as a stagnant layer. If the dominant resistance is the
molecular diffusion resistance caused by the virtually stagnant air trapped within the
membrane pores, one obtains the following relationship for the gas flow in a single pore,
πd p2 M v Dva
FO = Δp v (9.30)
4 RT τδ
where the ordinary diffusion coefficient of water vapor molecule in air is expressed by
[17,29]
C a T 1.75 1 1
D0 = +
( )
(9.31)
1/ 3 2
Pm v1v / 3 + v a
Mv Ma
where Ca=3.203×10-4. Pm is the mean total pressure in pores. The terms vv and va are
molecular diffusion volumes and are calculated by summing the atomic contributions: va
=20.1, and vv =12.7 [17]. Mv and Ma are molecule weight of vapor and air in kg/mol. M is
0.018 kg/mol for water vapor and 0.029 kg/mol for air respectively. Under room pressure, air
is trapped in membrane pores, therefore this mechanism exists.
Equation (9.30) can be re-written as
1
FO = GO Δp v (9.32)
τδ
πd p2 M v Dva
GO = (9.33)
4 RT
Combined Flow
Between the two limits of Knudsen diffusion and Poisseuille flow, i.e., 0.01≤Kn<10, the
above mentioned three mechanims may coexist. The combined flow is
FPKO =
1
τδ
[ ]
GP + (GK-1 + GO-1 ) Δp v
−1
(9.34)
When Kn<0.01, the Poisseuille flow is dominant, the Knudsen mechanism may be
neglected, then the combined flow is
FPO =
1
[GP + GO ]Δp v (9.35)
τδ
When Kn≥10, the Knudsen flow is dominant, the Poisseuille mechanism may be
neglected, then the combined flow is
FKO =
τδ
(
1 −1
GK + GO−1 ) Δp v
−1
(9.36)
Total Mass Flux Across a Membrane
In the case of a membrane with a pore size distribution, all the above mechanisms may
exist, but to different extents, depending on the operating conditions and membrane
morphological characteristics. Finally, considering the various diameters of pores in the
membrane, the moisture flux across the membrane Jm (kg/m2s) is
⎛ m ( d =0.1λ ) p ( d =100 λ ) n ( d = d max)

⎞
J m = N ⎜⎜ ∑ f j FKOj + ∑ f j FPKOj + ∑ f j FPOj ⎟⎟ (9.37)
⎝ j =1 j =m j= p ⎠
where N is the number of pores per unit area (1/m2), m is the last class of pores in the
Knudsen region, p is the last class of pores transition region.
Substituting Eqs.(9.34) to (9.36) into Eq.(9.37), one gets
N⎡ n ⎤
Jm = ⎢∑ ( f jG j )⎥ Δp v (9.38)
τδ ⎣ j =1 ⎦
where
Kn<0.01
⎧GP + GO 0.1≤Kn<10
⎪⎪
G = ⎨GP + (GK-1 + GO-1 )
−1
Kn≥10 (9.39)
⎪ −1
⎪⎩(GK + GO−1 )
−1
212 Li-Zhi Zhang
Humidity ratio in ambient air is in the range of 0.005kg/kg to 0.035 kg/kg, therefore the
above equation can be simplified to
p v = 1.608ωP (9.40)
Substituting the relations of partial pressure and humidity ratio into Eq.(9.38), one gets
Δω m
J m = ρ a Deff (9.41)
δ
where ρa is density of dry air (kg/m3) and the effective moisture diffusion coefficient is
defined as (m2/s)
n
1.608 PN
Deff =
ρa τ
∑fG
j =1
j j (9.42)
Effective Diffusivity in the First and the Third Layer
The two protective layers on both sides of The liquid layer are highly hydrophobic. The
established theory of gas diffusion in such membranes considers three mechanisms: Poiseuille
flow, ordinary molecular diffusion and Knudsen diffusion, or a combination of them.
As discussed previously, when Kn (ratio of the pore size to the mean free path) ≥10, the
Knudsen flow is dominant, the Poisseuille mechanism may be neglected. Actually, in most
cases for air conditioning industry with microporous membranes, Knudsen number is larger
than 10, and Poisseuille flow can be neglected, then the flow is considered to be combined
Knudsen and ordinary diffusion.
In a simple form, Knudsen diffusion coefficient
dP 8 RT
DK = (9.43)
3 πM v
where R is gas constant, 8.314 J/(mol K).

The effective diffusivity of combined Knudsen and ordinary flow is
-1
DKO (
= DK-1 + DO-1 ) −1
(9.44)
Effective diffusivity in this layer
εi
Dei = DKO, i , i=1, 3 (9.45)
τi
Resistance in these two layers
δ 1,3
r1, 3 = (9.46)
ρ a De1,3
The resistance through these two layers is called the diffusion resistance, which is
calculated by diffusion distance divided by air density and moisture diffusivity in membrane.
Effective Diffusivity in the Second Layer
This layer is the supported liquid layer. Water transfer in liquid layer is described by:
ε2 ΔC w
J= Dwlq (9.47)
τ2 δ2
where Dwlq is water diffusivity in liquid membrane (m2/s), ΔCw is the difference of water
concentration in liquid membrane solution (kg/m3) between the two sides of liquid
membrane.
Water vapor partial pressure, temperature, and LiCl solution concentration are governed
by a thermodynamic equation [25]
B ( m) C ( m)
log p v = A(m) + + 2 (9.48)
T T
A(m) = A0 + A1 m + A2 m 2 + A3 m 3 (9.49)
B (m) = B0 + B1 m + B2 m 2 + B3 m 3 (9.50)
C (m) = C 0 + C1 m + C 2 m 2 + C 3 m 3 (9.51)
where in this equation pv is in kPa, T is in K, and m is molality of the electrolyte (mol LiCl
/kg water).
x
m= (9.52)
0.0425(1 − x)
where x is mass fraction of solute (kg LiCl/kg solution).The constants in Eqs. (9.48) to (9.51)
are given by
A0= 7.3233550, A1=−0.0623661,

214 Li-Zhi Zhang
A2= 0.0061613, A3= −0.0001042,

B0=−1718.1570, B1=8.2255,
B2=−2.2131, B3=0.0246,
C0=−97575.680, C1=3839.979,
C2=−421.429, C3=16.731,
Water concentration in solution is
C w = (1 − x )ρ sol (9.53)
where ρsol is solution density (kg/m3), and it is calculated by the following equation
i
3
⎛ x ⎞
ρ sol = ρ w ∑ ρi ⎜ ⎟ (9.54)
i =0 ⎝1− x ⎠
where ρw is pure water density at temperature T, and ρi are given below:
ρ0 =1.0, ρ1 =0.540966
ρ2 =−0.303792, ρ3 =0.100791
The thermodynamic equilibrium chart of LiCl solution dictated by Eqs.(9.48) to (9.52)

could be represented by a series of linear equations as
Table 9.3. Values of kp and Cw0 for LiCl solution
T kp Cw0
(°C) (kgm-3Pa-1) (kgm-3)
14 0.1428 727.8
24 0.0867 706.4
35 0.0463 700.2
45 0.0293 693.2
C w = k p p v + C w0 (9.55)
where kp is called the Henry coefficient (kgm-3Pa-1), and Cw0 is a constant (kg/m3). They are
functions of temperature as given in Table 9.3.
Substituting Eq. (9.47) and psychrometric relation (9.40) into (9.47) gives the effective
moisture diffusivity
1.608Pk p ε 2
De2 = Dwlq (9.56)
ρ aτ 2
Resistance in this layer
δ2
r2 = (9.57)
ρ a De 2
As seen, the resistance in this layer is similar to the resistance in the two protective
layers. It is calculated by diffusion distance divided by dry air density and the effective
moisture diffusivity in this layer.
Moisture Diffusion in the Air Gap
Moisture transfer below the membrane can be expressed by
ω1 − ω 2
J = ρ a Dva (9.58)
L
where L is the height of air gap (m).

Resistance in this component
δL
rL = (9.59)
ρ a Dva
It is also calculated by diffusion distance divided by moisture diffucsivity and dry air
density.
Moisture Permeability
Mean moisture permeability across the whole membrane surface, (kgm-1s-1)/(kg/kg), is

calculated by
u a ρ a Ac (ωO − ωi )
Pe = (δ1 + δ 2 + δ 3 ) (9.60)
At Δω lm
where Ac is the cross section area of air duct (m2), At is the transfer area of membrane in the
cell (m2), Δωlm is the logarithmic mean humidity difference between the solution surface and
air stream, and it is calculated by
ωo − ωi
Δω lm = (9.61)
⎛ ω − ωi ⎞
ln⎜⎜ L ⎟⎟
⎝ ωL − ωo ⎠
216 Li-Zhi Zhang
where subscripts o, i represent outlet and inlet of air stream, respectively.

The permeability Pe here represents moisture transfer rate (kg/s) for unit area of
membrane under unit transmembrane humidity difference (kg vapor/kg dry air), times total
thickness. It reflects the performance of membrane.
Dimensionless radius
r
r* = 1− (9.62)
r0
In the simulations, it is assumed that heat effects are negligible due to slow water
evaporation rates from the solution. Further, the physical properties like moisture diffusivity
are uniform and constant in the membrane.
Resistance Analysis
Both experimental and numerical values of outlet RH are obtained. Table 9.4 lists the
values of operating conditions and system configurations. The tested mean permeation rate
across the whole membrane surface is 1.1×10-4 kgm-2s-1, which is 3 times higher than the
value obtained with a highly hydrophilic solid polymer membrane of comparable thickness
(2.5×10-5 kgm-2s-1). The corresponding permeability with this supported liquid membrane is
5.2×10-6 (kgm-1s-1)/(kg/kg).
rL
rD 9%
23%
r1
28%
r3
28%
r2
12%
Figure 9.20. Percentages of various resistances to total moisture transfer resistance.
Figure 9.20 shows the percentages of various resistances to total resistance. As seen, the
convective moisture transfer resistance accounts for 23% of the total resistance. The two
protective layers account for 28% of the total resistance each. The air gap diffusion resistance
amounts for less than 10% of the total resistance. The supported liquid layer, LiCl solution
layer, only accounts for 12% of the total resistance.
Table 9.4. Parameters used in the test and analysis
Symbol Unit Value Symbol Unit Value

T °C 26.0 ε1, ε3 0.66
δ1, δ3 μm 48 ε2 0.51
δ2 μm 52 τ1, τ3 2.0
dp1, dp3 μm 0.18 τ2 2.8
dp2 μm 0.23 V L/min 10.0∼30
r0 mm 75 L mm 0.1
Hd mm 1.0
Dwl m2/s (3×10-9)
Effects of Protective Layers
The protective layers have great influences on membrane performance. Figure 9.21
shows the effects of two protective layer thicknesses on mean permeation rate and the
permeability through the membrane. The permeation rate decreases with an increase in
protective layer thickness, due to the increased moisture resistance. As for the permeability, it
first decreases, and then increases with thickness increasing. The reason is that the
permeability is co-determined by two contradicting factors: membrane resistance and
membrane thickness. The final permeability is the result of balance between these two
factors. When the thickness is less than 20μm, increased resistance with thickness is more
influential, while when the protective thickness is greater than 20μm, increases of thickness is
more influential. When the two protective thicknesses are reduced from 100μm to 5μm, the
permeation rate can be improved by 20%.
Figure 9.22 shows the effects of protective layer porosity on mean permeation rate and
the permeability. Porosity plays a big role. Both permeation rate and permeability increases
with porosity, due to decreased resistance. The mean permeation rate increases 1.3 folds with
a porosity increase from 0.2 to 0.8.
Figure 9.23 shows the influence of mean pore diameters of protective layers on
performance. The smaller the pore sizes, the greater the resistance, and the better the
performance. When the pores are greater than 0.4μm, performance improvement with larger
pores becomes slower. On the other hand, with larger pores, the use with protective layers to
support and stabilize the liquid layer becomes limited. Therefore membrane with 0.4 μm
diameter pores is a good optimization.
218 Li-Zhi Zhang
1.5E-04 6.0E-06
Jm
Pe
5.8E-06
Pe (kgm-1s-1)/(kg/kg)
Jm (kgm-2s-1)
5.6E-06
1.0E-04
5.4E-06
5.2E-06
5.0E-05 5.0E-06
0 20 40 60 80 100 120
δ1, δ3 (μm)
Figure 9.21. Effects of protective thickness on mean moisture permeation rate and permeability.
1.5E-04 6.0E-06
Jm
Pe
5.0E-06
Jm (kgm-2s-1)
1.0E-04 4.0E-06
3.0E-06
5.0E-05 2.0E-06
0 0.2 0.4 0.6 0.8 1
ε1, ε3
Figure 9.22. Effects of protective layer porosity on mean permeation rate and permeability.
Effects of Liquid Layer
The liquid layer has a major impact on performance, since it is not only the barrier, but
also the active layer facilitates moisture transfer. Figures 9.242 and 9.25 show the effects of
liquid layer thickness and porosity on performances, respectively. As seen, permeation rate
and permeability decrease with thickness increasing, due to resistance increasing. The
performance increases with porosity increasing, due to resistance decreasing. The porosity
has greater impacts on performance than thickness does. An increase of porosity from 0.2 to
0.8 has a 90% permeation rate improvement.
1.5E-04 7.0E-06
Jm
Pe
6.0E-06
Jm (kgm-2s-1)
5.0E-06
1.0E-04
4.0E-06
3.0E-06
5.0E-05 2.0E-06
0 0.2 0.4 0.6 0.8 1
dp1, dp3 (μm )
Figure 9.23. Effects of nominal pore diameter of protective layer on mean permeation rate and
permeability.
220 Li-Zhi Zhang
1.5E-04 6.0E-06
Jm
Pe
Jm (kgm-2s-1)
1.0E-04 5.0E-06
5.0E-05 4.0E-06
0 20 40 60 80 100
δ2 (μm )
Figure 9.24. Effects of liquid layer thickness on mean permeation rate and permeability.
1.5E-04 6.0E-06
5.0E-06
Jm (kgm-2s-1)
1.0E-04 4.0E-06
Jm 3.0E-06
Pe
5.0E-05 2.0E-06
0 0.2 0.4 0.6 0.8 1
ε2
Figure 9.25. Effects of liquid layer porosity on mean permeation rate and permeability.
9.4. HEAT CONDUCTIVITY OF MEMBRANES
Composite Supported Liquid Membrane
As illustrated in Figure 9.17, heat flux through the composite membrane
λe (T2 − T5 )
q= (9.63)
δ1 + δ 2 + δ 3
where 2 and 5 denote two membrane surfaces. Heat conductivity
δ1 + δ 2 + δ 3
λe = (9.64)
δ1 δ 2 δ 3
+ +
λ1 λ 2 λ 3
The heat conductivity in the first layer can be analyzed by [30]
λ 1 = λ a ε1 + λ s (1 − ε1 ) (9.65)
where subscripts a and s denote moist air and solid material part, respectively. The effective
heat conductivity of the third layer is assumed as the same to the first layer, since they are the
same material.
Similarly, the heat conductivity in the second layer, where a liquid solution is stationed in
the porous media, can be analyzed by
λ 2 = λlq ε 2 + λs (1 − ε 2 ) (9.66)
where subscripts lq denote liquid solution.
Composite Hydrophobic-hydrophilic Membrane
As illustrated in Figure 9.26, heat flux through a composite membrane is
λe (T2 − T4 )
q= (9.67)
δ1 + δ 2
where 2 and 4 denote the two surfaces of the composite membrane. Heat conductivity
222 Li-Zhi Zhang
δ1 + δ 2
λe = (9.68)
δ1 δ 2
+
λ1 λ2
The heat conductivity in the first porous layer can be analyzed by
λ 1 = λ a ε1 + λ s (1 − ε1 ) (9.69)
The second layer is the dense layer. Its heat conductivity is λ2.
1
2
q
3
4
L1 L2 5
O Z
Figure 9.26. Heat transfer through a composite membrane.
9.5. MEMBRANE SELECTIVITY

An ideal membrane should let vapor permeates freely, but prevent other unwanted gases
to permeate. CO2 is a typical unwanted gas in air conditioning industry. It is an index for
indoor air pollution. Usually, indoor air CO2 concentrations should be less than 0.5%.
Therefore membrane selectivity is defined by permeations of H2O over CO2. The relation is
expressed by
Pev
α= (9.70)
PeCO 2
It indicates that selectivity is the ratio of vapor permeability to CO2 permeability. A test
rig shown in Figure 9.26 has been set up to measure the permeability and selectivity of CO2
and H2O. Membrane is sandwiched by two cells as demonstrated in Figure 9.28. Selectivity
can be calculated by the ratio of permeation rates. The membrane and the inner surface of
each cell form a cone-shaped cavity. Two air streams are supplied through the air slits in each
cell. Each stream is introduced through two diametrically positioned inlets (symmetrically
placed) into a circular-shaped channel at the perimeter, from where the air is distributed over
the whole membrane surface through the circular air slit. The air flow inward uniformly in
radial direction, until they exit the cell outlets in the center. One air stream has no inlet CO2,
which represents the fresh air. The other air stream has 2-4% CO2, which represents polluted
exhaust air. The inlet humidity is also different for two streams, representing outdoor humid
air and indoor dry air, respectively. The humidity is adjusted by two bubblers after the air
sources. When flowing across the membrane, the air streams exchange moisture and CO2
simultaneously. The convective mass transfer coefficients on both sides of membrane in the
cells are rather large and the resistance can be neglected. After measurement of the inlets and
outlets humidity and CO2 concentrations, the permeability and selectivity can be estimated.
The moisture permeability is calculated by
V (ω fi − ω fo )δ
Pev = (9.71)
AΔω m
(ω fi − ω eo ) − (ω fo − ω ei )
Δω m = (9.72)
⎛ ω − ω eo ⎞
ln⎜⎜ fi ⎟⎟
⎝ ω fo − ω ei ⎠
where V is the volumetric air flow rate of fresh air (m3/s), ω is air humidity ratio (kg vapor/kg
dry air), A is membrane area (m2), Δωm is the log mean humidity difference between the two
air streams. Subscripts “i", “o”, “f”, “e” denote inlet, outlet, fresh and exhaust air respectively.
The unit of permeability is [mol/(m⋅s)]/[mol/m3 humidity difference], meaning vapor flux in
mol/(m2s)) times the membrane thickness (m) divided by humidity difference in mol/m3.
Similarly, the permeability for CO2
V (cfi − cfo )δ
PeCO2 = (9.73)
AΔc m
(c − x eo ) − (c fo − xei )
Δc m =
fi
(9.74)
⎛ c fi − ceo ⎞
ln⎜ ⎟
⎜c −c ⎟
⎝ fo ei ⎠
where c is volumetric concentration of CO2 (m3 CO2/m3 air), Δcm is the log mean CO2
concentration difference between the two air streams. The unit of permeability for CO2 is the
same as for vapor. Different units of humidity and CO2 concentration are used because their
conventional units are different in air conditioning industry and in measurement.
Now the newly developed composite supported liquid membrane is measured. Table 9.5
lists the values of permeability and selectivity under five operating conditions. For each test,
several minutes are needed for the system to become fully steady state. After outlet RH
224 Li-Zhi Zhang
reaches the steady state, water vapor permeability and CO2 permeability are calculated. The
water vapor permeability is in the order of 1.39e-7 to 3.44e-7 [mol/(m⋅s)]/[mol/m3], and the
CO2 permeability is in the order of 3.19e-11 to 1.21e-10 [mol/(m⋅s)]/[mol/m3].The resulted
selectivity is ranging from 2845 to 4355.
The permeability of water vapor decreases with increasing air flow rates. The reason may
be that with increasing air flow rates, the operating fresh air inlet humidity increases, and the
log mean humidity difference increases. But the permeation increases slowly,
correspondingly, the permeability decreases a little bit. The permeability of CO2 decreases
more rapidly with flow rates, resulting an increased selectivity. When the CSLM is replaced
by the hydrophobic-hydrophilic composite membrane, the measured selectivity is higher than
5000. This indicates that both membranes are ideal for total heat exchanger.
Table 9.5. Permeability and selectivity for composite supported liquid membrane
V Pev PeCO2 α
ml/min [mol/(m⋅s)]/[mol/m3] [mol/(m⋅s)]/[mol/m3]
121 3.44e-7 1.21e-10 2845
174 3.55e-7 1.11e-10 3218

235 2.85e-7 8.09e-11 3521
335 1.87e-7 4.45e-11 4206
454 1.39e-7 3.19e-11 4355
Flow M eter and

Pressure M eter Controller/Pum p
Valve
Fresh air
Distilled
W ater
Hum idity/CO 2
Sensor
Com pressed air Bubbler
M em brane sandwiched
by two cells Exhaust air Hum idity/CO 2 Sensor
Com pressed CO 2
Com pressed air
Figure 9.27. Test rig for H2O/CO2 selectivity.

Inlet Outlet Inlet

Cells
Membrane
Inlet Outlet Inlet
Figure 9.28. Two cells to sandwich membranes.
9.6. CONCLUSION
Two novel membranes are developed for total heat exchangers. One employs a thin
active layer on a support porous layer. The other one employs a liquid layer supported in one
porous media. The two are highly permeable for water vapor. They provide a solution to total
heat exchangers, which demand cheap, permselective, high vapor permeable membranes.
Moisture diffusivity and heat conductivity are evaluated. For sensible heat transfer,
membrane resistance is very small and it can be neglected. However for moisture transfer,
membrane resistance is the major part of total resistance. Therefore it cannot be neglected.
Novel membranes should be developed to reduce this part of resistance. It has a determining
effect of latent effectiveness of a total heat exchanger.
REFERENCES
[1] Harriman, L.G.; Judge, J. Dehumidification equipment advances. ASHRAE Journal,
2002, 44, 22-29.
[2] Zhang, L.Z.; Niu,J.L. Energy requirements for conditioning fresh air and the long-term
2001, 26, 119-135.
[3] Reineke, C.R.; Moll, D.J.; Reddy, D.; Wessling, R.A. Functional Polymers. New York:
Plenum Press; 1989.
[4] Ye, X.H.; Levan, M.D. Water transport properties of Nafion membranes Part I. Single-
161.
[5] Cha, J.S.; Li, R.; Sirkar, K.K. Removal of water vapor and VOCs from nitrogen in a
hydrophilic hollow fiber gel membrane permeator. Journal of Membrane Science,
1996, 119, 139-153.
226 Li-Zhi Zhang
[6] Pan, C.Y.; Jensen, C.D.; Bielech, C.; Habgood, H.W. Permeation of water vapor
through cellulose triacetate membranes in hollow fiber form. Journal of Applied
Polymer Science, 1978, 22, 2307-2323.
[7] Hu, H.; Jia, J. Xu, J. Studies on the sulfonation of poly(phenylene oxide) (PPO) and
permeation behavior of gases and water vapor through sulfonated PPO membranes, II.
Permeation behavior of gases and water vapor through sulfonated PPO membranes.
Journal of Applied Polymer Science, 1995, 51, 1405-1409.
[8] Dilandro, L.; Pegoraro, M. Bordogna, L. Interaction of polyether-polyurethane with
water vapor and water-methane separation selectivity. Journal of Membrane Science,
1991, 64, 229-236.
[9] Wang, K.L.; McCray, S.H. Newbold, D.D. Cussler, E.L. Hollow fiber air drying.
[10] Aranda, P. Chen, W.J. Martin, C.R. Water transport across polystyrenesulfonate/
alumina composite membranes. Journal of Membrane Science, 1995, 99, 185-195.
[11] Scovazzo, P.; Hoehn, A.; Todd, P. Membrane porosity and hydrophilic membrane
based dehumidification performance. Journal of Membrane Science, 2000, 167, 217-
225.
[12] Morillon, V.; Debeaufort, F.; Blond, G.; Voilley, A. Temperature influence on moisture
transfer through synthetic films. Journal of Membrane Science, 2000, 168, 223-233.
[13] Kistler, K.R.; Cussler, E.L. Membrane modules for building ventilation. Chemical
Engineering Research & Design, 2002, 80, 53-64.
[15] Zhang, Li-Zhi; Wang, Yuan-Yuan; Wang, Cai-Ling; Xiang, Hui. Synthesis and
characterization of a PVA/LiCl blend membrane for air dehumidification. Journal of
Membrane Science, 2008, 308, 198-206.
[16] Zhang, L.Z. Fabrication of a Lithium Chloride solution based composite supported
liquid membrane and its moisture permeation analysis. Journal of Membrane Science,
2006, 276, 91-100.
[17] Cussler, E.L. Diffusion-Mass Transfer in Fluid systems. Cambridge: Cambridge
University Press; 2000.
[18] Isetti, C.; Nannei, E.; Magrini, A. On the application of a membrane air-liquid
contactor for air dehumidification. Energy and Buildings, 1997, 15, 185-193.
[19] Ito, A. Dehumidification of air by a hygroscopic liquid membrane supported on surface
of a hydrophobic microporous membrane. Journal of Membrane Science, 2000, 175,
35-42.
[20] Sengupta, A.; Raghuraman, B.; Sirkar, K.K. Liquid membranes for flue gas
desulfurization. Journal of Membrane Science, 1990, 51, 105-126.
[21] Quinn, R.; Appleby, J.B.; Pez, G.P. Hydrogen sulfide separation from gas streams
using salt hydrate chemical absorbents and immobilized liquid membranes. Separation
Science and Technology, 2002, 37, 627-638.
[22] Lin, S.H.; Pan, C.L.; Leu, H.G. Equilibrium and mass transfer characteristics of 2-
chlorophenol removal from aqueous solution by liquid membrane. Chemical
Engineering Journal, 2002, 87, 163-169.
[23] Dreher, T.M.; Stevens, G.W. Instability mechanisms of supported liquid membranes.
Separation Science and Technology, 1998, 33, 835-853.
[24] Fortunato, R.; Afonso, C.A.M.; Reis, M.A.M.; Crespo, J.G. Supported liquid
membranes using ionic liquids: study of stability and transport mechanisms. Journal of
Membrane Science, 2004, 242, 197-209.
[25] Zhang, L.Z. Fabrication of a Lithium Chloride solution based composite supported
liquid membrane and its moisture permeation analysis. Journal of Membrane Science,
2006, 276, 91-100.
[26] Zhang, L.Z. Effects of membrane parameters on performance of vapor permeation
through a composite supported liquid membrane. Separation Science and Technology,
2006, 41, 3517-3538.
[27] Zhang, L.Z. Evaluation of moisture diffusivity in hydrophilic polymer membranes: a
new approach. Journal of Membrane Science, 2006, 269, 75-83.
[28] Khayet, M.; Matsuura, T. Pervaporation and vacuum membrane distillation processes:
modeling and experiments. AIChe Journal, 2004, 50, 1697-1712.
[29] Tomaszewska, M.; Gryta, M.; Morawski, A.W. Mass transfer of HCl and H2O across
hydrophobic membrane during membrane distillation. Journal of Membrane Science,
2000, 166, 149-157.
[30] Datta, A.K. Porous media approaches to studying simultaneous heat and mass transfer
in food processes. I: problem formulations. Journal of Food Engineering, 2007, 80, 80-
95.
Chapter 10
HEAT MASS TRANSFER IN CROSS-CORRUGATED

TRIANGULAR DUCTS
ABSTRACT
To increase sensible and latent effectiveness of a stationary total heat exchanger with
limited transfer area, heat and mass transfer intensification should be performed. There
are two directions for heat mass transfer intensification: material side intensification and
air side intensification. Cross corrugated triangular ducts are a structure to intensify heat
mass transfer in air side. They belong to the primary surface heat exchangers. In this
chapter, fluid flow and convective heat mass in the cross-corrugated triangular ducts are
numerically calculated. Due to the different flow regimes under different Reynolds
numbers, various momentum models from laminar, transitional, to fully turbulent are
considered.
NOMENCLATURE
Aci cross-sectional area at inlet or outlet of a cycle, m2
Acyc surface area of the channel, m2
cp specific heat, kJ/(kgK)
Dh hydraulic diameter of the channel, m
Dva vapor diffusivity in air, m2/s
f friction factor
h heat transfer coefficient, kW/(m2K)
k turbulent kinetic energy, m2/s2
Lcyc length of a cycle in flow direction, m
Nu Nusselt Number
P time average pressure, Pa
Pr Prandtl Numbers
q heat flux, W/m2; mass flux, kgm-2s-1
Re Reynolds number,
Sc Schmidt Numbers
230 Li-Zhi Zhang
T temperature, K
U time average velocity, m/s
Vcyc volume of channel, m3
x, y or z coordinates, m
ycyc pitch of the repeated segment of the duct, m
zcyc width of the repeated segment of the duct, m
Greek letters
ρ density, kg/m3
ν kinematic viscosity, m2/s
μ molecular viscosity, Ns/m2
τ shear stress, N/m2
ω specific dissipation, s-1
ε turbulent dissipation rate, m2/s3
Superscripts
* dimensionless
‘ fluctuation
Subscripts
0 inlet
i inlet
m mean
o outlet
t turbulent
w wall
10.1. INTRODUCTION
Parallel plates and plate-fin have been the main structure for total heat exchangers.
Parallel-plates are simple, however their heat mass transfer capability is limited. Additional
spacers are required to ensure the channel pitch not narrowed by neighboring membranes
collapsing, contacting, or oscillating under air flow. Plate-fin is strong, stable and compact.
However as discussed in previous chapters, due to the finite fin conductance both for sensible
heat and latent heat, the fin efficiency is quite limited.
To enhance the heat and mass transfer, a structure named the cross-corrugated triangular
membrane duct has been proposed [1-3]. It belongs to a type of primary surface heat
exchangers, which have been used for air-to-air sensible heat exchangers. The concept is
shown in Figure 10.1. Flat membrane sheets are corrugated to form a series of parallel
equilateral triangular ducts. Sheets of the corrugated plates are then stacked together to form a
90 degree orientation angle between the neighboring plates, which guarantees the same flow
Heat Mass Transfer in Cross-corrugated Triangular Ducts 231
pattern for both fluids. The shaded area is blocked. The membranes are very thin and soft,
which requires plastic frame cases to support them, as seen in Figure 10.2. Consequently,
with a pre-designed plastic frame, triangular shaped duct walls are formed to construct the
required geometry. The structure gives better heat mass transfer. This efficiency improvement
is attributed to the pattern of flow that undergoes abrupt turnaround, contraction, and
expansion.
Cross corrugated ducts with sinusoidal cross sections, which are mainly used for rotary
regenerators, have been investigated extensively by various investigators [4-10].This chapter
gives the recent research results of cross-corrugated triangular ducts. Triangular cross sections
are naturally formed by the corrugations of ultra-thin materials like paper, plastic films, tinsel,
and hydrophilic membranes, which are increasingly used in air conditioning industries, due to
their superiorities in weight-lightness, cheapness, and abilities in selective transfer.
CFD modeling is a cost-effective means to study the flow field and mass transport
phenomenon in fine yet complex geometries, where it’s usually hard to get information of
local flow and mass fraction distributions, either for technical or economical reasons. With
appropriate models, the flow and mass transfer can be simulated rather accurately. In fact,
CFD has become the most efficient tool for design of chemical processes and equipments.
Using CFD to study mass transport in heat mass exchangers is one of the fast developing
technologies. In this chapter, the convective heat mass transfer coefficients and the friction
factors, which are the major parameters affecting membrane total heat exchanger
performance, were investigated with this efficient tool.
Fresh
Exhaust
Figure 10.1. Cross-corrugated triangular ducts.

232 Li-Zhi Zhang
Spacer Membrane
Figure 10.2. The spacer/case to form the channel.
10.2. LAMINAR FLOW
Mathematical Model
A computational domain is selected as Figure 10.3. It is a representing unit cell in the

total heat exchanger.
The basic set of equations that require solving comprise of equations for:
• conservation of mass;
• conservation of momentum, in three coordinate directions and
• conservation of energy.
The general form of the mass continuity equation as shown below is valid for
compressible and incompressible flows:
∂ρ ∂
+ (ρu i ) = 0 (10.1)
∂t ∂xi
where ρ is the fluid density (kg/m3), t the time (s), u the flow velocity (m/s), subscript i
denotes coordinates directions, say, x, y or z.
Air out
Air in
z y
x Symmetry
Duct wall
Figure 10.3. The single cross-corrugated channel segment for computation.

The conservation of momentum in the ith direction in an inertial reference frame is

governed by
∂ ∂σ
(ρui ) + ∂ (ρui u j ) = ij + Bi (10.2)
∂t ∂x j ∂x j
where Bi is a body force in the ith direction. It includes contributions from gravitational
acceleration and external body forces. The stress tensor, σij, is given by
⎡ ∂u i ∂u j ⎤ 2 ∂u k
σ ij = − pδ ij + μ ⎢ + ⎥− μ δ ij (10.3)
⎣⎢ ∂x j ∂xi ⎦⎥ 3 ∂x k
where the final term is the effect of volume dilation (zero for an incompressible fluid). The
pressure and molecular viscosity are denoted by p (Pa) and μ (Ns/m2), respectively; δij is the
Kroneker operator, which equals to 1 when i=j, and 0 when i≠j.
In addition, a general scalar advection-diffusion equation for a dependent variable, φ, is
given by
∂ (ρφ) ∂ ⎡ ∂φ ⎤
+ ⎢ρu i φ − Γ ⎥=S (10.4)
∂t ∂xi ⎣ ∂xi ⎦
Γ is the diffusion coefficient and S a source or sink term representing creation or

destruction of Γ.
For fully compressible flow, the energy transport equation is solved for the total enthalpy,
H, according to
∂ (ρH ) ∂ ⎡ ∂T ⎤ ∂ρ
+ ⎢ρu i H − λ ⎥= (10.5)
∂t ∂xi ⎣ ∂xi ⎦ ∂t
where λ is the thermal conductivity [kW/(mK)] and T the temperature (K). H is expressed in
terms of the static enthalpy, h, according to:
1
H = h + u2 (10.6)
2
p
h = U ie + (10.7)
ρ
234 Li-Zhi Zhang
For weakly compressible and incompressible flow, the kinetic energy term (1/2ρu2) is
assumed to be negligible compared to the internal term, Uie. The pressure work term, p/ρ,
may also be safely ignored.
The above equations, also known as the Navier-Stokes equations, represent five transport
equations in the seven unknown field variables ux, uy, uz, p, ρ, h, and T. The thermal equation
of state provides a sixth equation relating density, ρ, to temperature T and pressure p. For air,
this equation is called the idea gas equation. The seventh equation required to close the entire
system is a thermodynamic relation between the state variables. For air, this equation defines
the function of static enthalpy in terms of temperature and pressure, i.e., h=h(T,P). Since the
fluid is assumed to be thermally perfect, the static enthalpy is a function of temperature only.
The hydraulic diameter of the channel is defined as
4Vcyc
Dh = (10.8)
Acyc
where Vcyc and Acyc are the volume and the surface area of the channel, respectively.
The Reynolds number, Re, is
ρu m Dh
Re = (10.9)
μ
where um is the area-weighted mean velocity through a cross-section, (m/s).

The cycle-average heat transfer coefficient is evaluated from the temperature difference
between the inlet and the outlet of a cycle
ρu m c p Aci (Ti − To )
hm = (10.10)
Acyc ΔT
where cp is the specific heat of fluid, kJ/(kgK); Aci is the cross-sectional area at inlet or outlet
of a cycle, (m2); Ti and To are fluid mass weighted temperature at inlet and outlet of a cycle,
respectively (K); ΔT is the logarithmic temperature difference between the wall and the fluid,
which is calculated by
(Ti − Tw ) − (To − Tw )
ΔT = (10.11)
T − Tw
ln i
To − Tw
where Tw is the wall temperature (K).

The cycle-average Nusselt number, Num, is defined as
hm Dh
Nu m = (10.12)
λ
The cycle-average friction factor is calculated by
pi − p o
Dh
Lcyc
fm = (10.13)
1 2
ρu m
2
where Lcyc is the length of a cycle, (m); pi and po are pressure at inlet and outlet of a cycle,
respectively, (Pa).
In a cycle, the heat transfer coefficient and friction factor have different local values. The
local heat transfer coefficient along the flow is defined by
1 ∂T
hL = − (10.14)
(Tb − Tw ) ∂n w
∂T
where Tb is the bulk temperature, (K); is the temperature gradient at the wall surface in
∂n w
normal position.
The local Nusselt number
hL Dh
Nu L = (10.15)
λ
The local Darcy friction factor
⎛ ∂u ⎞
8μ ⎜ i ⎟
⎝ ∂n ⎠ w
fL = (10.16)
ρu m2
⎛ ∂u i ⎞
where ⎜ ⎟ is the velocity gradient at the wall surface in normal position.
⎝ ∂n ⎠ w
x
x* = (10.17)
Lcyc
236 Li-Zhi Zhang
y
y* = (10.18)
y0
z
z* = (10.19)
z0
where y0, z0 are the pitch and width of the repeated segment of the duct (m), respectively.
Solution Method and Validation
The problem is solved with a commercial CFD code Fluent. The meshes on the outside
walls of a computational block are shown in Figure 10.4. The graph depicts the meshes only
for the first 3 and a half cycles, to get an amplified view of the mesh structure. Totally there
are 10 cycles in this block.
Boundary conditions are defined. A uniform temperature and non-slip velocity wall
conditions are assumed. At the inlet, velocity is set to uniform and parallel to the corrugation
of the upper wall.
The governing equations are solved by using standard finite difference methods that
employ control-volume based discretization techniques along with a pressure-correction
algorithm. The N-S equations are solved by SIMPLEC scheme, while the convective term in
the energy equation is solved by first-order upwind implicit approximation, and the diffusive
term is by second-order central difference scheme. The fluid is selected as air.
Figure 10.4. The grid distribution for the computation domain, showing 2 and a half cycles, totally
10cycles.
Table 10.1. Comparisons of (fD⋅Re) and NuD of fully developed laminar flow for some
ducts from present study and those from literature
Duct geometry τ (fD⋅Re) NuD

Present Refs Deviation Present Refs Deviation
study (%) study (%)
Circular 64.334 64.000 0.51 3.667 3.657 0.55
Square 56.470 56.908 0.78 3.011 2.976 0.84
Elliptic 0.5 67.754 67.292 0.70 3.753 3.742 0.56
Equilateral Triangular 53.334 53.284 0.11 2.476 2.46 0.65
Isosceles Triangular 0.5 52.718 52.612 0.18 2.331 2.34 0.90
Sine 2.0 58.314 58.212 0.16 2.658 Unavailable
Sine 1.5 55.726 56.088 0.63 2.614 2.6 0.54
Sine 1.0 51.688 52.092 0.77 2.463 2.45 0.53
Sine 0.75 49.304 48.936 0.75 2.317 2.33 0.56
Sine 0.50 44.682 44.828 0.30 2.141 2.12 0.99
Notes: τ, aspect ratio (duct height/duct width).
The grid independency test has been done. The calculations were primarily carried out
with three different grid densities, 219985, 114992, and 429970 mesh points. The channel
fully developed periodic mean pressure drop and temperature change for the two fine grids
are almost the same and 10% higher than that for the coarse grid. For the finest grids, 429970,
the solution time is very long, which is hard to use practically. Based on the above
experience, which establishes the grid independency, the final calculations are performed for
the 219985 grids and the results obtained in this paper refer to the grid geometry mentioned
above.
To validate the solution procedure, the methodology has been used to calculate the
friction factor and Nusselt numbers of fully-developed flows in ducts of various cross-
sections. For the fully developed laminar flow in such ducts, the local values of (fL⋅Re) and
NuL are constants. In such cases, these constant values are denoted as (fD⋅Re) and NuD,
respectively. Table 10.1 shows the values calculated and give in refs [11,12]. From this table,
it can be concluded that maximum errors are less than 0.8% for (fD⋅Re) and less than 0.99%
for NuD.
Flow Distribution and Friction Factor
Figure 10.5 shows the vector plot of the velocity in the x-y plane at z*=0.5 (the center in
width), for Re=1000. As seen from this figure, the flow has two distinct patterns: in the
corrugation troughs of the upper wall, the flow is parallel steady flow, while in the troughs of
the lower wall, fluid re-circulation or swirl flows are generated due to the reason that the fluid
separates from the rear-facing facet and reattaches to the front facing facet. The shapes of
flow re-circulation in the valleys become almost identical to each other, after 3-5 cycles,
indicating a cyclic manner. The maximum value of the velocity occurs near the peaks of the
lower wall, where the flow has the least cross section area. The velocity is the smallest where
238 Li-Zhi Zhang
the duct expansion occurs. The vortexes resulted from duct expansion and contraction are
helpful for momentum transfer.
The variations of the local wall friction factor along the perimeter of the duct cross
section that is perpendicular to x axis are shown in Figure 10.6 for the upper wall. A double
climax pattern is found in the figure. The local wall friction factor has the maximum value at
the center of the border edge of the upper wall (points B and D), where the velocity is the
highest. It has the least value at the ends of the border edge of duct walls (Points A, C, E),
where the velocity is also the least. Due to the symmetry of duct cross section, the local wall
friction factor also distributes symmetrically.
Figure 10.5. Velocity contours in the x-y plane at z*=0.5, Re=1000.
B D
A E
(a) The geometry of duct cross section and positions.
fL*
B D
1
E
A C
0 1 z*
(b) The local wall friction factor.
Figure 10.6. Local wall friction factor on the cross-section plane at the center of 5th cycle, Re=1000; (a)
the geometry of duct cross section and (b) the local wall friction factor distribution.
2
Re=1000
Re=533
fm 1.5 Re=100
0.5
0
0 2 4 6 8 10 12
Cycles
Figure 10.7. Mean friction factor of each cycle along the duct。
Variations of the local wall friction factor along duct length also have a cyclic manner:
the local wall friction factor becomes the least at the valley of the cycle, while reaches the
highest at the crest. The reason is that at the valley, flow has the least velocity; while at the
crest, flow has the highest velocity. After 3-5 cycles, the behavior of local wall friction factor
gets stable on a cycle-by-cycle basis.
For the cross-corrugated triangular ducts, the cyclic mean friction factor is one of the
most important parameters affecting heat exchanger design. In Figure 10.7 is shown the mean
friction factor of each cycle along the flow direction, for 3 different flow conditions. As
shown, the cyclic mean friction factor is very high at the entrance, but it decreases rapidly in
this region. After 3-5 cycles, it decreases gradually to a stable value, which is called the fully
developed cyclic mean friction factor, fD. Also indicated in this figure is that the higher the
Re, the greater the fD.
Figure 10.8 shows the variations of the fully developed cyclic mean friction factor with
increasing Re for ducts of three cross sectional shapes: the corrugated duct in this study, the
parallel flat plates duct, and a constant cross sectional area triangular duct. It should be noted
that for the straight ducts of constant cross section, fD denotes the fully developed local
friction factor. Generally speaking, the friction factor for a corrugated duct is greater than that
for a triangular duct. When the Reynolds numbers are less than 200, fD of the corrugated duct
is less than the parallel plates duct. At higher flow rates, i.e., bigger Reynolds numbers, fD of
the corrugated duct becomes greater than the parallel plates duct. This characteristics disclose
a fact that flow field in the corrugated duct has been enhanced, in comparison to a traditional
duct, especially when at higher Re numbers, due to the existence of vortexes. The growth of
steady swirl in the corrugation valleys with Reynolds numbers and the concomitant periodic
240 Li-Zhi Zhang
disruption and thinning of the boundary layer promote enhanced transport of momentum. In
contrary, in the non-swirl flow regime, the enlarged surface area of the corrugated plate is
solely responsible for the enhancement, which is significantly less.
A correlation has been proposed to establish the relations between the periodically mean
friction factor of the cross-corrugated triangular duct with Reynolds numbers as following:
f D = 11.03 Re −0.5121 (10.20)
for 10≤Re≤2000
Temperature Distribution and Nusselt Number
The isotherms shown in Figure 10.9 in an x-y midplane for Re=1000 clearly show a
cyclic manner. After 5-7 cycles, the thermal boundary layer has become fully developed. The
temperature profiles in a y-z cross section are shown in Figure 10.10. The temperature
gradients have the highest values near the center of the boundary edge of the upper wall,
while have the least values on the valley of the lower wall. Flow conditions also have an
influence on temperature distribution. At smaller Re numbers, temperature gradients in the
valleys of the lower wall are smaller, due to the reduced intensities of swirls.
101
Corrugated
Flat parallel
Triangular
0
fD
10
10-1
101 102 103
Re
Figure 10.8. Friction factor for fully developed flow for three ducts.
Figure 10.9. Temperature contour on the x-y plane at z*=0.5, Re=1000.
Figure 10.10. Temperature contour on the z-y plane at x*=5.5 (center of the 5th cycle), Re=1000.
Figure 10.11 plots the distribution of the normalized local wall heat transfer coefficients
(h*=h/hmax) in an x-y plane at z*=0.5, and Re=1000. The values shown in the figure is for the
lower wall. The local wall heat transfer coefficients demonstrate a cyclic pattern: the shapes
of the local wall heat transfer coefficients distribution resemble each other on a cycle-by-
cycle basis. The heat transfer coefficients have the highest values at the crests of each cycle,
while have the least values at the valleys of each cycle. In combining the analysis of flow
fields, it can be concluded that heat transfer coefficients have the highest values where the
velocity are the highest.
242 Li-Zhi Zhang
Figure 10.11. Distribution of normalized local wall heat transfer coefficient on the x-y plane at z*=0.5,
Re=1000.
Figure 10.12 shows the local wall heat transfer coefficients in an y-z plane, for the upper
wall. A double climax pattern is observed, similar to that observed in local friction factor
analysis. The local wall heat transfer coefficients are the greatest at the center of the triangular
edges (Points B, D, in Figure 10.6(a)), while they are the least neat the sharp corners of the
cross section. This is due to the dead spaces in the sharp corners. After about 5-7 cycles, the
magnitudes of the variations of the local heat transfer coefficients become stable, which
implicates a fully developed thermal boundary condition.
Figure 10.12. Distribution of normalized local wall heat transfer coefficient in the z-y plane at x*=5.5.
Figure 10.13 shows the mean Nusselt numbers for each cycle, Num, under different flow
rates. At the entrance, the cyclic mean Nusselt numbers are very high, due to very thin
boundary layers at the entrance. Along the flow direction, the cyclic mean Nusselt numbers
decrease rapidly in the first 5-7 cycles, and then arrive gradually at some stable values, which
is denoted as the fully developed value, NuD. For parallel flat plates, NuD=7.54; for triangular
straight duct, NuD=2.47 (Incropera and Dewitt, 1996). Contrary to the fully developed flow in
straight ducts of constant cross sections, the fully developed Nusselt numbers, NuD, are not
constants. Rather, they are variables influenced by Reynolds numbers.
18
16 Re=1000
Re=533
14
Re=100
12
10
Num
8
6
4
2
0
0 2 4 6 8 10 12
Cycles
Figure 10.13. Cyclic mean Nusselt numbers for the flow.
16
Corrugated
14 Flat parallel
Triangular
12
10
NuD
0
101 102 10 3
Re
Figure 10.14. Comparison of fully developed Nusselt numbers for three ducts.
244 Li-Zhi Zhang
In Figure 10.14 is shown the comparisons of fully developed Nusselt numbers of three
ducts: the parallel flat plates, equilateral triangular straight duct, and the proposed cross-
corrugated triangular duct. As is shown, the Nusselt numbers of the corrugated duct are
higher than a common straight triangular duct. In Re range between 100 and 2000, compared
to a triangular straight duct, the heat transfer for the cross-corrugated duct is enhanced by
60% to 480%. As expected, the lower Nusselt numbers of the triangular straight duct is due to
large dead spaces in the sharp corners.
At lower Re numbers, the Nusselt numbers of the cross corrugated duct are lower than
parallel flat plates. However, at higher Reynolds numbers, i.e., 500≤Re≤2000, the Nusselt
numbers of the cross-corrugated duct grow rapidly, and surpass the parallel flat plates with
large margins. The reason behind this fact is that at lower flow rates, the swirls in the valleys
of the cross-corrugated duct is not strong enough, and the cyclic heat transfer is mainly
determined by the parallel flows in the corrugations of the upper wall, which is sacrificed by
the dead spaces in the corners. However, when the flow rates are high, strong swirls are
generated in the valleys, which effectively reduces the thickness of the boundary layers and
offsets the influences of dead spaces. In such cases, the swirls in the valleys are the dominant
factor for heat transfer. The heat transfer is enhanced by a factor of 2 when Re=2000, and is
expected to rise sharply with further increases in Re.
Studies of the flow with different flow rates and several gases have helped to propose a
correlation to estimate the fully developed cyclic mean Nusselt numbers as following:
Nu D = 0.5528 Re 0.2157 Pr 0.333 (10.21)
for 10≤Re≤2000. Pr is the Prandtl number.
10.3. TURBULENT FLOW

When Reynolds numbers are high, laminar model is not appropriate. This section is to
study turbulent flow for Re=2000~20000. The computational domain and grid structure are
the same as previous laminar flow. However the model is different.
Turbulence Models
The equations describing the fluid flow and heat transfer are transport equations for the
continuity, momentum, and energy, which are developed from conservation laws of physics.
The fluid flow is described by conservation of mass (the continuity equation), momentum
(Navier-Stokes equations) and energy (the temperature equation for the fluid). The velocities
and temperatures are time-averaged and divided into a mean and a fluctuating value,
u j = U j + u 'j and T = T + T ' . Together with the boundary conditions, they form the
steady state governing equations for incompressible flow with negligible external and viscous
forces:
∂ (ρU j )
=0 (10.22)
∂x j
∂
(ρU iU j ) = − ∂P + ∂ (τ ij + τ ijt ) (10.23)
∂x j ∂xi ∂x j
⎛
∂U ∂U ⎞
τ ij = μ ⎜⎜ i + ⎟ ; τ ijt = − ρ u i' u 'j
j
(10.24)
⎝ ∂x j ∂xi ⎟⎠
∂
(ρc pU j T ) = ∂ q j + q 'j
( ) (10.25)
∂x j ∂x j
μc p ∂T
qj = ; q j = − ρc p u j T
t ' '
(10.26)
Pr ∂x j
where ρ, μ, and cp are density, viscosity, specific heat, respectively. It would be impossible to
solve these equations analytically because of non-linearity and the stochastic nature of
turbulence. The extra terms that appear due to averaging the velocity and temperature are the
Reynolds stress and the turbulent heat flux. Modeling these is known as the closure problem
of turbulence. Various turbulence models have been proposed and totally 4 models are
considered.
The definition of hydraulic diameter, Reynolds number, heat transfer coefficient, Nusselt
number, and friction factor are the same as laminar flow. Velocity, temperature are selected
as the time averaged mean values.
Standard k-ε Model
The k-ε model is the most popular of the two-equation models and has produced
qualitatively satisfactory results for a number of complex flows. According to this concept,
the turbulent shear stress in Eq.(10.23) is determined by
⎛ ∂U i ∂U j ⎞ 2
τ ijt = μ t ⎜⎜ + ⎟ − δ ij kρ (10.27)
⎝ ∂x j ∂xi ⎟⎠ 3
where δ ij is the Kronecker delta function, δ ij = 1 when i=j and zero when i ≠ j . The
turbulent viscosity μt is determined by Prandtl-Kolmogorov equation [13]
μ t = C μ ρk 2 / ε (10.28)
246 Li-Zhi Zhang
where the turbulence kinetic energy k and dissipation rate ε are calculated by
∂k 1 ∂ ⎛⎜ μ t ∂k ⎞⎟ μ t ∂U i ⎛ ∂U i ∂U j
⎜
⎞
⎟−ε
Uj = + + (10.29)
∂x j ρ ∂x j ⎜⎝ σ k ∂x j ⎟⎠ ρ ∂x j ⎜ ∂x
⎝ j ∂xi ⎟
⎠
∂ε 1 ∂ ⎛⎜ μ t ∂ε ⎞⎟ ε μ t ∂U i ⎛ ∂U i ∂U j
⎜
⎞
⎟ − C ε2 ε
2
Uj = + C ε1 + (10.30)
∂x j ρ ∂x j ⎜⎝ σ ε ∂x j ⎟⎠ k ρ ∂x j ⎜ ∂x
⎝ j ∂xi ⎟
⎠ k
The turbulent heat transfer term in Eq.(10.26) is determined by the following equation
μ t ∂T
q' = (10.31)
σ θ ∂x j
The constants in the above model take following values [13]:
Cμ=0.09; Cε1=1.44; Cε2=1.92;

σk=1.3; σε=1.3; σθ=1.3
Renormalized k-ε model
The RNG-based k-ε model follows the same framework as the above two equations
model but uses Renormalization Group methods [14]. The model is to provide improved
predictions of near-wall flows and flows with high streamline curvature. The momentum and
energy equations can be re-written in the following form:
∂ (ρU iU j ) ⎡ ⎛ ∂U i ∂U j ⎞ 2 ⎤
=−
∂P
+
∂
⎢ μ eff ⎜⎜ + ⎟ − μ eff ∂u k ⎥ (10.32)
∂x j ∂xi ∂x j ⎢⎣ ⎝ ∂x j ∂xi ⎟ 3 ∂x k ⎥⎦
⎠
∂(U j ρc pT ) ⎡ ⎛ ∂T ⎞⎟⎤ ∂U j ⎡ ⎛ ∂U i ∂U j ⎞ 2 ⎤
=
∂
⎢αT ⎜⎜ μeff ⎥+ ⎢ μeff ⎜⎜ + ⎟ − μeff ∂u k ⎥ (10.33)
∂x j ∂x j ⎢⎣ ⎝ ∂x j ⎟⎠⎥⎦ ∂xi ⎢⎣ ⎝ ∂x j ∂xi ⎟ 3
⎠ ∂xk ⎥⎦
where
2
⎡ Cμ k ⎤
μ eff = μ ⎢1 + ⎥ (10.34)
⎣⎢ μ ε ⎦⎥
∂k τ ijt ∂U j ∂ ⎡ ∂k ⎤
Uj = −ε+ ⎢(α k μ eff ) ⎥ (10.35)
∂x j ρ ∂x j ∂x j ⎢⎣ ∂x j ⎥⎦
ε τ ij ∂U i ε2 ⎛ ⎞
∂ε ∂ ⎜ α ε μ eff ∂ε ⎟ − R
t
Uj = C ε1 − C ε2 + (10.36)
∂x j k ρ ∂x j k ∂x j ⎜ ∂x j ⎟⎠
⎝
where R in the ε equation is given by
⎛ η ⎞
C μ η 3 ⎜⎜1 − ⎟⎟
⎝ η0 ⎠ε
2
R= (10.37)
1 + βη 3 k
Sk
with η = , and η0=4.38, β=0.012. Other constants are [15]:
ε
Cμ=0.085; Cε1=1.42; Cε2=1.68
The term S is the modulus of the mean rate-of-strain tensor, Sij, which is defined as
S = 2S ij S ij (10.38)
where
1 ⎛⎜ ∂U i ∂U j ⎞
⎟
S ij = + (10.39)
2 ⎜⎝ ∂x j ∂xi ⎟
⎠
The RNG k-ε model yields an accurate description of how the effective turbulent
transport varies with the effective Reynolds number (or eddy scale). The coefficients αT, αk,
αε, in Eqs. (10.34)-(10.36) are the inverse effect Prandtl number for T, k and ε respectively.
They are computed using the following formula:
0.6321 0.3679
α − 1.3929 α + 2.3929 μ
= (10.40)
α 0 − 1.3929 α 0 + 2.3929 μ eff
where α0 is equal to 1/Pr, 1.0, and 1.0, for the computation of αT, αk, αε, respectively.
Low Reynolds k-ω Model
Unlike the k-ε model, it is easier to prescribe the boundary conditions in the k-ω model.
We know that k=0 on solid boundaries, and ω can be specified at the first few grid points
248 Li-Zhi Zhang
6μ
away from the wall as ω = , (y the distance to wall). The resulting equations for k, ω,
βy 2
and μt are:
k
ν t = α* (10.41)
ω
The equations for kinetic energy and specific dissipation rate are given as:
∂k ∂U i ∂ ⎡ ∂k ⎤
Uj = τ ij − β * ωk + ⎢(ν + σ k ν t ) ⎥ (10.42)
∂x j ∂x j ∂x j ⎢⎣ ∂x j ⎥⎦
∂ω σ ∂U i ∂ ⎡ ∂ω ⎤ 1 ∂k ∂ω
Uj = τ ij − β 2 ω2 + ⎢(ν + σ ω ν t ) ⎥ + 2σ ω2 (10.43)
∂x j ν t ∂x j ∂x j ⎣⎢ ∂x j ⎦⎥ ω ∂x j ∂x j
The model constants are provided in Ref [16, 17]:
4
5 ⎛⎜ Re t ⎞⎟
+
9 18 ⎜⎝ Rβ ⎟⎠
β =
*
4
(10.44)
100 ⎛ Re t ⎞
1+ ⎜ ⎟
⎜ R ⎟
⎝ β ⎠
Re t
α *0 +
Rk
α* = (10.45)
Re
1+ t
Rk
Re t
α0 +
α=
5
9
Rω
Re t
α* ( ) −1
(10.46)
1+
Rω
β=0.075, α0=0.1, σk=0.1, σω=0.5, Rβ=8, Rk=6, α0*=β/3, Rω=2.7

Full Reynolds Stress Modeling
The Reynolds stress model (RSM) is also considered as a choice. The RSM model
equation for the transport of Reynolds stresses is given by the following equation:
∂
∂x k
( )
ρU k u i' u 'j = −
∂
∂x k
[ (
ρ u i' u 'j u k' + p δ kj u i' + δ ik u 'j )]
∂ ⎡ ∂ ' ' ⎤ ⎛ ' ' ∂U j ∂U i ⎞
+
∂x k
⎢ μ u (
i j ⎥
u − )
ρ ⎜⎜ u i u k + u 'j u k' ⎟
∂x k ⎟⎠
(10.47)
⎣ ∂x k ⎦ ⎝ ∂x k
⎛ ∂u i' ∂u 'j ⎞ ∂u 'j
⎟ − 2μ ∂u i
'
+ p⎜ +
⎜ ∂x j ∂xi ⎟ ∂x k ∂x k
⎝ ⎠
The summation convention is used in the above equations. Therefore, there are totally 6
equations. In addition to the Reynolds stress transport equations, the dissipation rate is
modeled by the dissipation equation as in the standard k-ε model. A detailed description of
the model is given by Moore et al. [18] and Rokni et al. [19]. The default constants supplied
by CFD code for simulation are:
Cμ=0.09; Cε1=1.44; Cε2=1.92; C1ps=1.8
Boundary Conditions
At the walls, the no-slip condition is used. For k-ε and Reynolds stress models, the two-
layer-based, non-equilibrium wall functions methods are also employed. The key elements in
the non-equilibrium wall functions are as follows:
1. The log-law for mean velocity is sensitized to pressure-gradient effects.

2. The two-layer-based concept is adopted to compute the budget of turbulence kinetic
energy in the wall-neighboring cells.
3. The law of the wall for mean temperature remains the same as in the standard wall
function.
The wall-neighboring cells are assumed to consist of a viscous sublayer and a fully
turbulent layer, where the logarithmic law of the wall applies. This method requires some
consideration of mesh, i.e., the cell adjacent to the wall should be located to ensure that the
⎛ ρu τ y ⎞ ⎛ ρC μ1/4 k 1/2
p y
⎞
parameter y+ ⎜⎜ ≡ ⎟⎟ (uτ, friction velocity) or y* ⎜ ≡ ⎟ falls into the 30-60
⎝ μ ⎠ ⎜ μ ⎟
⎝ ⎠
range. In the present study, the y+ is adapted into 35-55 range. In addition, a uniform wall
temperature condition is assumed.
250 Li-Zhi Zhang
At the inlet, all dependent variables are assumed to enter the pipe with uniform profile in
the direction parallel to the corrugation of the upper wall, i.e,
u=u0, T=T0, k=k0, ε=ε0 (10.48)
The inlet boundary values of k and ε are computed from an estimated turbulence
intensity, I, and turbulent length scale, l, as follows:
3/ 2
k0 =
2
(u 0 I )2 , ε 0 = C μ3 / 4 k 0 (10.49)
3 l
'
The turbulent intensity I, defined as u / u , is equal to 5%, and the length scale l is set to
be 0.07⋅Dh in the present study, as suggested by Li et al. [15].
The exit boundary condition is treated as an outflow condition, which means that the
diffusion flux for all dependent variables are set to zero at the exit and an overall mass
balance is obeyed. This outflow boundary condition is true if the flow becomes fully
developed at a position far upstream from the exit because the accuracy of the exit boundary
condition should not affect the flow and heat transfer fields far upstream from the exit. The
results computed afterwards indicated that the flow becomes fully developed after 3-5 cycles
after the inlet. Therefore, the assumption is considered valid.
Solution Method
term is by second-order central difference scheme. The fluid in the study is selected as air.
Because of the intensive nonlinearity and coupling features of this problem, the under-
relaxation technique is applied to the iteration process to accelerate convergence. The
convergence criterion of
Rφn
n −1
≤ 10 − 4 (10.50)
Rφ
is applied for all equations, where Rφn refers to the maximum residual value summed over all
the computation cells after nth iteration. To test the criterion independence, another
convergence criterion of 10-5 is applied to a case. The difference of computed periodic Nu
numbers of the two convergence criterion is within 1%.
The grid independency test has also been done. The calculations were primarily carried
out with three different grid densities, 209985, 104992, and 419970 mesh points. The channel
fully developed periodic mean pressure drop and temperature change for the two fine grids
are almost the same and 10% higher than that for the coarse grid. For the finest grids, 419970,
the solution is too time-consuming, which is hard to use practically. Based on the above
experience, which establishes the grid independency, the final calculations are performed for
the 209985 grids and the results obtained in this paper refer to the grid geometry mentioned
above.
Model Validation
Selection of an appropriate turbulence model for numerical simulation requires

consideration of computational cost, anticipated flow phenomena, and the variables of
primary interest. The Nusselt numbers in the duct are of major concern in this case. Therefore
the agreements of the calculated Nu and experimental data are the main criterion for selecting
turbulence models.
For validation and comparison of the 4 turbulence models used, a cross-corrugated
triangular duct with geometric parameters identical to Scott and Lobato [1,2] is numerically
simulated: z0=2mm; Lcyc=2mm; y0=1mm. Figure 10.15 shows the experimentally obtained
(the discrete data were obtained from the experimentally obtained Sherwood correlations for a
90 degrees corrugation angle) and the calculated fully developed periodic mean Nusselt
numbers. As seen from this figure, of the 4 turbulence models employed, generally, the
Reynolds stress model (RSM) fits the experiment the best. The differences are from 5% to
11%, for Re ranging from 2000 to 20000. At lower Re, the low Reynolds k-ω model (LKW)
agrees the best, while the standard k-ε fits poorly. At higher Re, i.e., Re≥6000, the standard k-
ε model (SKE) gives the best prediction. As for the Renormalized group k-ε model (RNG
KE), it seems inappropriate to use such model for the corrugated triangular duct: the
differences between the calculated and experimentally found are substantial. In the low Re
range, the model over-predicts the Nusselt number by 58%, while in the high Re range, it
under-predicts the Nu by 21%.
In the following analysis, for Re ranging from 2000 to 2000, the RSM model is employed
to investigate the fluid flow and heat transfer. The imperfection with this model is that the
computational time is very long [17], due to large memory size required.
Turbulent Flow and Heat Transfer
width), for Re=10000. In the figure, the flow has two distinct patterns: in the corrugation
troughs of the upper wall, parallel flow is predominant, while in the troughs of the lower wall,
clockwise strong swirls are generated due to the reason that the fluid turns abruptly when
facing the trough walls of the lower wall. The shapes of the swirls in the valleys become
almost identical to each other, after 3-5 cycles, indicating a cyclic fully developed manner.
These swirls intensify the momentum transfer in the duct.
252 Li-Zhi Zhang
80
70 SKE
RNG SKE
60 LKW
RSM
50
Experimental
Nu d
40
30
20
10
0
0 5000 10000 15000 20000 25000
Re
Figure 10.15. Predictions of fully developed periodic mean Nusselt numbers with various turbulence
models.
Figure 10.16. Velocity vectors in the x-y midplane showing 5 cycles.
Figure 10.17. Velocity vectors in the y-z plane at x*=4.5.

Figure 10.17 shows the velocity vectors in the y-z plane at x*=4.5. This plane is
perpendicular to the main flow direction, and is located at the center of the 4th cycle. It can be
seen that in the corner regions, there exist appreciable secondary flows. These secondary
flows all exhibit the same pattern: departing from one wall, arriving at the other of the same
corner and leaving a small region very close to the corner where the fluid flows are almost
retarded. In the central part of the upper wall, the secondary flows are very weak and cannot
be observed clearly, while in the lower wall, the secondary flows are relatively strong, even in
the central part. There are semi-swirls around each corner. The interactions of these semi-
swirls generate demonstrable secondary flows in the central part of the lower wall valley.
These secondary flows generated swirls will intensify momentum and heat transfer.
For the cross-corrugated triangular ducts, the cyclic mean friction factor is one of the
most important parameters affecting heat exchanger design. Figure 10.18 shows the
calculated periodic mean friction factor along the duct length with four turbulence models.
The trend is like most of the developing flows: near the inlet region, the friction factor is very
high; with the flow propagates, after 3-5 cycles, it decreases gradually to the fully developed
value, fd. At the location near the outlet, the mean friction factor rises somewhat, which may
be attributed to the influence of outflow boundary conditions.
The corrugation usually leads to increased pressure drop. Figure 10.19 shows the
calculated fully-developed periodic mean friction factor with varying Reynolds numbers.
Figure 10.20 demonstrates comparisons between the friction factors in parallel flat plates
passages and in corrugated ducts. Under the same flow rates, the friction factor for the cross-
corrugated ducts is 3 times that for a parallel flat plates duct. Another feature with this figure
is that the fd decreases with an increasing Re.
To summarize the relations between the periodic mean friction factors with Renolds
numbers, a correlation has been proposed, with data obtained from the RSM model. The
correlation is:
f D = 1.9398 Re −0.27 (10.51)
for 2000≤Re≤10000
In contrast, friction factor in parallel flat plates is correlated by [11]
f D = (0.790 ln Re− 1.64)

−2
(10.52)
Temperature Distribution and Nusselt Number
After 3-5 cycles, the thermal flow has become fully developed, and the shapes of
isotherms in different valleys are similar. It’s clear that the isotherms in the last 1-2 valleys
have some distortion, which may also be attributed to the exit flow boundary conditions. The
temperature gradients at the walls are very high, indicating an enhanced heat transfer, due to
strong turbulence. Contrary to laminar flow, which usually has lower heat transfer on the
lower walls, the turbulence flow in this case has high heat transfer both on the upper walls
and on the lower walls, as a result from turbulence.
254 Li-Zhi Zhang
0.3
SKE
0.25 RNG SKE
LKW
RSM
fm 0.2
0.15
0.1
0.05
0
0 2 4 6 8 10 12
Cycles (x *)
Figure 10.18. Distribution of periodic mean friction factor along flow direction.
0.35
SKE
0.3 RNG SKE
LKW
0.25 RSM
0.2
fd
0.15
0.1
0.05
0
0 5000 10000 15000 20000 25000
Re
Figure 10.19. Calculated fully developed periodic mean friction factor.

0.4
Corrugated
Parallel flat plate
0.3
0.2
fd
0.1
0
5000 10000 15000 20000
Re
Figure 10.20. Comparisons of periodic mean friction factors with Reynolds numbers, cross-corrugated
and parallel flat plates.
70
65
60
55
50
Nu m
45
40
SKE
35
RNG SKE
30
LKW
25 RSM
20
0 2 4 6 8 10 12
Cycles (x *)
Figure 10.21. Distribution of periodic mean Nusselt numbers along flow direction.
Figure 10.21 shows the mean Nusselt numbers for each cycle, Num, along the duct length,
with four turbulence models. At the entrance, the cyclic mean Nusselt numbers are very high,
due to very thin boundary layers at the entrance. Along the flow direction, the cyclic mean
256 Li-Zhi Zhang
Nusselt numbers decrease rapidly in the first 3-5 cycles, and then arrive gradually at some
stable values, which is denoted as the fully developed value, NuD. Using the data from RSM
model and a least square curve fit technique, a correlation has been formulated, which is:
Nu D = 0.2338 Re 0.599 Pr 0.333 (10.52)
for 2000≤Re≤20000.
In contrast, heat transfer of the turbulence flow in parallel flat plates are governed by the
Dittus and Boelter equation [20]
Nu D = 0.023 Re 0.8 Pr 0.333 (10.53)
The variations of fully-developed cyclic mean Nusselt numbers with Re, which is
dictated by the above correlation, is depicted in Figure 10.22. The Nusselt numbers increase,
almost linearly with Re. For comparison, the fully developed Nusselt numbers in parallel flat
plates passages [20] are also plotted in this figure. As can be seen, the corrugation results in a
40-60% heat transfer enhancement. This is mainly due to the enhanced momentum transfer in
the cross-corrugations.
80
70 Corrugated
Parallel flat plates
60
50
Nud
40
30
20
10
0
5000 10000 15000 20000
Re
Figure 10.22. Comparisons of the fully developed Nusselt numbers for the corrugated duct and parallel
flat plates.
Three-dimensional turbulence flow and convective heat transfer in the entrance region of
a cross corrugated triangular duct which is proposed for a novel membrane module has been
studied numerically employing 4 turbulence models: standard k-ε, Renormalized group k-ε,
low Reynolds k-ω, and Reynolds stress model. Comparing to available experimental data, the
Reynolds stress model seems superior to others during the whole Reynolds range of 2000 to
20000. The cross-corrugation nature of the duct generates strong turbulence and secondary
flows to enhance heat transfer on the duct walls. Comparing to a parallel flat plate geometry,
the cross-corrugation can obtain a 40 to 60% heat transfer augmentation, however with a
penalty of 2 times more friction pressure drop. In addition, both the friction factor and the
heat transfer exhibit a cyclic manner, and the generated swirls rotate clockwise in the
corrugation valleys. The turbulence is almost uniform in the upper corrugation, but has large
variations in the lower corrugations and especially near the crests of valleys.
10.4. TRANSITIONAL FLOW

The aim of this section is to examine numerically flow structure and heat transfer
characteristics in a periodically fully developed cross-corrugated triangular segment. Some
correlations will be provided to estimate the cyclic fully developed friction factors and
Nusselt numbers in typical transitional flow regimes for Re=100 to 6000.
The flow in the cross corrugated duct is rather complex because of periodic convergent-
divergent nature. Various studies have found that transitions from laminar to turbulence occur
at Reynolds numbers as low as 150-500 [5-8], much lower than conventional flat-plates ducts.
Due to the small characteristic length and relative low air velocity (<5m/s), typical Reynolds
numbers for a cross-corrugated plate heat mass exchanger are in the range of 100 to 6000. It
is therefore not wise to use a simple laminar modeling technique. In contrary, since the flow
in the duct is transitional, numerical modeling should take into account of the turbulence
behaviors in the geometry.
Usually, as fully turbulent flows, transitional flows are modeled with turbulence models.
This section use the same turbulence models as described in above section. A fully developed
cyclic flow is considered.
Geometry
For a corrugation angle of 90 degrees between two neighboring plates, the fluid flows
predominantly along the furrows, i.e. in the corrugation on each of the plates (Focke et al.
[5]). The flow in this study is assumed parallel to a corrugation, and periodically fully
developed. It has been found that after 3-5 cycles, periodicity will be set up. The smallest
volume which is worthwhile resolving is one which represents a repeat or periodic building
block. The geometry of the problem considered here is shown in Figure 10.23.
258 Li-Zhi Zhang
Air flow y
z x
Figure 10.23. The unit cell used for CFD modeling.
Figure 10.24. The external mesh on the plate surfaces.

A computational mesh has been constructed to resolve this geometry; the external mesh
on the solid plates is shown in Figure 10.24. A body fitted mesh has been adopted for this
work so as to obtain good resolution of the corrugations. Due to the triangular nature of the
block, the computational domain is meshed with tetrahedral grids.
Three-dimensional numerical simulations of fluid flow and heat transfer in the
computational domain are conducted. The solution technique is based on a finite
difference/finite volume representation, while allowing for general body-fitted grids. The
SIMPLEC pressure-velocity coupling algorithm is used.
Boundary Conditions
The basic fluid dynamics equations are discretised onto the mesh described above and
solved numerically. The equations are elliptic in nature and as such, require boundary
conditions to be defined on all boundaries of the computational domain. At the walls, the no-
slip condition is used. Both constant temperature and constant heat flux are employed as
energy boundary conditions.
The flow is periodically fully developed, which means that the flow in the computational
domain is identical to that in the adjacent repeat segment. The flow field entering the domain
is identical to that leaving it, and the shape of the pressure distribution leaving the domain is
the same as that entering it, although the absolute value will be different. This difference is
required to sustain the flow, and it is in fact the pressure drop across the element. In other
words, periodicity is assumed for inlet and outlet boundary conditions. More specifically, a
constant mean pressure drop is imposed as a boundary condition for CFD calculations.
Various values of pressure drop give results under different air flow rates or Reynolds
numbers.
40
35
30
25
Nu m
20
Experimantal
15
LKW
10 SKE
RNG SKE
5 Laminar
0
0 1000 2000 3000 4000 5000 6000 7000
Re
Figure 10.25. Comparisons of 10-cycles-mean Nusselt numbers with different models.

260 Li-Zhi Zhang
term is by second-order central difference scheme. The convergence criterion and grid
independence test are the same for fully turbulence modeling.
Model Validation
For validation of the turbulence model, a cross-corrugated triangular duct with geometric
parameters identical to Scott and Lobato [1,2] is numerically simulated: zcyc=2mm;
Lcyc=2mm; ycyc=1mm, in another separate study. Totally 10 consecutive and repeated cycles
considering inlet developing conditions are modeled. The 10-cycles-mean Nusselt numbers
are calculated. Figure 10.25 shows the experimentally obtained values from the literature and
the calculated mean Nusselt numbers. For comparison, other 2 turbulence models, standard k-
ε (SKE), Renormalized Group k-ε (RNG KE), and a laminar model are used. For the k-ε
models, wall functions are needed to treat boundary conditions. As seen from this figure, of
the 4 turbulence models employed, only the LKW model fits the experiment well in the whole
transitional regime. The deviations are from 5% to 12%. For Re ranging from 100 to 500, the
laminar model is still acceptable. At high Re numbers, laminar prediction deteriorates
drastically, especially when Re>2000. Standard k-ε and RNG k-ε models are only applicable
at higher Re numbers greater than 6000. The deviations of experimental values from literature
and the predicted Nusselt numbers with these two models could be as high as 65%.
Transitional Flow and Heat Transfer
width), for Re=1000. In the figure, the flow has two distinct patterns: in the corrugation
troughs of the upper wall, parallel flow is predominant, while in the troughs of the lower wall,
clockwise strong swirls are generated due to the reason that the fluid turns abruptly when
facing the trough walls of the lower wall. These swirls intensify the momentum transfer in the
duct.
Figure 10.27 shows the velocity vectors in the y-z plane at x*=0.5. This plane is
perpendicular to the main flow direction, and is located at the center of the unit cell. It can be
seen that the flow is rather complex, but with regular and interesting patterns. There are
strong double swirls in the corner regions of the lower wall, which rotate in a clockwise and a
counter-clockwise manner, respectively. The secondary flows in the upper corrugation are
relatively weak. In summary, in the upper corrugations, parallel flows are predominant, while
in the lower corrugation, secondary swirl flows are dominant. These secondary flows
generated swirls will intensify momentum and heat transfer.
Figure 10.26. Velocity vectors in x-y plane at z*=1.5.
Figure 10.27. Velocity vectors in y-z plane at x*=1.5.
Temperature contours are shown in Figure 10.28 for an x-y plane at z*=0.5 when
Re=1000. The temperature gradients near the lower walls are very high, indicating an
enhanced heat transfer, due to strong turbulence. The secondary flows are able to bring the
main stream fluid closer to the solid surfaces of the plates and therefore increase the rate of
heat transfer. Enhancing these secondary flows lead to an increase in heat transfer coefficient.
262 Li-Zhi Zhang
Figure 10.28. Temperature contours in x-y plane at z*=0.5.
Figure 10.29 shows the cyclic mean friction factors under various Reynolds numbers for
the developed flow. The friction factor decreases drastically with Re when the Re is below
2000, but decreases gradually afterwards. A correlation has been formulated to reflect the f-
Re relations as
f = 6.5336 Re −0.421 (10.54)
To make comparisons with the cross corrugated sinusoidal passages, the experimental
data from Ref [5] are also plotted in the figure. It can be seen that when Re<500, two
geometries have similar friction factors. When at higher Reynolds numbers, the data for
cross-corrugated sinusoidal ducts are much higher than the corrugated triangular ducts. It
should be mentioned that the measurements have 5-7% uncertainties [5].
Figure 10.30 shows the cyclic mean Nusselt numbers for various Reynolds numbers. The
Nusselt numbers increase, almost linearly with Re. This is mainly due to the enhanced
momentum transfer in the cross-corrugations, more specifically, by secondary flows and
swirls. Heat transfer at lower Reynolds numbers is relatively low since the laminar free shear
layers of the separated regions are an additional resistance. Transfer across turbulent-free
shear layers is, however, usually very effective owing to the absence of a restraining effect
such as solid wall. The additional resistance is therefore effectively removed when the free
shear layers become turbulent.
Two correlations are proposed for uniform temperature and uniform heat flux boundary
conditions. They are:
Nu = 0.1922 Re 0.599 Pr 0.333 (10.55)
for uniform temperature, and
Nu = 0.2743 Re 0.569 Pr 0.333 (10.56)
for uniform heat flux conditions, respectively.

1.2
Cross Triangular
1 Cross Sinusoidal, ref[5]
0.8
0.6
f
0.4
0.2
0
0 1000 2000 3000 4000 5000 6000
Re
Figure 10.29. Variations of cell mean friction factor with Reynolds numbers.
45
uniform flux
40
uniform tem
35 Cross Sinusoidal, ref [5]
30
25
Nu
20
15
10
0
0 1000 2000 3000 4000 5000 6000
Re
Figure 10.30. Variations of cyclic mean Nusselt numbers with Reynolds numbers.
264 Li-Zhi Zhang
The experimental data from Ref [5] for a cross-corrugated sinusoidal duct are also plotted
in the figure. Generally, the Nusselt numbers for a cross-corrugated sinusoidal ducts lie
between the Nusselt numbers of triangular ducts under uniform temperature and uniform heat
flux boundary conditions. Certainly differences are expected due to the disparities in the two
geometries, but the trends are similar.
10.5. CONVECTIVE MASS TRANSFER

Besides heat transfer, mass transfer can be modeled with the same strategy. The turbulent
mass concentration conservation equation is [21]
∂
(ρU j Yv ) = ∂ q j + q tj
( ) (10.57)
∂x j ∂x j
μ ∂Yv
qj = ; q j = −ρu j Yv
t ' '
(10.58)
Sc ∂x j
where ρ, μ, and Sc are density, viscosity, Schmidt number, respectively, and qj is mass flux in
kg/m2. The turbulent mass transfer term in Eq.(10.58) is determined by the following equation
μ t ∂Yv
q tj = (10.59)
Sct ∂x j
where the turbulent Schmidt number Sct=1.0 [21]. These equations can be combined with
turbulence model to be solved.
Two correlations have been proposed to estimate the fully developed cyclic mean friction
factors and cyclic mean Sherwood numbers. They are:
f D = 9.665 Re −0.486 (10.60)
ShD = 0.266 Re 0.539 Sc 0.333 (10.61)
For 100≤Re≤3000.
For comparison, the experimental Sherwood correlation given in ref [1] for 90 degrees
corrugation is
ShD = 0.268 Re 0.56 Sc 0.333 (10.62)

Table 10.2. Structural parameters of total heat exchanger
Parameters Unit Value

Exchanger length mm 200.0
Exchanger width mm 200.0
Channel height mm 1.95
Number of channels for each flow 100
Exhaust air temperature °C 27
Exhaust air humidity kg/kg 0.010
Fresh air temperature °C 35
Fresh air humidity kg/kg 0.021
Plate thickness μm 105
Table 10.3. Performance comparison of four total heat exchanger configurations
Case 1 Case 2 Case 3 Case 4

Structure Parallel-plates Parallel-plates Cross- Cross-
corrugated corrugated
Plate type Solid membrane CSLM Solid CSLM
membrane
Dh (mm) 3.96 3.96 2.0 2.0
Re 249 249 126 126
NuD 7.54 7.54 4.67 4.67
ShD 7.10 7.10 4.39 4.39
Air side heat transfer 0.05 0.05 0.061 0.061
coefficients (kWm-2K-1)
Air side mass transfer 0.051 0.051 0.062 0.062
coefficients (m/s)
fD 0.3855 0.3855 1.1856 1.1856
Plate heat resistance 0.3125 0.3852 0.3125 0.3852
(m2s/kW)
Plate moisture resistance 114.5 61.8 114.5 61.8
(s/m)
Total transfer area (m2) 8.0 8.0 16.0 16.0
Number of Transfer Units 4.15 4.15 10.0 10.0
for sensible heat
Number of Transfer Units 1.08 1.65 1.42 2.29
for latent heat
Sensible effectiveness 0.72 0.72 0.82 0.82
Latent effectiveness 0.47 0.57 0.55 0.64
Pressure drop (Pa) 12 12 71 71
266 Li-Zhi Zhang
10.6. COMBINED WITH MEMBRANES

Considering four design strategies for a total heat exchanger: (1) Parallel plates duct
structure with solid polymer membrane; (2) Parallel plates duct structure with composite
supported liquid membrane (CSLM); (3) Cross-corrugated triangular duct structure with solid
polymer membrane; and (4) cross-corrugated triangular duct structure with CSLM. The
structural parameters and operating conditions are listed in Table 10.2. For the traditional
solid polymer membrane, the membrane thickness is selected as 100μm.
Table 10.3 lists the heat and mass transfer properties of the four total heat exchangers. As
seen, the cross-corrugated structure intensifies air side convective heat and mass transfer
substantially. Compared to a parallel-plates duct, the convective heat and mass transfer
coefficients in a cross-corrugated duct increased by 22%. Besides, transfer area increased by
1 fold. These two factors make the new structure more compact and have a higher heat mass
transfer capabilities. As a penalty, the pressure drop increased from 12 Pa to 71 Pa.
Compared to a common solid membrane, the composite supported liquid membrane has a
46% less membrane side mass transfer resistance. Consequently, the CSLM could increase
latent effectiveness by 8%, even with the same traditional parallel-plates duct structure. The
effect of CSLM on decreasing heat transfer resistance is not obvious, since the dominant
resistance in the unit is in air side. However, when the CSLM is combined with the cross-
corrugated triangular duct, it will have the highest sensible effectiveness of 0.82, and the
highest latent effectiveness of 0.64, which is highly desired in market.
10.7. CONCLUSION
Convective heat mass transfer coefficients in cross-corrugated triangular ducts are solved
with CFD codes. Laminar model and turbulence models are used. Correlations are obtained
for friction factor and Nusselt numbers. Laminar model is only useful under very low
Reynolds numbers. Standard k-ε model is only appropriate for high Reynolds numbers. Low
Reynolds number k-ω turbulence model predicts the results well under transitional flow
regime.
Cross-corrugated triangular ducts are one kind of primary surface heat exchangers. The
cross-corrugation nature of the duct generates strong turbulence and secondary flows to
enhance heat transfer on the duct walls. Comparing to a parallel flat plate geometry, the cross-
corrugation can obtain a 40 to 60% heat transfer augmentation, however with a penalty of 2
times more friction pressure drop. In addition, both the friction factor and the heat transfer
exhibit a cyclic manner, and the generated swirls rotate clockwise in the corrugation valleys.
The turbulence is almost uniform in the upper corrugation, but has large variations in the
lower corrugations and especially near the crests of valleys.
REFERENCES
[1] Scott, K.; Lobato, J. Mass transport in cross-corrugated membranes and the influence
of TiO2 for separation processes. Industrial Engineering Chemical Research, 2003, 42,
5697-5701.
[2] Scott K., Mahmood A.J., Jachuck R.J., Hu B., 2000. Intensified membrane filtration
with corrugated membranes. Journal of Membrane Science 173, 1-16.
[3] Zhang, L.Z. Convective mass transport in cross-corrugated membrane exchangers.
[4] Okada, K.; Ono, M.; Tomimara, T.; Okuma, T.; Konno, H.; Ohtani, S. Design and Heat
transfer characteristics of new plate heat exchanger. Heat Transfer-Japanese Research,
1972, 1, 90-95.
[5] Focke, W.W.; Zachariades, J.; Olivier, I. The effect of the corrugation inclination angle
on the thermohydraulic performance of plate heat exchangers. International Journal of
Heat and Mass Transfer, 28, 1469-1479, 1985.
[6] Gaiser, G.; Kottke, V. Effects of corrugation parameters on local and integral heat
transfer in plate heat exchangers and regenerators. Proceedings of 9th International
Heat Transfer Conference, Jerusalem, vol.5, pp.85-90, 1990.
[7] Ciofalo, M.; Stasiek, J.; Collins, M.W. Investigation of flow and heat transfer in
corrugated passages-2. Numerical simulations. International Journal of Heat and Mass
Transfer, 1996, 39, 165-192.
[8] Stasiek, J. Ciofalo, M.; Smith, I.K.; Collins, M.W. Investigation of flow and heat
transfer in corrugated passages-I Experimental results. International Journal of Heat
Mass Transfer, 1996, 39, 149-192.
[9] Muley, A.; Manglik, R.M. Enhanced heat transfer characteristics of single-phase flows
in a plate heat exchanger with mixed chevron plates. Journal of Enhanced Heat
Transfer, 1997, 4, 187-201.
[10] Ergin, S.; Ota, A.; Yamaguchi, H. Numerical study of periodic turbulent flow through a
corrugated duct. Numerical Heat Transfer Part A-Applications, 2001, 40, 139-156.
[12] Sherony, D.F.; Solbrig, C.W. Analytical investigation of heat or mass transfer and
friction factors in a corrugated duct heat or mass exchanger. International Journal of
Heat and Mass Transfer, 1970, 13, 145-159.
[13] Patel, C.V.; Rodi, W.; Scheuerer, G. Turbulence models for near wall and low
Reynolds number flows: a review. AIAA Journal, 1984, 23, 1308-1319.
[14] Mompean, G. Numerical simulation of a turbulent flow near a right-angled corner
using the speziale non-linear model with RNG k-ε equations. Computers and Fluids,
1998, 27, 847-859.
[15] Li, L.J.; Lin, C.X.; Ebadian, M. A. Turbulent mixed convective heat transfer in the
entrance region of a curved pipe with uniform wall temperature. International Journal
of Heat and Mass Transfer, 1998, 41, 3793-3805.
[16] Jones, R.M.; Harvey, A.D.; Acharya, S. Two equation turbulence modeling for
impeller stirred tanks. ASME Journal of Fluids Engineering, 2001, 123, 640-648.
268 Li-Zhi Zhang
[17] Song, B.; Amano, R.S. Application of non-linear k-ω model to a turbulent flow inside a
sharp U-bend. Computational Mechanics, 2000, 26, 344-351.
[18] Moore, E.M.; Shambaugh, R.L.; Papavassiliou, D.V. Analysis of isothermal annular
jets: comparison of computational fluid dynamics and experimental data. Journal of
Applied Polymer Science, 2004, 94, 909-992.
[19] Rokni, M.; Sunden, B. Calculation of turbulent fluid flow and heat transfer in ducts by
a full Reynolds stress model. International Journal of Numerical Methods in Fluids,
2003, 42, 147-162.
[20] Incropera, F.P.; Dewitt, D. P. Introduction to Heat Transfer. New York: John Wiley &
Sons, Chapter 8, pp. 392. 1996.
[21] Longest Jr, P.W.; Kleinstreuer, C.; Kinsey, J.S. Turbulent three-dimensional air flow
and trace gas distribution in an inhalation test chamber. Journal of Fluids Engineering,
2000, 122, 403-411.
Chapter 11
APPLICATIONS OF TOTAL HEAT RECOVERY
ABSTRACT
Conditioning of fresh air constitutes 20-40% of the total load for air conditioning
industry, which is expanding rapidly. How to combine the total heat recovery technology
with air conditioning industry is an attracting task. In this chapter, some novel air
conditioning technologies that combining fresh air total heat recovery are introduced and
analyzed. System modeling is performed to have a comparison of energy savings effect.
The results found that energy performance could be greatly improved with total heat
recovery measures. Further, besides energy savings, indoor air quality could also be
improved.
NOMENCLATURE
Δp total pressure rise (Pa),

ACH Air infiltration rate (h-1)
COP Coefficient of performance
cp Specific heat (kJkg-1K-1)
h Specific enthalpy (kJ/kg)
m s Indoor moisture generation rate (kg/h)
Q Energy or moisture (W or kg)
q Heat (kW)
T Temperature (K)
Va Volumetric flow rate (m3/h)
Vr Room volume (m3)
x Degree of dryness
Greek Letters
ε Effectiveness
τ time (s)
270 Li-Zhi Zhang
η Efficiency
ρ Density (kg/m3)
ε Effectiveness
ω Humidity (kg vapor/kg air)
Subscripts
a Air stream
c Condenser
com Compressor
e Evaporator
f Refrigerant
fan Fan
motor Motor
s isentropic
11.1. INTRODUCTION
Approximately one-third of the primary energy resources are consumed in the air-
conditioning sector. The depleting nature of primary energy resources, negative
environmental impact of fossil fuels and low exergetic efficiencies obtained in conventional
space heating and cooling are the main incentives for developing alternative heating,
ventilating and air-conditioning (HVAC) techniques which can either save energy or employ
low-grade thermal energy sources. Novel air conditioning systems with total heat recovery
are the directions for sustainable development of HVAC industry.
Humidity has been increasingly raised as a big issue. Besides temperature, humidity is
another important parameter influencing people’s feeling of thermal comfort. Figure 11.1
shows the comfort zone in a psychrometric chart [1]. As seen, in summer, the narrow zone
between operative temperature 24°C and 27°C, humidity between 4g/kg and 20°C wet bulb
are the acceptable levels of thermal comfort. People will feel uncomfortable whether it’s too
dry or too humid. The design of HVAC systems for thermal comfort requires increasing
attention, especially in the light of recent regulations and standardization on ventilation [2], so
that an optimal level of indoor humidity may be reached and maintained to ensure a
comfortable and healthy environment and to avoid condensation damage for building
envelope and furnishings.
Applications in Total Heat Recovery 271
Humidity ratio (g/kg)

0 5 10 15 20
35
18
20
°C
RH
RH
°C
WB
％
WB
50
60
Operative temperature (°C)
30
25 Summer
Winter RH
0％
10
20
H
30％R
15
Figure 11.1. Comfort zone in a psychrometric chart.
Air-conditioning in hot and humid environments has become an essential requirement for
support of daily human activities. Humidity problems can be found in many applications
including office buildings, supermarkets, art galleries, museums, libraries, electronics
manufacturing facilities, pharmaceutical clean rooms, indoor swimming pools and other
commercial facilities. According to Figure 11.1, for thermal comfort reasons, indoor air
conditions around 25°C temperature and 10g/kg humidity ratio are the accepted set points.
However, the Southern China and other Southeast Asia countries have a long summer season
with a daily average temperature of 30°C, and humidity ratio above 20g/kg. Outdoor relative
humidity often exceeds 80% continuously for a dozen of days, leading to mildew growth on
wall and furniture surfaces, which affects people’s life seriously. In spring in Southern China,
there is a period named “Plum raining seasons” when it rains continuously for one to two
months. People can not see sun for a long time and stuff from quilts to grains gets moldy
easily. Consequently, mechanical air dehumidification plays a major role in air conditioning
industry in these regions. In many of these countries, the energy used to cool and dehumidify
the ventilation air ranges from 20 to 40% of the total energy consumption for air conditioning,
and can be even higher where 100% fresh air ventilation is required, such as kitchen, hospital,
factories.
Part load is another problem. In hot and humid climates, air conditioning is an
indispensable component to maintain a comfort indoor environment with lower temperature
and humidity than outdoor conditions. Operating under hot and humid outdoor conditions, air
conditioning has to deal with both sensible and latent loads in a space. In many cases to deal
272 Li-Zhi Zhang
with space latent cooling load using a small HVAC system is often challenging and difficult
[3]. The air conditioning system’s design load is calculated based on the number of occupants
and their level of activity, types and quantity of equipments used in space, solar irradiation
experienced, heat transmission through the building materials, heat gained from infiltrated
outdoor air and many other factors. In reality, the space loads are always below their design
values. Under part-load conditions, the common practice is to employ control method to
maintain the space temperature while allowing the space humidity to vary. In part load
conditions, supply air temperature is reduced. It is still enough to extract sensible load, but is
insufficient to extract latent load. The indoor humidity is out of control. In full load seasons
from June to September, humidity is controlled very well, but in other transit seasons,
humidity of the space may drift towards a high value that causes human discomfort while
supporting the growth of pathogenic or allergenic organisms. It is also believed that the
emission rate of formaldehyde from furniture and building materials is higher when humidity
rises, resulting in poor indoor air quality.
Stringent ventilation regulations make the humidity problem more serious. In modern
society, people spend most of their time in built environments. More attention has been paid
to indoor air quality and indoor thermal comfort. HVAC systems are necessary for almost all
buildings. However, conventional air conditioning modes, such as constant air volume (CAV)
systems and variable air volume (VAV) systems, face great challenges in effective outdoor air
ventilation and precision indoor air humidity control. From the view points of ventilation, the
main problems with conventional air conditioning systems are analyzed as follows.
(1) The outdoor air in conventional air conditioning systems mixes with the re-circulated
air, which causes transmission of bacterium and virus among multiple zones. In this
situation, occupants are at high risk of infection when diseases breakout, like SARS
and bird flu. Human’s expectation of effective ventilation with 100% outdoor air has
been increasingly rising.
(2) The indoor relative humidity tends to rise under part load operation because the air
conditioning systems usually control the indoor temperature by reducing their
cooling capacities. To control load, cooling-reheating processes are required, which
are very much energy intensive. This problem is very serious in hot and humid
regions, like Canton. To improve humidity control, the method of decoupling
temperature control and humidity control has attracted much attention. To realize
independent humidity control, an independent fresh air conditioning system, or
known as dedicated outdoor air system, is always required.
(3) New technologies for a more comfortable and energy efficient indoor environment,
such as chilled ceiling/beam, thermal storage and VRV (Variable Refrigerants
Volume), require parallel independent fresh air conditioning systems to meet
demands on effective ventilation and removal of latent load. However, the energy
consumption for dehumidifying fresh air is huge, which often accounts for 20%-40%
of the total energy for air conditioning in hot and humid areas. The unaffordable
energy cost for treatment of fresh air, particularly for fresh air dehumidification
seriously restricts the application of independent fresh air conditioners. Total heat
recovery becomes a necessity.
Conventionally cooling coils are used to cool and dehumidify supply air. It is called the
coupled cooling since cooling and dehumidification are accomplished simultaneously in a
coupled way. To dehumidify air, air temperature must be cooled to below dew point
temperature like 10°C. Dehumidified air of such low temperature cannot be supplied to the
space directly because people may feel cold draft under the cold air stream. Reheating has
been widely used in many applications behind a cooling coil to prevent this problem.
However this cooling-reheating process is energy intensive. Energy is needed to overcool the
air across the cooling coil and also to reheat the off-coil air to the desired space humidity and
temperature. Although reheating is able to maintain a space at its design temperature and
humidity during pert-load conditions, it is not often a recommended practice chiefly due to its
high energy use.
To solve this problem, nowadays there is a trend to separate the treatment of sensible load
and latent load. This is the so called independent humidity control. According to this scheme,
sensible load is treated by chilled-ceilings, cooling coils, or air handling units which still
cools the supply air but doesn’t necessarily cool it as low as to dew point. Supply air
temperature can be adjusted as sensible load requires. The latent load is accomplished by an
independent dehumidification unit, which is to treat all latent load alone. How to combine
these systems with total heat recovery is a challenging yet interesting work.
11.2. DESICCANT WHEEL WITH CHILLED CEILING

Desiccant wheels have two uses: active air dehumidification and total heat recovery.
Energy wheels are simple, but incoming air cannot be dehumidified to states drier than
exhaust air. In most air-conditioning systems, incoming air should be dehumidified to be drier
than room air, say, 7g/kg, to further extract the moisture load of a building. In this case, to
generate dry air, hot air is required to regenerate wheels. Desiccant cooling is a cycle
involving desiccant wheel that can produce dry and cool air. In a desiccant cooling system,
enthalpy from exhaust air is recovered through evaporative cooling of exhaust air and the
subsequent sensible heat exchange.
Almost all materials have the capacity to adsorb and hold water vapor, but commercial
desiccants have significant capacity for holding the water. A commercial desiccant takes up
between 10 and 1100 percent [4] of its dry weight in water vapor, depending on its type and
the moisture available in the environment. Desiccants remove moisture from the surrounding
air until they reach equilibrium with it. This moisture can be removed from the desiccant by
heating it to temperatures between 50°C and 260°C and exposing it to a scavenging air
stream. The desiccant is then cooled so that it can adsorb moisture again. The transfer of
moisture is due to the difference in vapor pressure at the desiccant surface and that of the
surrounding air. When the vapor pressure at the desiccant surface is less than that of air, the
desiccant attracts moisture and releases it when its vapor pressure is greater than that of air.
The major advantages of desiccant cooling are:
a) Only air and water are required as working fluids. Fluorocarbons are not required;
thus, there is no impact on the ozone layer.
274 Li-Zhi Zhang
b) The source of thermal energy can be diverse (i.e., solar, waste heat, natural gas). The
electrical energy requirement can be less than 25 percent of conventional
refrigeration systems.
c) Since desiccant systems operate near atmospheric pressure, maintenance and
construction are simplified.
Because of these advantages, much effort has been spent in the researches and
applications of desiccant cooling cycles [5-7]. This cycle is very attractive since a relatively
low temperature heat source, such as can be provided by solar energy, is suitable for the
regeneration of the desiccant in the cycle. However, there are still two shortcomings that need
to be addressed. Firstly, the supply air temperature in such systems is 4°C higher than that in
a conventional all-air system, which means that a 60% greater volume of air needs to be
pumped and circulated through the system. As a result, fan energy, which is already
astonishingly large in a conventional all-air system, would be further aggravated. This would
compromise the energy savings effect of the system. Secondly, in hot and humid regions like
Canton, the effectiveness of desiccant dehumidification would be quite limited, which means
that a heavier desiccant wheel and/or a higher regenerating temperature are required to realize
efficient dehumidification and cooling. The system would become bulky and the initial
investment would skyrocket.
Decoupled cooling, in which room sensible load is extracted by chilled-ceiling, whereas
the latent load is extracted by a desiccant system, is introduced here. In the system, aluminum
ceiling panels are installed under the cement ceiling. Chilled water with 17 °C flows through
the metal tubes connected with metal-sheet panels, removing heat collected by the panels,
mainly radiantly. Certainly the metal tubes and plates can be replaced by plastics, to reduce
cost. With this desiccant cooling combined with a chilled-ceiling system, the volume of the
process air becomes the minimum: the process air is also the fresh air for ventilation.
Therefore, large fractions of the fan energy can be saved. Other advantages of this combined
system include:
a) Chilled-ceiling systems improve thermal comfort because heating and cooling are
provided directly and more evenly to the occupants without causing drafts.
b) Radiant cooling by ceiling panels results in a temperature gradient in the occupied
zone varying from 0 to 2°C/m. Consequently, chilled ceiling system creates a “cool
head and warm feet” radiant environment, which is more comfortable than a “warm
head and cool feet” environment created by conventional all air systems.
c) The air temperature in the room is higher at the same operative temperature due to
the radiant effect, in comparison with an all-air system. This will result in energy
savings.
d) The latent load is treated by a desiccant cooling cycle. The chilled water to the
ceiling panels is supplied by a chiller. Without the need to dehumidify air, the
evaporating temperature for the chiller could be raised from conventional around 5°C
to as high as 15°C, which means large improvements in COP and large quantities of
energy savings.
Conventional all-Air System
To have a comparison, a conventional constant volume all-air system is first introduced.

The schematic of the system and the air treatment processes in a psychrometric chart are
shown in Figures 11.2a and 11.2b respectively. Ambient air is mixed with return air first
before the mixed air (state 2) flows through an AHU (Air Handling Units) cooling coils
where it is cooled and dehumidified. Air leaving cooling coils at state 4 (13°C, and 95% RH
in summer design conditions) is then pumped to the conditioned space through supply air
ducts where air temperature rises about 1°C due to heat gains on the trip (4 to 5). After
absorbing the building’s sensible and latent loads, air flows to the return air ducts to be mixed
with fresh air or exhausted to the surroundings. Energy wasting process of re-heating after
cooling coills are omitted and the supply air temperature is determined by sensible load.
Therefore, temperature is actively controlled, while humidity is passively controlled. The
result of this approach is a loss of humidity control while the temperature is maintained at
setpoint. In winter and transient seasons, sensible load becomes smaller and air is not
sufficiently dehumidified. As a consquence, indoor humidity becomes higher in these periods.
It is believed that humidity levels above 70% RH in the supply air duct rapidly increase the
effect of bacteria, viruses, fungi, and mites. The system has no total heat recovery.
Desiccant Cooling (DC) with Chilled-Ceiling
This is a de-coupled cooling strategy [8]. The schematic of the system components and
the corresponding psychrometric processes are shown in Figure 11.3. Chilled-ceiling panels
are equipped to extract sensible load, and dehumidified fresh air is supplied by a desiccant
cooling system. The principles of the desiccant cooling cycle is well-known. Ambient outdoor
air at state 1 enters the supply air duct. This air passes through a desiccant wheel and hot, dry
air exits at state 2. This increase in temperature is due to the heat of sorption and some
sensible heat transfer. The hot, dry supply air transfers much of this heat to the return air
stream in process 2 to 3 involving a sensible heat wheel. Unlike a commonly used desiccant
cooling, where air at state 3 is evaporatively cooled to state 4, in this system, the warm, dry
air at state 3 is cooled by a cooling coil, to keep the dryness of supply air at state 4. The cool,
dry air at state 4 is then distributed to the room. After accepting the building latent load and a
small amount of sensible load, the air then returns to the desiccant system through return air
ducting. This is the state of the air which corresponds to state 5. This somewhat cool, fairly
dry air is evaporatively cooled to as low a temperature as possible at state 6. This cold, damp
air is then preheated by the rotary sensible wheel to state 7 while cooling the supply air
stream. State 7 is the state of the moist air as it enters the heating coil. Hot, humid air exits at
state 8 and regenerates the desiccant wheel. Warm, very humid air at state 9 is then exhausted
to the surroundings. Chilled water flows through the cooling coil and chilled ceiling panels in
series, implying that only one refrigerator with an evaporating temperature as high as 15°C is
needed. In this scheme, partial total heat recovery is realized by the evaporative cooling of
exhaust air.
276 Li-Zhi Zhang
3 2 1
4
Ambient
Room Cooling Coil
5
Exhaust
(a) Schematics.
0.03
0.025
Humidity ratio (kg/kg)
0.02 φ =100%
1
0.015
2
0.01
5
3 4
0.005
0
10 15 20 25 30 35 40
Dry bulb temperature (°C)

(°C)
(b) Psychrometrics.
Figure 11.2. Conventional constant volume all-air system, (a) schematics and (b) psychrometrics.
Ventilation distribution is provided by displacement ventilation. Displacement ventilation

brings air into the space near floor level at a low velocity and air is exhausted at the ceiling.
This strategy delivers fresh air to where the occupants require it and odors and airborne
contaminants are carried to the ceiling and exhausted instead of being recirculated as is
common with traditional HVAC systems.
12 11 Sensible Heat Wheel
6 7 8 9
5
Exhaust
Chilled Ceiling
Evaporative Cooler Gas Heater
Room
3 2 1
4
Ambient
10 11 Desiccant Wheel
Cooling Coil
(a) Schematics.
Figure 11.3. (Continued on next page.)
0.03
φ =100%
0.025 9

1
0.02
0.015 6 7
8
0.01 5
4 3 2
0.005
0
10 20 30 40 50 60 70 80 90 100
(b) Psychrometrics.
Figure 11.3. A desiccant cooling combined with chilled-ceiling, (a) schematics and (b) psychrometric
processes.
Pre-cooling Desiccant Cycle with Chilled-Ceiling
It was discovered that in hot and humid regions, very high regenerating temperatures are
required, otherwise the performance of the desiccant system would deteriorate seriously in
extreme weather conditions. To address this problem, a modification of a desiccant cycle, a
Pre-cooling desiccant cooling cycle (PCDC), has been proposed and combined with chilled
ceiling panels for an office building [9]. A PCDC cycle could use lower regenerating
temperatures in comparison with a DC cycle.
Figure 11.4a illustrates a schematic of this cycle, while Figure 11.4b shows this cycle on
a psychrometric chart. The components are similar to a DC cycle, but an additional total heat
exchanger, which could be either an energy wheel or a membrane-based total heat exchanger,
is added to the system. Ambient air at state 1 first passes through a total heat exchanger where
it exchanges sensible heat and moisture with the exhaust air. Air at state 2 then passes through
the rotating dehumidifier matrix and hot, dry air exits at state 3. The hot, dry supply air
transfers much of its heat to the return air stream in process 3-4 involving a sensible heat
wheel. The warm, dry supply air at state 4 is now cooled by a cooling coil to state 5 and is
supplied to the conditioned space.
After accepting the building latent load and some sensible load, the air then returns to the
air ducting. This room air, which is still relatively cool and dry, is first ushered into the total
heat exchanger to pre-cool and pre-dry the outside fresh air, through process 6-7. Air exited
from the total heat exchanger at state 7 is evaporatively cooled to state 8 and is used to cool
the supply air in the sensible heat exchanger involving process 8-9. The air at this stage is
further heated by a gas heater to state 10 and is used to regenerate the desiccant wheel. After
278 Li-Zhi Zhang
passing through the regenerating desiccants matrix passages, air becomes hot and humid and
is exhausted to the outside.
14
Total Heat Exchanger
13
7
6
Chilled Ceiling
12 Sensible Heat Wheel

Room 13
4 3 2 1
5
Ambient
Cooling coil
Gas Heater
8 9 10 11
Exhaust
Evaporative Cooler Desiccant Wheel
(a) Schematics.
0.03
φ =100%
0.025 11
1
0.02 8
10
9
7
0.015
2
0.01 6
3
5 4
0.005
0
10 20 30 40 50 60 70 80 90 100
(b) Psychrometrics.
Figure 11.4. A Pre-cooling desiccant cooling combined with chilled-ceiling, (a) schematics and (b)
psychrometric processes.
The temperature of the return air at state 9 required to regenerate the desiccant is a
property of the type of desiccant used and the amount of dehumidification required in process
2-3. Many different desiccants have been considered for use in dehumidification. Some of the
more common ones include silica gel, lithium chloride, lithium bromide, zeolites. Most
industrial rotary dehumidifiers use silica gel or lithium chloride as the desiccant when lower
regeneration temperatures are desired.
This system takes full use of the total heat recovery.
Component Modeling
Ceiling Panels and Cooling Load

Thermal performance is evaluated by a special cooling load program for cooled ceiling
systems ACCURACY [8] which is developed and validated previously at Delft University of
Technology in Netherlands. This is a room-energy-balance method, according to which,
chilled-ceiling panels are treated as individual surfaces that exchange heat convectively with
room air and radiantly with other building surfaces [10]. Heat conduction within the ceiling
panels is treated in a one-dimensional way using a three-nodal point model. The program
calculates not only the cooling load, but also the required supply water or air temperatures for
different panel installation areas. The program also works out the respective convective and
radiant heat extractions by cooling panels, and the radiant and convective heat from all the
window and wall surfaces. Adopting such building dynamics simulation techniques, year
round simulation of rooms with or without chilled-ceilings provides hour-by-hour data of
cooling/heating loads, and temperatures of room air and other components.
Moisture balance in the room is calculated by [11],
V r ρ r dω r
= −vs ρ s (ω r − ω s ) + ACH ⋅ Vr ρ o (ω o − ω r ) + m s − m ad/de (11.1)
dτ
Moisture load in the room is calculated by,
QL = ACH ⋅ Vr ρ o (ω o − ω r ) + m s − m ad/de (11.2)
where Vr is the room volume (m3), ρ is air density (kg/m3), ω is humidity ratio (kg/kg), τ is
time (hr), vs is supply air volumetric flow rate (m3/h), ACH is air infiltration rate (h-1), m
s is
indoor moisture generation rate (kg/h), subscripts “r”, “o”, “s” refer to “room”, “outside”, and
“supply” respectively. The last term m ad/de is moisture adsorption and absorption by room
surfaces and furniture (kg/h). In general, the moisture adsorption/absorption properties of
building materials are less understood, therefore, overall, precision of moisture prediction lags
behind that for thermal prediction. The indoor surface and furniture adsorption and desorption
effects are not included. In reality, such effects may help smooth the RH fluctuations, but will
not significantly affect the hourly averages. In simulations, thermal analysis is coupled with
moisture analysis to take into account the interactions between thermal and moisture
performance.
280 Li-Zhi Zhang
At this point, the governing equations will be developed for each of the component in the
above proposed PCDC cycle combined with chilled-ceiling panels.
The Total Heat Exchanger

The total heat exchanger could be either an energy wheel, or a membrane exchanger
previously described. Two effectiveness: sensible effectiveness (εS) and latent effectiveness
(εL) are defined. Air state at point 2 is calculated by
T2 = T1 − ε S (T1 − T6 ) (11.3)
ω 2 = ω1 − ε L (ω1 − ω 6 ) (11.4)
A Sensible effectiveness of 0.8 is easily obtained with a membrane system, and the latent
effectiveness is about 0.65.
The Rotary Dehumidifier

The purpose of the rotary dehumidifier is to dehumidify the supply air stream as it passes
from state 2 to 3. In doing so, the temperature of the supply air is raised. The humidity ratio of
the return air is increased and the temperature decreased as the desiccant is regenerated in
process 10-11.
The models for the desiccant wheel can be classified into two categories: finite difference
models, and correlation models. The former models are described in detail in Chapter 3, the
latter one is given in Ref [12]. Detailed finite difference modeling of desiccant wheels were
conducted in Chapter 3 for energy wheels. The only difference is that fresh air is replaced by
regenerating air. Exhaust air is replaced by supply air. In this chapter, a simpler model
provided in [12] is used. The finite difference models are detailed, but complex and difficult
to find stable solutions. The correlations, which are used in this study, are simple, yet
sufficient for an energy analysis. The correlations are summarized as follows:
f1so − f1si
η f1 = (11.5)
f1ei − f1si
f 2so − f 2si
η f2 = (11.6)
f 2ei − f 2si
f1j = −2865T j-1.49 + 4.344w 0.8624

j (11.7)
f 2j = T j1.49 / 6360 − 1.127 w 0.07969

j (11.8)
where subscripts “s” and “e” mean “supply” and “exhaust” respectively, and “i” and “o”
mean “inlet” and “outlet” respectively. Effectiveness ηf1 and ηf2 could be pre-determined by
the finite difference equations, considering wheel structure and dimensions.
The desiccant dehumidifier configurations are: wheel weight, 2.5kg; duct wall thickness,
0.2mm; duct geometry, sinusoidal; material, silica gel; effective material fraction, 0.7; wheel
length, 0.2m, rotary speed, 30RPH. The coefficients are: ηf1=0.29; ηf2=0.85.
The Rotary Regenerator

The purpose of the regenerative heat exchanger is to transfer the heat of sorption present
in the supply air stream after dehumidification to the return air stream. As can be seen from
Figure 11.4b, in sensibly cooling the supply air stream from state 3 to 4, the return air stream
is preheated from state 8 to 9. The more effective this component the less energy required
from the heat source to regenerate the desiccant.
The effective method can be applied to a compact heat exchanger of this type in the
conventional manner. By assuming equal heat capacities for the two air flows, temperature at
point 4 is calculated by:
T4 = T3 − ε R (T3 − T8 ) (11.9)
Based on information on the construction details of the regenerator, effectiveness values

can be estimated from previously developed models. A constant effectiveness of 0.85 is
selected for this study.
The Evaporative Cooler

In the return air side and after the total heat exchanger, a rigid media air cooler is used to
cool the air stream. This particular type of evaporative cooler consists of rigid, corrugated
packing material which form the wetted surface. Moist air flows through the corrugations.
Water enters the top of evaporative cooler and flows by gravity through the wetted surfaces.
Commercially available evaporative coolers are rated according to their saturation
effectiveness, εC, defined as
ε C = (Ti − To ) / (Ti − Tiwb )(100% ) (11.10)
where Ti is the entering air dry bulb temperature, Tiwb is the entering air wet bulb temperature,
and To is the exiting air dry bulb temperature.
Therefore, exit air state at point 8 is
T8 = T7 − ε C (T7 − T7wb ) (11.11)
Process 7-8 is a constant enthalpy process. Moisture content at point 8 can be calculated
from psychrometrics. Saturation effectiveness εC in the range of 0.7-0.9 is attainable.
282 Li-Zhi Zhang
The Heating/cooling Coil

The air cooling coil is a conventional cross flow, water-to-air heat exchanger. Its
performance is described by the effectiveness relationship:
εHC=[ΔT (minimum fluid)]/[ΔT (inlets)] (11.12)
where the minimum fluid is the one with the minimum value for heat capacity. The moisture
content is constant through the heating/cooling processes. An effectiveness of 0.85 is assumed
for the cooling coil.
Primary Equipments
The required electric power input for a chiller can be calculated by
Chiller power=Qc/COP (11.13)
where Qc is cooling energy (W) and COP is the coefficient of performance of the water
chiller. For a specific system, the COP is the function of the required chilled water
temperature, the ambient temperature and humidity, as well as operation capacity.
For wheel regeneration, a boiler is needed, energy input
Boiler power=Qh/ηb (11.14)
where Qh is heating energy required (W) and ηb is the boiler efficiency.

Fan and pump energy is an important factor in the annual energy consumption of an
HVAC system. Fan (pump) performance can be characterized by its efficiency, which itself is
dependent on operational air flow rate. Mostly, rated volumetric flow rate, pressure rise and
efficiency are available from the manufacturer. Then rated power can be calculated as
Fan (pump) power=VaΔp/(3600ηf) (W) (11.15)
where Va is air (water) volumetric flow rate (m3/h), Δp is total pressure rise (Pa), ηf is fan
(pump) efficiency.
Effectiveness of the main equipments relates to design and operating conditions. When
the operating conditions fluctuate near design conditions, the effectiveness changes in a small
range. To simplify analysis, constant effectiveness values for various equipments are
assumed.
Building Configurations
To compare the energy consumption and indoor humidity behaviors of the different air-
conditioning systems, a typical office room in a south-facing high-rise building in Guangzhou
city, Canton, is considered. The office is 5.1m long, 3.6m wide, and 2.6m high. The thickness
of the envelope is 260mm and the resulting heat transmission coefficient U for the opaque
part of the facade is 2.91W/m2K. The glazing area is 2.88m2 with double-pane windows of
which center-of-glass U value is 1.31 W/m2K. Louver curtain is installed behind windows. It
is determined that about 70% of the ceiling is covered by radiant ceiling panels when the
radiant ceiling panel system is applied. Indoor set points are: 25°C operative temperature,
50% RH in summer and 23°C OT, 50% RH in winter.
An occupancy pattern of 2 persons with a schedule from 9 to 18h is simulated in the
office. When present, each person generates 75W sensible heat and 57.6g/h moisture. Of the
sensible heat generated, 50W is radiative and 25W is convective. An air infiltration rate of 0.2
ACH are modeled during the time when the ventilation system was switched off and the
building is not pressurized.
Besides, constant loads of 459W of equipment and lighting with a schedule from 9 to 18h
are modeled in the room. Half of these loads are considered as radiative and half convective.
The operating hours of the ceiling panels and ventilation systems are also from 9 to 18h.
However, in summer, to prevent water condensation on ceiling panel surfaces, the desiccant
system is operated one-hour in advance of ceiling panels.
The supplied chilled water flow rate to the ceiling panels is 0.5t/h, resulting a temperature
rise of about 1.6°C through the panels. The processed fresh air is supplied at a rate of 67m3/h,
which complies with ASHRAE standard 62-2001 [2]. The desiccant dehumidifier
configurations are: wheel weight, 2.5kg; duct wall thickness, 0.2mm; duct geometry,
sinusoidal; material, silica gel; effective material fraction, 0.7; wheel length, 0.2m, rotary
speed, 0.5rpm.
Operating Parameters in the Cycle
Detailed computer models of the desiccant cooling system are developed and combined
with the building energy simulation code ACCURACY [10]. Weather data in Canton is used
to evaluate the system performance by hour-by-hour calculations. The hourly operating states
in the DC cycle and PCDC cycle are calculated. Table 11.1 and 11.2 show the operating
states of the two systems at 14:00 of July 15 respectively. The total sensible load of the room
in this hour is 64.6W/m2, of which, 86% is accounted for by the ceiling panels and 14% is
accounted for by the desiccant cooling. The total latent load is 5.1W/m2, which is treated by
the desiccant cooling system solely.
Table 11.1. Operating point for the DC cycle at 14:00, July 15
State Point Temperature (°C) Humidity ratio (kg/kg)

1 32.8 0.0201
2 77 0.0085
3 28 0.0085
4 18 0.0085
5 25.7 0.0103
6 19.4 0.0128
7 68.4 0.0128
8 92.9 0.0128
9 48.7 0.0244
284 Li-Zhi Zhang
Table 11.2. Operating point for the PCDC cycle at 14:00, July 15
State Point Temperature (°C) Humidity ratio (kg/kg)

1 32.8 0.0201
2 26.8 0.0126
3 57.7 0.0085
4 30.8 0.0085
5 18 0.0085
6 25.7 0.0103
7 31.7 0.0178
8 26.1 0.0201
9 52.9 0.0201
10 77.3 0.0201
11 46.9 0.0241
110
DCMEC
PMEC
PCDC
90
T h (?(°C)
)
70
50
30
1 2 3 4 5 6 7 8 9 10 11 12
Months
Figure 11.5. Monthly averaged regenerating temperatures for the DC and PCDC cycles.
From these two tables, it is seen that the heat absorbed in the gas heater for wheel
regeneration are the same for the PCDC and DC cycles, however, the regenerating
temperature in the former cycle is about 15.6°C lower than that in the latter one. Therefore,
lower grade heat can be used in a PCDC cycle. This indicates that pre-cooling improves the
performance by increasing the dehumidification efficiency.
Figure 11.5 shows the monthly-averaged regenerating temperatures in a year for the DC
and PCDC cycles. August is the most humid month, therefore, highest regenerating
temperatures are required in this month. Similarly, it is driest in February, and the
regenerating temperatures are the lowest at this time. In winter, the two cycles have the
similar regenerating temperatures, however, in summer when it’s very humid, the PCDC
cycle has much lower regenerating temperatures. In other words, the more humid it is, the
more superior an PCDC in comparison with a DC cycle. The distributions of annual operating
hours with various regenerating temperatures are shown in Figure 11.6. In a DC cycle, only
70% of annual operating hours are with regenerating temperatures below 80°C; while in a
PCDC cycle, nearly 99% of operating hours are with less than 80°C regenerating
temperatures. Lower heat source usually mean higher efficiency. This is very interesting,
indicating that renewable energy such as solar energy can be more efficiently employed in a
pre-cooling desiccant system.
100~120°C
0.3% <40°C
15.3%
80~100°C
30.2%
40~60°C
29.5%
60~80°C
24.7%
(a) DC cycle.
80~100°C
1.2% <40°C
12.4%
60~80°C
46.5%
40~60°C
39.9%
(b) PCDC cycle.
Figure 11.6. Annual percentage of operating hours with regenerating temperatures, (a) DC cycle and (b)
PCDC cycle.
286 Li-Zhi Zhang
Annual Primary Energy Consumptions
The energy required for the primary equipments is calculated. The performances of the
primary equipments are modeled with certain constant indices [8]: the COP of the chiller is
4.39 when evaporating at 15°C and 3.31 when evaporating at 5°C; the boiler efficiency is
0.75 and fan (pump) efficiency 0.60; the fan pressure rise is 1400Pa for all-air, and 1600Pa
for desiccant cooling; pump pressure rise is 0.3bar for all-air and 0.5bar for ceiling panels. In
the analysis, electricity consumptions are converted into the equivalent primary energy by
multiplying a factor of 3. The energy used by motors driving the regenerative wheels is
categorized as auxiliary energy.
The annual primary energy consumptions per floor area are illustrated in Figure 11.7.
Compared to a conventional all-air system, chilled-ceiling combined with precooling
desiccant cooling (PCDC+CC) saves 71.6% of fan energy, 50% of chiller energy, while
consumes 1.1 times of pump energy, and additional 90.4kWh/m2 of P.E. due to desiccant
regeneration and wheel driving, resulting in a total energy saving of 30%. In an all-air system,
air is supplied at 14°C, while in a desiccant system combined with chilled-ceiling system, air
supply temperature is 4°C higher. This increased supply air temperature helps to eliminate
thermal discomforts like cold drafts. In the combined system, air is mainly for ventilation and
dehumidification purposes and its volume is only 23% of that in an all-air system. Fan
energy, which is usually in the same magnitude with cooling energy, is thus drastically
reduced. As a result, large amount of energy is saved.
400
Annual P.E. consumption (kWh/m2
Auxiliary
Boiler
300 Fan
Pump
200 Chiller
100
0
All-air PCDC+CC DC+CC
Figure 11.7. Annual primary energy consumptions of a PCDC desiccant cooling combined with chilled-
ceiling (PCDC+CC), and a DC+CC cycle, in comparison with a conventional all-air system.
The electricity consumptions for the three systems are 114.7kwh/m2, 76.1kwh/m2, and
67.6kwh/m2, for conventional all-air system, PCDC+CC, and DC+CC, respectively. A
PCDC+CC system saves 33.7% of electricity compared to a conventional all-air system. This
would be of particular significance since “peak-saving” is a hot issue these days and the
cooling load usually is the highest during peak hours. A PCDC+CC system uses more low-
grade heat, less electricity.
It could also be concluded from this figure that a PCDC+CC cycle further saves about
10% primary energy, compared to a DC+CC cycle. The largest fraction of energy saving
comes from reduced boiler energy. Therefore, a PCDC+CC cycle is more efficient than a
DC+CC system.
Indoor Humidity
The distribution of indoor relative humidity during working hours are shown in Figure
11.8 for CAV (constant volume all-air) and PCDC+CC. The optimum zone is from 40% to
60%RH and 50% RH is ideal for building occupants to avoid the hazards of fungi, bacteria,
viruses, and respiratory difficulties. The figure shows that the all-air system has less annual
hours in the comfort region, and has more hours in either the dry (<40%) or the humid (>60%)
regions. The chilled-ceiling combined with pre-cooling desiccant cooling controls the indoor
humidity well: 90% of annual operating hours is in the optimum region.
For conventional constant volume all air system, the minimum RH is in July and the RHs
are well controlled to around 50% in summer season. In winter and transient seasons, the RHs
rise up. RHs greater than 70% occasionally occur in these seasons. This phenomenon is due
to the fact that in a conventional all air system, temperature is intentionally controlled, while
the humidity is passively controlled, because the supply air state is determined by the sensible
load. In winter and transient seasons, sensible load becomes smaller and air is not sufficiently
dehumidified. In contrast, PCDC +CC controls indoor temperature and humidity
independently.
Therefore, it can be concluded that chilled-ceiling with desiccant cooling has better
indoor humidity controls with reduced energy use than its all-air counterpart does.
Even though PCDC+chilled-ceiling system realizes independent humidity control with
reduced energy use, when used in hot and humid regions, condensation on ceiling panels
remains a troublesome issue. To illustrate this problem, the variations of room temperature
and RH on a typical summer day are shown in Figure 11.9 for a system combining chilled-
ceiling with AHU. In this situation, the air infiltration rate at night ACH is set to 0.2. The
operation schedule is that both the AHU and the ceiling panels are operated from 9:00 to
18:00 working hours. The maximum RH occurs on the coldest location of ceiling panels. It is
indicated that both the room temperature and room mean humidity drop to the set points in
the hour when the systems are started, and stay near set points afterwards. When the systems
are closed after 18:00, both the temperature and humidity rise steadily due to air infiltration,
to above 27°C and 70%RH. At 9:00, because the inlet cooling water to ceiling panels is the
coldest and the indoor moisture level is the highest after one night’s infiltration, maximum
local RH above 100% results. In other words, condensation occurs at this moment.
288 Li-Zhi Zhang
4000
All air
Annual operating hours (Hrs 3000 PCDC+CC
2000
1000
0
<40% 40~60% >60%
RH
Figure 11.8. Indoor humidity distributions for two systems: all air and PCDC+CC.
100 30
90
Room Temperature (? )
80
70
RH (%)
60 25
50
40 Average RH
30 Max RH
Room Temperature
20 20
0 5 10 15 20 25
Time (Hrs)
Figure 11.9. Variations of room temperature and RH on a typical summer day with chilled-ceiling and
PCDC desiccant cooling both operated from 9:00 to 18:00.
100
90
80
70
RH (%)
60
50
40 ACH=0.1
ACH=0.2
30
ACH=0.4
20
0 5 10 15 20 25
Time (Hrs)
Figure 11.10. Variations of Maximum RH with 1 hour in advance dehumidification.
To prevent condensation, room air needs to be dehumidified prior to the operation of

ceiling panels. A one-hour in advance dehumidification/ventilation strategy is simulated in
which ceiling panels are operated from 9:00-18:00 and the desiccant system is operated from
8:00-18:00 [11]. The variations of the maximum RH with different night air infiltration rates
are plotted in Figure 11.10 under this strategy. It is clearly shown that room moisture levels
are substantially pulled down during the first dehumidification hour, and when the ceiling
panels begin to operate, the maximum RHs are below 90%. Therefore, water condensation is
prevented. The higher the ACH is, the larger the room humidity is at night. The influences of
ACH on indoor humidity in the working hours are negligible if a one-hour in advance
dehumidification is implemented. On the other hand, if a building is well enclosed and the air
infiltration rate is less than 0.05 at night, no condensation occurs even under simultaneous
dehumidification and ceiling cooling. In a summary, condensation is the most important issue
in ceiling cooling. Unless properly controlled regarding dehumidification strategy,
condensation remains a problem.
Capital Cost
An exact capital cost difference between the combined PCDC+CC system and the
equivalent all-air system is difficult to determine. It depends mainly on market prices and
contractors, and varies considerably for different countries. In Canton, it is expected that the
PCDC+CC system and the DC+CC system are 30% and 25% more expensive, respectively,
than a traditional all-air system for cooling load of 60W/m2 (floor area). The increased cost
mainly comes from ceiling panels and desiccant wheels. The estimated payback period for a
PCDC+CC system is 15 years for a 500m2 office building. We believe the capital cost and the
payback period will decrease in future, if the market penetration of the combined system
accelerates.
290 Li-Zhi Zhang
11.3. INDEPENDENT AIR DEHUMIDIFICATION

There are various techniques for air dehumidification [2]. Traditionally, latent load and
sensible load are treated in a coupled way, like the CAV system described before. Because air
is not only for ventilation, but also a heat transfer medium, and a large quantity of air is
needed to extract the sensible load, energy requirements are very high. Another problem with
this technique is that in transition seasons, when it is at part load conditions, humidity control
will be lost.
There is an increasing trend to separate the treatment of sensible and latent load by using
an independent humidity control system. According to this concept, the latent load of a room
is treated by an independent humidity control system, while the sensible load is treated with
some other alternative cooling techniques such as chilled-ceiling panels, or phase-change
materials. Since the circulating air is dramatically reduced, energy consumption can be
reduced substantially. Another benefit is that chilled water or suction temperatures can be
raised, resulting in increased equipment efficiency and decreased operating costs. It is
estimated that with a new system of chilled-ceiling panels combined with independent
humidity control, 30% of energy could be saved in comparison to a traditional coupled
system. Nevertheless, due to the hot and humid climates in south China, energy for moisture
control with an independent humidity control system still accounts for 25-40% of the total
energy for air conditioning. To further reduce the energy consumption in the treatment of
fresh air, energy recovery measures must be combined to an independent humidity control
system.
In this section, four independent humidity control systems with heat recovery measures
are compared. These systems are: System 1, Mechanical dehumidification with a heat pump;
System 2, Mechanical dehumidification with a sensible heat exchanger; System 3,
Mechanical dehumidification with a membrane-based total heat exchanger; and System 4:
Desiccant wheel driven by a heat pump. Through hour-by-hour analysis, the annual primary
energy consumption for the four systems is discussed.
As an example, an office building with 5 occupants in a 20m2 room is considered. The set
points for indoor air are as follows: temperature, 25°C; relative humidity, 50%. Fresh air is
supplied at 37.5m3/hr, 20°C in temperature and 7g/kg in humidity ratio. The air
dehumidification systems are operated when the offices is opened, i.e., from 9:00 to 18:00. At
nights, the systems are shut down, to save energy. The air flow rates are selected according to
ASHRAE Standard 62-2001 [2], which is determined by several factors such as room area,
building occupancy pattern, and building types. The sensible load of the room is treated by
chilled-ceiling panels, as previously did. The total ventilation load includes two fractions:
moisture load from fresh air and moisture load from human activities. The fresh air sensible
and latent load is a variable relating to weather conditions, while the load from human
activities is assumed to be kept at 50g/hr.
System Descriptions
Four independent air dehumidification systems with heat recovery measures are
proposed. The operations of the four systems are controlled to satisfy the load and weather
conditions. The air flow rates are fixed. For mechanical dehumidification, both the
evaporating and condensing temperatures are controlled according to the load and outside
temperatures. For desiccant wheels, the regeneration temperature is adjusted to fit the load.
These systems can be operated as stand-alone independent air dehumidification units.
Compressor
Condenser
Indoor Air
D
Valve
Fresh Air Supply Air
A B C Fan
Evaporator Condenser
(a)
0.03
0.025
Humidity ratio (kg/kg
0.02
A
0.015
0.01 D
0.005 B C
0
5 10 15 20 25 30 35 40
(b)
Figure 11.11. Schematics (a) and Psychrometrics (b) of System 1: Mechanical dehumidification with
heat pump.
System 1: Mechanical Dehumidification with Heat Pump

A schematic and the corresponding psychrometric charts are shown in Figure 11.11.
Fresh air (Point A) first flows through a cooling coil where it is dehumidified below the dew
point (Point B). Then the air flows through a heating coil where it is heated to the set points of
292 Li-Zhi Zhang
supply air. The system comprises a heat pump: the cooling coil acts as the evaporator, and the
heating coil acts as a condenser. The heat that should be rejected is larger than that that is
required to heat the supply air, so an additional condenser is necessary to reject the surplus
heat of the heat pump. Exhaust air from indoor space is pumped through this condenser to
enhance the performance of the heat pump system.
System 2: Mechanical Dehumidification with Sensible Heat Exchange

This concept uses a sensible heat exchanger to recover the heat of the supply air itself
after it flows through the dehumidification cooling coil. As shown in Figure 11.12, fresh air at
point A flows through the sensible heat exchanger, where it is cooled down (in some cases, it
also loses some water content due to condensation) to point B. Subsequently, it is
dehumidified by a cooling coil and returns to the sensible exchanger at point C. After heated
up, it is supplied to the room. The cooling coil can be an evaporator of a small refrigeration
system. Then the incoming fresh air is dehumidified and subsequently heated by a heat pump
system that is similar to System 1.
Without any use of the exhaust air, this design uses a sensible-only air-to-air heat
exchanger to pre-cool and reheat the outdoor air that is dried with a mechanical dehumidifier.
Heat pipes, coil run-around loops and plate type heat exchangers are used for this purpose.
System 3: Mechanical Dehumidification with a Membrane-based Total Heat Exchange

In this system, a membrane based total heat exchanger is used before the fresh air is
pumped to a heat pump for air dehumidification. The total heat exchanger has a membrane
core where the incoming fresh air exchanges moisture and temperature simultaneously with
the exhaust air. In this manner, the total heat or enthalpy from the exhaust is recovered. The
schematic and the processes are shown in Figure 11.13 for this system. This system is also
relatively simple, since the membrane system has no moving parts, and is compact.
System 4: Mechanical Dehumidification with a Desiccant Wheel

This two-stage equipment uses the condenser heat from a mechanical dehumidifier to re-
active a desiccant wheel. First, fresh air is pre-cooled and partially dried by the mechanical
dehumidifier. Then the air is dried more deeply and also reheated by the desiccant wheel.
This arrangement uses both technologies at favorable points of performance. At higher inlet
humidities, the mechanical refrigeration system can operate at a higher coil temperature and
suction pressure, thereby saving energy. Dehumidified air from a desiccant wheel is very hot.
Therefore before it is supplied to rooms, it should be cooled down to set points first. To
recover the energy from exhaust air, an evaporative cooler is used to cool down the exhaust
air, which is then used to cool the supply air from desiccant wheel. Under extremely humid
ambient conditions, energy required for re-activation exceeds available heat from the heat
pump. In such cases, an auxiliary electric heater is used to accomplish the regeneration of the
desiccant wheel. This system is relatively complex due to the rotating character of the wheel
and the reactivation of the desiccants. If low grade waste heat is available, the desiccant
wheel system becomes superior to other systems in energy efficiency. The system is shown in
Figure 11.14. Total heat recovery is realized by evaporative cooling of room exhaust air.
Besides above four systems with partial or full total heat recovery, to have a comparison,
a mechanical dehumidification system with no heat recovery shown in Figure 11.15 is also
considered. It is the base system.
Sensible heat
exchanger
A Fan
Fresh Air C
Supply Air B
Cooling Coil
(a)
0.03
0.025
0.02
B A
0.015
0.01
0.005 C D
0
5 10 15 20 25 30 35 40
(b)
sensible heat exchange.
294 Li-Zhi Zhang
Compressor
Total heat Condenser

exchanger
A Indoor Air
FreshAir E
Valve
SupplyAir
F
B C D Fan
Evaporator Condenser
(a)
0.03
0.025
A
0.02
F
0.015 B
E
0.01
0.005 C D
0
5 10 15 20 25 30 35 40
(b)
membrane-based total heat exchanger.
Compressor
Electric
Heater
Valve
Condenser
I H G E
F Indoor Air
A Supply Air
Fresh Air C D Evaporative cooler
Evaporator B
Desiccant wheel
(a)
0.03
0.025
A
0.02
0.015
0.01
B
C
D
0.005
0
5 10 15 20 25 30 35 40
(b)
Figure 11.14. Schematics (a) and Psychrometrics (b) of System 4: Mechanical dehumidification +
Desiccant wheel.
296 Li-Zhi Zhang
Compressor
Condenser
Valve
Supply Air
Fresh Air Fan

Evaporator Heater
Figure 11.15. The traditional system, mechanical air dehumidification with no recovery.
Component Modeling
Energy performance of the four building dehumidification system were calculated by

thermodynamic calculations. The modeling methods of total heat exchangers, evaporative
coolers, desiccant wheels, sensible only heat exchanger are the same as those described in the
above section. Here gives the modeling for the newly added components.
2’ ° 2
°
4 ° 3
°
°
5 ° ° 1
6
O s
Figure 11.16. T-s diagram of the refrigeration system.
Refrigeration Cycle
A thermodynamic model of the refrigeration system is formulated based on the processes
of refrigerant R134a shown in Figure 11.16. Saturated R134a liquid at point 4 flows through
an expansion valve and becomes wet vapor at point 5. Refrigerant at this state flows to the
evaporator (also the dehumidifier) where it chills and dehumidifies the fresh air and
evaporates to point 6 and further superheats to point 1. Then the refrigerant vapor is pumped
by a compressor to point 2 where the vapor is displaced to the condenser and condensates
from state 2 to 4 through 3. The superheat is set to 5°C. It should be noted that the exact
degree of superheating may be affected by many factors, such as evaporator, expansion valve,
and compressor, and are strongly related to operating conditions, control strategies.
Experiments found superheating are in the range of 2 to 16°C; and the superheating increases
with air temperature in the evaporator, from 3°C to 14°C when the expansion valve is fully
open. From viewpoint of energy use, too large superheating is not good. Generally, 3-14°C
superheating are possible. Refrigerant superheating refers to the superheating in the
evaporator. Superheating from tubes can be neglected with well tube-insulating, since
superheating from such sources is harmful to energy performance and should be prevented.
The effects of heating in the compressor are included in the compressor’s isentropic
efficiency.
The specific enthalpy of the refrigerant at the compressor exit is calculated by
hf2' − hf1
hf2 = hf1 + (11.16)
ηs
where ηs is the isentropic efficiency; h2’ is the specific enthalpy at the condensing pressure by
isentropic compression from the evaporating pressure. Experimental results have shown that
the isentropic efficiency is a weak function of the displacement volume, and varies linearly
with the compressor speed. In this analysis, a constant isentropic efficiency of 0.75 is
assumed, neglecting the rotational speed of the compressor
The specific enthalpy after the expansion valve is calculated by
hf5 = hf4 (11.17)
The degree of dryness at the inlet of the evaporator
h5 − h5'
x= (11.18)
h6 − h5'
where h5’ is the specific enthalpy of the saturated liquid refrigerant at the evaporating
pressure.
The electricity consumed by the compressor
q com = m f (hf2 − hf1 ) / η motar (11.19)
f is the mass flow rate of refrigerant (kg/s); ηmotor is the motor efficiency, which is
where m
considered as 0.75.
298 Li-Zhi Zhang
The energy load of the evaporator
q e = m f (hf1 − hf5 ) (11.20)
Heat rejected in the condenser
q c = m f (hf2 − hf4 ) (11.21)
Refrigeration efficiency
qe
COP = (11.22)
q com
Evaporator and Condenser

Cooling and dehumidification of the incoming fresh air are performed in the evaporator.
A detailed modeling of the evaporator and the condenser are rather complicated [15,16].
Usually, they are divided into regions associated to the phase of the refrigerant. Each region
constitutes a separate heat exchanger. In the case of the condenser, the superheated vapor, the
condensation and subcooled liquid regions are considered, whereas for the evaporator it is
divided into the evaporating and superheated vapor regions. For each region, the refrigerant
side and air side convective heat transfer coefficients need be calculated from the established
correlations for single-phase and two-phase flow. For the evaporator, when the average fin
surface temperature is calculated to be less than the local water dew point of the air stream,
moisture condensation will occur. Under these conditions, the air heat transfer coefficients
can no longer be calculated as in dry conditions, and a water mass balance must be carried
out. In this case, the enthalpy method, proposed by Threlkeld [17] and introduced in
ASHRAE Handbook [1] was most adequate for use. According to this procedure, the driving
force for heat transfer is assumed to be the difference between the saturated enthalpy of the
air flowing over the fins and a fictitious saturated air enthalpy evaluated at the refrigerant
temperature.
The analysis of air cooling and dehumidifying coils requires coupled, non-linear heat
mass transfer relationships. While the complex heat mass transfer theory that is presented in
many textbooks is often required for cooling coil design, simpler models based on
effectiveness concepts are usually more appropriate for energy estimation. These techniques
are resulted from basic heat and moisture transfer equations for simultaneous heat and
moisture transport. Therefore, in this study, to ease the analysis, thermal performance of the
heat exchangers regions is evaluated by the (ε, NTU) method. According to this procedure,
the heat exchanger effectiveness is defined as
Qactual
ε= (11.23)
(m c p ) min (Thi − Tci )
where Thi and Tci are the inlet temperatures of the hot and cold fluids, respectively (K).
Because the operations are set to fluctuate around the design points, constant evaporator and
condenser effectiveness are assumed in the simulations. Then the air states at the outlets of
evaporator and condenser could be obtained.
The heat extracted in the evaporator is calculated by
q e = m a (haB − haC ) (11.24)
a is the mass flow rate of fresh air stream (kg/s); haB and haC are the specific
where m
enthalpies of air at point B and C respectively.
Similarly, heat rejected at this portion of the condenser is governed by
q c1 = m a (haC − haD ) (11.25)
It should be noted that qc1 calculated from Eq.(11.25) is only a portion of that calculated
by Eq.(11.21). The evaporating temperature is fixed to 5°C, while the refrigerant flow rate
and the condensing temperature varies according to the cooling load of the evaporator and the
outside weather conditions. Usually, a 10°C log mean temperature difference between the
condensing refrigerant and the air flowing through it is required.
Heat Pump
The cooling coil acts as an evaporator of a heat pump, and the heating coil acts as a
condenser for the heat pump. The efficiencies vary with evaporating and condensing
temperatures. The heat pump efficiency is defined as
q Con
ε HP = (11.26)
q Ele2
where qCon is the heat rejected at the condenser side (kW), and qEle2 is electricity consumed by
the compressor (kW). The above equation is used to calculate the electric energy to drive the
heat pump, from the condensing energy required and heat pump efficiencies. Depending on
the operating and condensing temperatures, the heat pump efficiencies are in the range of 3-5.
Effectiveness of the main components are related to design and operating conditions.
When the operating conditions fluctuate near design conditions, the effectiveness changes
only in a small range. To simplify analysis, constant effectiveness for various components is
assumed.
The simulations are conducted on an hour-by-hour basis. The operating hours are from
9:00 to 18:00. Fan efficiency is selected as 0.6. For the convenience of comparison, energy
consumed in the form of electricity is converted to primary energy by a factor of 3.3.
The dehumidified supply air temperature is set to and fixed at 20°C. The indoor is 25 °C.
So the dehumidified air has no sensible load. Rather, it will extract a small fraction of the
sensible load from the building.
300 Li-Zhi Zhang
Performance Analysis
The COP varies with both the evaporating temperature and the condensing temperature.
The influence of the condensing temperatures on the COP is shown in Figure 11.17. As can
be seen, the COP decreases with increasing condensing temperatures. When the condensing
temperature increases from 20°C to 50°C, the system COP decreases from around 8.0 to 2.5.
Following correlation could be formulated for the relation for the refrigeration evaporating at
5°C:
COP = 20.041e −0.0417 tc (11.27)
The effects of the evaporating temperature on the COP are shown in Figure 11.18. The
system COP rises with increasing evaporating temperatures. In fact, when the evaporating
temperature increases from -10°C to 25°C, COP is improved from 2.0 to 6.5. A correlation
has been formulated for the analysis of the system performance:
COP = 0.0001t e3 + 0.0016t e2 + 0.0746t e + 2.6377 (11.28)
where the condensing temperature is fixed as 45°C.

Figure 11.19 shows the distribution of the COP of the refrigeration system during a year.
As indicated, in winter, the system has higher performance, in contrast, when it’s hot in
summer, the system COP decreases, which will in return deteriorate the energy requirements.
12
10
8 ♦
♦
COP
0
10 20 30 40 50 60
(C)
TTc c(°C)
Figure 11.17. Variations of COP versus condensing temperature, when Te=4.5°C.

7 ♦
♦
6
5
COP
0
-10 0 10 20 30
T e (C)
T (°C)
e
Figure 11.18. Variations of COP versus evaporating temperature, when Tc=46°C.
10
7
COP
2
0 2 4 6 8 10 12
Months
Figure 11.19. Distribution of COP of the refrigeration cycle in a year, by hourly calculations.
Annual Primary Energy Requirements
The four systems are used to treat ventilation fresh air. To make comparisons, the annual
energy requirements of the four proposed systems and a dehumidification system with no heat
302 Li-Zhi Zhang
recovery measures are computed and listed in Table 11.3. As can be seen, the total energy of
the systems with heat recovery saved 30% to 43%, depending on the systems involved. In the
table, the values in the column “cooling” refers the energy required in the cooling coil for air
dehumidification; “Heating” refers to the energy need to heat the dehumidified air to the set
points; “Electricity” refers to the converted primary energy of electricity used by
compressors; “Condensing” refers to the energy rejected by the condenser of the heat pump;
“Auxiliary” means the converted primary energy used in electric heating, for example, the
electric heater in System 4. The column “Fan” refers to the converted primary energy used to
circulate the air. All the energy values are calculated on a per-person basis.
In the analysis, the systems are used to treat the latent load solely, the sensible load of the
room is around 50W/m2, which will be extracted by chilled-ceiling panels.
The comparisons of the different systems are plotted in Figure 11.20. It is shown that
among the 5 systems, System 3 consumes the least energy, and System 2 consumes the most.
Generally speaking, energy savings for the four systems are in the same order. This is due to
the reason that all the 4 systems newly proposed take into account the energy recovery
measures of the exhaust air.
Of the systems studied, three systems recover, more or less, the energy from exhaust air.
They are the same use as an economizer. Outside air is fresh air that needs to be dehumidified
and treated. In system 1, the indoor air is used to cool the condenser. In system 3, indoor air is
used to cool and dehumidify the fresh air in a total heat exchanger. In System 4, indoor air is
used to cool down the dehumidified air. System 3 is the best.
Table 11.3. Annual primary energy requirements (kJ/person) for each person under
various dehumidification strategies
Cooling Heating Electricity Condensing Auxiliary Fan Total
No 5.41E+06 2.11E+06 3.61E+06 6.52E+06 9.11E+04 2.63E+05 6.22E+06

recovery
System 5.41E+06 1.62E+06 3.62E+06 6.61E+06 3.82E+04 3.92E+05 4.13E+06

1
System 4.13E+06 1.64E+06 3.72E+06 5.32E+06 3.33E+04 5.93E+05 4.54E+06

2
System 4.32E+06 5.83E+05 3.01E+06 5.34E+06 4.54E+04 5.92E+05 3.64E+06

3
System 3.23E+06 2.33E+06 2.32E+06 4.16E+06 9.72E+05 8.81E+05 4.18E+06

4
7. 0E+06
Annual P. E. r equi r ement s ( kJ/ per son)

6. 0E+06
5. 0E+06
4. 0E+06
3. 0E+06
2. 0E+06
1. 0E+06
0. 0E+00
No Syst em 1 Syst em 2 Syst em 3 Syst em 4
r ecover y
St r at egi es
Figure 11.20. Annual primary energy consumptions by air dehumidification for each person with four
systems proposed.
11.4. CONCLUSION
The novel pre-cooling MEC desiccant cooling cycle in combination with chilled-ceiling
panels is a new generation of HVAC system. In this system, sensible load is treated by the
cooling panels and the latent load is treated by the desiccant system. The results found that
compared to a conventional all-air system, the proposed system saves much fan energy due to
reduced air volume, and saves much chiller energy due to raised evaporating temperatures.
The total primary energy savings amount to 30% for DC+CC system and 40% for PCDC+CC
system, respectively. In addition, in the combined system, the temperature and the indoor
humidity are decoupled and intentionally controlled independently. As a result, more annual
hours are in the comfort region.
Pre-cooling improves wheel’s dehumidification efficiency, therefore, lower regenerating
temperatures can be employed with a PCDC cycle. The more humid it is, the more superior a
PCDC in comparison with a DC cycle. With a PCDC cycle, nearly 99% of annual operating
hours are with less than 80°C regenerating temperatures. In contrast, a common DC cycle
needs 30% annual hours’ heat of higher than 80°C.
These discussions prove that the proposed PCDC system is a more efficient, energy
saving system, which could be used in hot and humid regions. With this new system, the
regenerating temperature could be 15°C lower than the common DC cycle.
Concerns on indoor air quality have prompted the research of novel air dehumidification
techniques. The systems of mechanical dehumidification are combined with energy recovery
measures like a heat pump, membrane enthalpy recovery, sensible heat exchanger and
304 Li-Zhi Zhang
desiccant wheel. An hour-by-hour simulation reveals that the independent air

dehumidification with heat recovery could save 29-42% of primary energy, depending on the
system involved. Of the systems proposed, the mechanical dehumidification with a sensible
heat exchanger consumes the largest energy, since only a small fraction of total heat is
recovered. In contrast, the mechanical dehumidification with a membrane total heat
exchanger consumes the least, because a full total heat recovery is realized. Because all the
four systems use recovery measures, their energy consumptions are in the same order. The
annual total primary energy used for independent air dehumidification is around 4.18E+06 kJ
per person.
REFERENCES
[1] ASHRAE. 2005 ASHRAE Handbook - Fundamental. Atlanta (GA): American Society
of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (ASHRAE), 2005.
[2] ASHRAE. ANSI/ASHRAE Standard 62-2001, Ventilation for acceptable indoor air
quality. Atlanta (GA): American Society of Heating, Refrigerating and Air-
Conditioning Engineers, Inc. (ASHRAE), 2001.
[3] Chua, K.J.; Ho, J.C.; Chou, S.K. A comparative study of different control strategies for
indoor air humidity. Energy and Buildings, 2007, 39, 537-545.
[4] Waugaman, D.G.; Kini, A.; Kettleborough, C.F. A review of desiccant cooling
systems. ASME Journal of Energy Resources Technology, 1993, 115, 1-8.
[5] Collier, R.K.; Barlow, R.S.; Arnold, F.H. An overview of open-cycle desiccant-cooling
systems and materials. ASME Journal of Solar Energy Engineering, 1982, 104, 28-34.
1347-1367.
[7] Warren, M.L.; Wahlig, M. Analysis and comparison of active solar desiccant and
absorption cooling systems: part I-model description. ASME Journal of Solar Energy
Engineering, 1991, 113, 25-30.
[8] Niu, J.L.; Zhang, L.Z.; Zuo, H.G. Energy savings potential of chilled-ceiling combined
with desiccant cooling in hot and humid climates. Energy and Buildings, 2002, 34,
487-495.
[9] Zhang, L.Z.; Niu, J.L. A pre-cooling Munters Environmental Control cooling cycle in
combination with chilled-ceiling panels. Energy, 2003, 28, 3, 275-292.
[10] Niu, J.L.; Kooi vd, J.; Ree, H.vd. Energy saving possibilities with cooled-ceiling
systems. Energy and Buildings, 1995, 23, 147-158.
[11] Zhang, L.Z.; Niu, J.L. Indoor humidity behaviors associated with decoupled cooling in
hot and humid climates. Building and Environment, 2003, 38, 99-107.
[12] Jurinak JJ, Mitchell JW, Beckman WA. Open cycle desiccant air conditioning as an
alternative to vapor compression cooling in residential applications. ASME Journal of
Solar Energy Engineering 1984; 106(3): 252-260.
[13] Zhang, L.Z.; Zhu, D.S.; Deng, X.H.; Hua, B. Thermodynamic modeling of a novel air
dehumidification system. Energy and Buildings, 2005, 37, 279-286.
[14] Zhang, L.Z. Energy performance of independent air dehumidification systems with
energy recovery measures. Energy, 2006, 31, 1228-1242.
[15] Bensafi, A.; Borg, S.; Parent, D.; Cyrano: a computational model for the detailed
design of plate-fin-and-tube heat exchangers using pure and mixed refrigerants.
International Journal of Refrigeration, 1997, 20, 218-228.
[16] Koury, R.N.N.; Machado, L.; Ismail, K.A.R. Numerical simulation of a variable speed
refrigeration system. International Journal of Refrigeration, 2001, 24, 192-200.
[17] Threlkeld, J.L. Thermal Environmental Engineering, 2nd Edn. Upper Saddle River,
New Jersey: Prentice-Hall, Inc.; 1970.
INDEX
attention, vii, 4, 5, 9, 76, 270, 272
A averaging, 245
absorbents, 226
absorption, 16, 279, 304 B
access, viii
accuracy, 20, 33, 44, 86, 89, 113, 179, 250 bacteria, 275, 287
acetate, 43, 45, 199 bacterium, 4, 272
acetic acid, 190 barrier, 219
acid, 190, 191 behavior, 226, 239
activation, 292 Beijing, 12
additives, 194, 197 bending, 90, 91, 92, 94, 96
adhesion, 199 benefits, 2, 50
adiabatic, 32, 58, 120 Bim, 64
adsorption, 21, 26, 56, 58, 60, 61, 66, 67, 71, 72, bird flu, vii, 4, 272
101, 134, 157, 163, 197, 198, 279 Boltzmann constant, 187, 208
adsorption isotherms, 21, 26 boundary conditions, 20, 32, 76, 96, 105, 109, 113,
advection-diffusion, 233 120, 126, 157, 161, 162, 163, 165, 172, 173, 186,
agent, 190 244, 247, 253, 259, 260, 262, 264
aid, 46 boundary surface, 164
air pollution, 222 breathing, vii
air quality, 1, 4, 7, 9, 269, 272, 303, 304 buildings, 1, 2, 4, 9, 13, 271, 272
alcohol, 201
algorithm, 33, 236, 250, 259, 260
alternative, vii, 3, 270, 290, 304 C
aluminium, 109
aluminum, 83, 274 capacity, 131, 273, 282
ambient air, 18, 192, 212 capital, 289
amide, 189 capital cost, 289
amines, 199 carbon, 44, 109, 120
anemometers, 181 cast, 190
application, viii, 5, 6, 73, 97, 103, 226, 272 casting, 190, 191, 194
aqueous solution, 226 catalyst, 190
Asia, 54 catheter, 77, 96, 185
aspect ratio, 89, 90, 91, 92, 94, 114, 117, 118, 119, cavities, 201
120, 127, 150, 151, 172, 176, 237 cell, 15, 16, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36,
assumptions, 43, 46, 133, 135, 157 37, 38, 39, 40, 41, 43, 44, 46, 47, 48, 49, 50, 51,
asymptotic, 94 52, 53, 54, 157, 188, 193, 202, 203, 204, 205,
asymptotically, 37, 41, 81, 89, 161 215, 222, 232, 249, 258, 260, 263
atmospheric pressure, 62, 137, 208, 274
308 Index
cellulose, 43, 45, 46, 53, 166, 189, 190, 199, 200, conduction, 58, 73, 103, 107, 124, 134, 148, 176,
226 279
cellulose triacetate (CTA), 190, 191, 192, 195, 196, conductive, 83, 99, 103, 109, 120, 203
197, 226 conductivity, 46, 55, 60, 100, 105, 109, 120, 122,
cement, 274 132, 149, 156, 163, 221, 233
CFD, 28, 231, 236, 249, 258, 259, 266 configuration, 128
CH4, 199 confinement, 40
channels, 57, 58, 59, 60, 65, 69, 73, 75, 78, 97, 99, conservation, vii, ix, 2, 6, 12, 13, 47, 59, 60, 79, 89,
101, 120, 131, 134, 149, 150, 157, 173, 174, 176, 105, 123, 158, 203, 232, 233, 244, 264
181, 185, 192, 265 constant load, 283
chemical, 140, 185, 190, 197, 226, 231 construction, 78, 274, 281
chemical industry, 140 consumption, 1, 271, 290
China, vii, ix, 2, 9, 10, 271, 290 contaminants, 276
chloride, 226, 227, 279 continuing, ix
chlorophenol, 226 continuity, 232, 244
circulation, 17, 18, 237 contractors, 289
classical, 176 control, viii, 1, 4, 5, 6, 17, 18, 31, 46, 54, 65, 81, 82,
classified, 6, 280 86, 89, 97, 107, 124, 128, 161, 164, 178, 236,
clay, 109 250, 260, 272, 273, 275, 287, 290, 297, 304
cleaning, 201 controlled, 4, 17, 34, 35, 44, 46, 52, 84, 111, 180,
closure, 245 181, 190, 203, 272, 275, 287, 289, 290, 303
CO2, 187, 195, 199, 222, 223, 224 convection, 27, 46, 73, 81, 96, 97, 103, 128, 134,
codes, 266 185, 186, 267
coefficient of performance (COP), 269, 274, 282, convective, 15, 27, 28, 31, 33, 34, 37, 41, 43, 48, 50,
286, 300, 301 59, 60, 75, 76, 77, 81, 103, 107, 113, 120, 122,
coil, 2, 6, 13, 18, 166, 192, 273, 275, 277, 282, 291, 123, 127, 128, 132, 134, 135, 137, 138, 140, 144,
292, 298, 299, 302 146, 148, 155, 157, 161, 164, 173, 175, 176, 187,
collisions, 209 194, 204, 205, 216, 223, 229, 231, 236, 250, 257,
comfort zone, 4, 270 260, 266, 267, 279, 283, 298
commercial, 2, 9, 174, 178, 179, 180, 181, 187, 190, convergence, 33, 86, 250, 260
200, 236, 271, 273 cooling, vii, viii, 1, 2, 3, 4, 5, 6, 7, 10, 11, 13, 17, 18,
complexity, 58, 78 57, 58, 60, 66, 67, 71, 96, 166, 185, 192, 270,
components, 1, 33, 164, 165, 197, 208, 275, 277, 272, 273, 274, 275, 277, 278, 279, 281, 282, 283,
279, 296, 299 284, 285, 286, 287, 288, 289, 290, 291, 292, 298,
composite, viii, 6, 58, 73, 127, 128, 187, 189, 190, 299, 302, 303, 304
192, 195, 196, 197, 198, 199, 200, 201, 202, 204, cooling process, 18, 71, 282
205, 206, 221, 222, 223, 224, 226, 227, 266 copolymer, 189
composition(s), 190, 191, 197 copper, 83, 109
compression, 297, 304 correlation(s), 21, 38, 39, 41, 48, 49, 131, 133, 135,
computation, 185, 232, 236, 247, 250 144, 145, 146, 147, 149, 153, 157, 171, 177, 240,
computational fluid dynamics, 268 244, 251, 253, 256, 257, 262, 264, 280, 298, 300
computational modeling, 157 cost-effective, 231
computer(s), 10, 18, 267, 283 costs, 1, 290
concentration, 19, 20, 25, 32, 33, 45, 93, 100, 124, coupling, 33, 109, 126, 157, 250, 259
126, 157, 161, 173, 187, 199, 213, 214, 223, 264 cross-linked, 58, 197
condensation, vii, 2, 4, 73, 270, 283, 287, 289, 292, cross-linking, 190
298 cross-sectional, 31, 60, 78, 82, 94, 229, 234
conditioning, vii, viii, ix, 1, 2, 3, 4, 5, 6, 9, 10, 12, cycles, 58, 68, 236, 237, 239, 240, 242, 243, 250,
13, 17, 125, 133, 157, 185, 189, 212, 222, 223, 251, 252, 253, 256, 257, 259, 260, 274, 284
225, 231, 269, 270, 271, 272, 273, 282, 290, 304
conductance, viii, 6, 99, 103, 108, 109, 114, 116,
117, 118, 119, 120, 121, 122, 123, 126, 127, 128, D
150, 151, 176, 182, 230
dead zones, 90, 92
Index 309
decoupling, viii, 5, 272 drying, 15, 18, 19, 20, 24, 25, 26, 43, 54, 128, 190,
deduction, 133, 144 225, 226
defects, 195 dynamic viscosity, 79, 105, 156
definition, 86, 112, 118, 139, 144, 149, 160, 245
degree, 52, 230, 297
delta, 245 E
demand, 1, 225
economies, vii
density, 16, 19, 30, 47, 59, 60, 69, 79, 101, 105, 122,
electric energy, 299
124, 149, 151, 156, 164, 188, 193, 204, 205, 207,
electric power, 282
209, 212, 213, 214, 215, 230, 232, 234, 245, 264,
electrical, 274
279
electricity, 286, 287, 297, 299, 302
dependent variable, 86, 233, 250
electrolyte, 187, 213
derivatives, 32
electronic(s), 2, 9, 19, 271
desorption, 58, 60, 66, 67, 68, 71, 72, 163, 197, 198,
emission, 4, 15, 26, 27, 28, 29, 30, 31, 32, 34, 37, 39,
279
40, 41, 46, 47, 49, 52, 53, 54, 138, 155, 163, 164,
destruction, 233
176, 182, 187, 204, 272
deviation, 37, 43
energy, vii, viii, ix, 1, 2, 3, 5, 6, 9, 12, 13, 16, 40, 55,
dew, 4, 5, 6, 17, 44, 181, 273, 291, 298
57, 58, 65, 66, 67, 73, 75, 77, 86, 96, 107, 131,
differential equations, 86, 142
134, 136, 140, 154, 157, 158, 161, 164, 174, 185,
differentiation, 20, 21, 138
225, 226, 232, 233, 236, 244, 246, 250, 259, 260,
diffusion, 6, 15, 17, 19, 20, 26, 27, 43, 46, 47, 50, 52,
269, 270, 271, 272, 273, 274, 277, 279, 280, 281,
53, 54, 58, 60, 61, 69, 72, 76, 113, 134, 148, 163,
282, 283, 286, 287, 290, 292, 297, 298, 299, 300,
173, 176, 188, 189, 190, 197, 199, 205, 206, 207,
301, 302, 303, 305
208, 210, 212, 213, 215, 216, 233, 250
energy consumption, vii, 1, 5, 272, 282, 286, 290,
diffusion mechanisms, 60, 197, 208
303, 304
diffusion process, 19
energy efficiency, 292
diffusion rates, 52
energy recovery, vii, 2, 12, 13, 65, 154, 185, 225,
diffusivities, 61, 127, 182
226, 290, 302, 303, 305
diffusivity, 15, 17, 19, 20, 21, 25, 26, 27, 30, 41, 43,
engineering, ix, 127, 131
47, 49, 50, 52, 53, 54, 55, 61, 79, 123, 124, 127,
enthalpy, 60, 136
129, 131, 134, 135, 147, 148, 149, 150, 151, 153,
envelope, vii, 4, 270, 282
155, 157, 164, 176, 185, 187, 199, 204, 206, 212,
environment, vii, 1, 4, 5, 18, 19, 270, 271, 272, 273,
213, 214, 215, 216, 225, 227, 229
274
dilation, 233
environmental, vii, 1, 3, 270
discipline, ix
environmental impact, vii, 1, 3, 270
discomfort, 4, 272
environmental issues, 1
discrete data, 251
environmental protection, vii
discretization, 236, 250, 260
epidemic, vii, 2
discs, 34
equilibrium, 18, 20, 23, 26, 44, 45, 46, 47, 134, 135,
diseases, vii, 2, 4, 272
148, 157, 176, 249, 273
displacement, 276, 297
equipment, 45, 225, 283, 290, 292
distillation, 185, 227
ethanol, 54
distillation processes, 227
Europe, 26
distilled water, 34, 43, 44, 202, 203
evaporation, 203
distribution, 41, 83, 84, 86, 97, 110, 111, 114, 128,
evolution, 33
206, 211, 236, 238, 240, 241, 259, 268, 276, 287,
exchange rate, 15, 48
300
experimental condition, 49, 50
draft, 6, 273
exponential, 21
dry, 4, 10, 11, 16, 18, 19, 35, 44, 47, 55, 56, 57, 58,
extraction, 185
60, 79, 100, 122, 123, 124, 132, 134, 156, 157,
164, 174, 192, 194, 203, 204, 205, 212, 215, 216,
223, 270, 273, 275, 277, 281, 287, 298
310 Index
gold, 201
F grains, 2, 9, 271
graph, 32, 71, 199, 200, 201, 236
fabric, ix
gravity, 281
fabrication, viii, 190
grids, 33, 84, 89, 110, 111, 113, 179, 237, 250, 259
factor analysis, 242
groups, 73
family, ix
growth, 2, 4, 9, 239, 271, 272
feet, 274
Guangzhou, 282
fiber, 225, 226
guidelines, ix
film(s), 54, 127, 129, 226, 231
filtration, 267
financial support, ix H
finite volume, 32, 164, 259
finite volume method, 164 handling, 6, 12, 13, 273
floating, 44 harmful, 297
flow field, 54, 231, 239, 241, 259 hazards, 287
flow rate, 16, 31, 33, 35, 36, 37, 40, 41, 44, 46, 50, head, 274
51, 55, 131, 134, 135, 137, 146, 149, 156, 171, heat capacity, 60, 137, 282
175, 179, 181, 182, 188, 192, 194, 195, 223, 224, heat conductivity, 127, 134, 148, 157, 176, 188, 221,
239, 243, 244, 253, 259, 269, 279, 282, 283, 290, 222, 225
291, 297, 299 heat exchangers, 131, 155
fluctuations, 279 heat loss, 166, 180
flue gas, 226 heat pumps, viii
fluid, ix, 27, 28, 40, 41, 48, 50, 54, 75, 77, 79, 81, heat transfer, viii, 6, 17, 55, 59, 60, 64, 69, 75, 77,
82, 96, 105, 106, 107, 108, 109, 113, 114, 117, 83, 89, 96, 97, 99, 100, 103, 107, 108, 114, 119,
120, 124, 126, 157, 160, 161, 229, 232, 233, 234, 120, 122, 126, 127, 128, 131, 132, 134, 135, 137,
236, 237, 244, 250, 251, 253, 257, 259, 260, 261, 138, 140, 142, 144, 148, 155, 160, 161, 163, 175,
268, 282 176, 185, 186, 225, 229, 234, 235, 241, 242, 244,
fluid interfaces, 108, 109, 126 245, 246, 250, 251, 253, 256, 257, 259, 260, 261,
fluoride, 189, 199 264, 265, 266, 267, 268, 275, 290, 298
foils, 12, 99, 132 heating, vii, 1, 2, 3, 10, 11, 13, 18, 19, 58, 60, 66, 67,
food, 54, 227 71, 166, 180, 192, 270, 273, 274, 275, 279, 282,
formaldehyde, 4, 272 291, 297, 299, 302
fossil fuel(s), vii, 3, 270 height, 15, 16, 30, 34, 47, 65, 75, 76, 78, 89, 93, 99,
Fourier, 20 101, 104, 109, 118, 120, 127, 149, 155, 157, 187,
friction, 75, 77, 81, 90, 91, 92, 96, 97, 106, 120, 128, 192, 204, 215, 237, 265
160, 185, 229, 231, 235, 237, 238, 239, 240, 242, high risk, 4, 272
245, 249, 253, 254, 255, 257, 262, 263, 264, 266, high temperature, 73
267 homogeneous, 58, 116, 180, 190, 199
fungi, 275, 287 Hong Kong, 10, 13, 185, 225
furniture, 2, 4, 9, 271, 272, 279 horizon, 34
hospital, 2, 271
hot water, 17
G human, 2, 4, 9, 10, 271, 272, 290
hydrate, 226
gas(es), 19, 26, 34, 35, 39, 43, 59, 60, 61, 135, 187,
hydro, viii, 6, 17, 21, 25, 26, 43, 45, 53, 54, 101,
188, 206, 208, 209, 210, 212, 222, 226, 234, 244,
157, 189, 190, 200, 216, 221, 224, 225, 226, 227,
268, 277, 284
231
gas diffusion, 212
hydrodynamic, 28, 59, 122, 176, 187
gauge, 18, 181
hydrodynamics, 53
gel, 18, 25, 54, 57, 225, 279, 281, 283
hydrogen bonds, 190
generation, 97, 113, 128, 180, 269, 279, 303
hydrophilic, viii, 6, 17, 21, 25, 26, 43, 45, 53, 54,
Germany, 194
101, 157, 189, 190, 200, 216, 221, 224, 225, 226,
glass, 34, 109, 120, 201, 283
227, 231
Index 311
hydrophilic groups, 190

hydrophilicity, 190, 197
L
hydrophobic, viii, 6, 180, 189, 190, 199, 200, 212,
lamina, viii, 6, 28, 31, 60, 73, 75, 76, 77, 78, 79, 89,
221, 224, 226, 227
96, 97, 99, 105, 113, 150, 157, 172, 181, 229,
237, 244, 245, 253, 257, 260, 262
I laminar, viii, 6, 28, 31, 54, 60, 73, 75, 76, 77, 78, 79,
89, 96, 96, 97, 99, 105, 113, 127, 128, 150, 157,
images, 195 172, 181, 185, 186, 229, 232, 237, 244, 245, 253,
in transition, 290 257, 260, 262, 266, 267
inactive, 68 Langmuir, 21
incentives, vii, 3, 270 Laplace transformation, 142
incompressible, 232, 233, 234, 244 law(s), 19, 43, 45, 46, 60, 244, 249
independence, 113, 250, 260 lead, 43, 94, 261
independent variable, 61 learning, ix
indices, 179, 286 limitations, 185
industrial, 45, 279 linear, 38, 84, 111, 125, 139, 146, 165, 214
industrial application, 45 linear regression, 38
industrialized countries, 1 links, 164
industry, vii, viii, 2, 4, 6, 9, 12, 68, 125, 197, 212, liquid nitrogen, 201
222, 223, 269, 270, 271 liquid water, 49, 60, 189
infection, 4, 272 literature, 17, 89, 127, 206, 237, 260
infinite, 142, 144, 176 lithium, 226, 227, 279
inhalation, 268 location, 40, 50, 107, 124, 253, 287
insight, ix, 135, 155 logging, 46
inspection, 94 London, 54, 77, 83, 96, 103, 128, 144, 185, 267
insulation, 166, 180 long-term, ix, 13, 101, 185, 225
intensity, 250 losses, 43
interaction(s), 21, 162, 226, 253, 279 lumen, 77, 96, 185
interface, 49, 202
inventions, 189
investment, 274
M
ionic, 201, 227
machines, 101
ionic liquids, 227
magnetic, 17
iron, 109
maintenance, 101, 274
irradiation, 4, 272
malic, 190, 191
isothermal, 25, 46, 268
manufacturer, 282
isotherms, 21, 26, 93, 114, 197, 240, 253
manufacturing, 2, 9, 58, 271
iteration, 86, 250
market, viii, 266, 289
market penetration, viii, 289
J market prices, 289
mass transfer process, 73, 185
Jacobian, 75, 86, 112 material surface, 26
Japan, 96 mathematical, ix, 26, 58, 155, 164, 185
Japanese, 267 matrix, 58, 86, 165, 277, 278
Jerusalem, 267 measurement, 17, 21, 27, 34, 41, 53, 179, 182, 223
measures, viii, 6, 269, 290, 302, 303, 305
mechanical, 2, 9, 45, 101, 128, 190, 195, 197, 271,
K 291, 292, 293, 296, 303
media, 6, 16, 99, 157, 200, 227, 281
kinetic energy, 40, 229, 234, 246, 248, 249 membrane permeability, 27
Kolmogorov, 245 membranes, viii, 6, 15, 17, 43, 50, 51, 52, 53, 54,
101, 128, 133, 146, 147, 150, 153, 157, 187, 189,
312 Index
190, 192, 194, 195, 197, 198, 199, 200, 201, 206, Nusselt, 60, 75, 83, 99, 105, 107, 113, 114, 117, 119,
207, 212, 224, 225, 226, 227, 230, 231, 267 120, 121, 122, 131, 135, 155, 157, 160, 161, 170,
memory, 251 171, 172, 173, 176, 229, 234, 235, 237, 240, 243,
metal ions, 199 244, 245, 251, 252, 253, 255, 256, 257, 259, 260,
metals, 122, 127 262, 263, 264, 266
methane, 226
methylcellulose, 25, 54, 127, 129
microstructure, 201 O
mirror, 17, 44, 181
observations, 93, 201
mites, 275
odors, 276
mixing, 35, 44, 134, 203
oil, viii
modeling, viii, ix, 6, 7, 32, 55, 58, 76, 133, 155, 157,
operator, 75, 86, 112, 233
164, 173, 227, 231, 257, 258, 260, 267, 269, 280,
optimization, 17, 96, 131, 133
296, 298, 304
organic, 54, 199
models, 26, 185, 206, 229, 231, 245, 251, 252, 253,
organic compounds, 54
255, 257, 259, 260, 266, 267, 280, 281, 283, 298
orientation, 230
modern society, 4, 9, 272
oxide, 189, 226
modules, 13, 185, 226
ozone, 273
modulus, 247
moisture content, 135, 282
moisture sorption, 15 P
molecules, 21, 190, 209
momentum, 40, 86, 113, 158, 164, 229, 232, 233, Pacific, 54
238, 240, 244, 246, 251, 253, 256, 260, 262 paper, 6, 57, 101, 103, 109, 120, 133, 149, 150, 153,
morphological, 211 174, 176, 177, 178, 180, 182, 184, 201, 231, 237,
motors, 286 251
parameter, 4, 50, 100, 103, 108, 109, 118, 121, 126,
150, 174, 249, 270
N
partial differential equations, 46, 164, 178
partition, 15, 45, 125, 127, 139, 149, 151
NaCl, 47, 205
passive, 57
Nafion, 43, 54, 127, 128, 189, 225
pathogenic, 4, 272
natural gas, 274
payback period, 289
Navier-Stokes equation, 28, 32, 79, 234, 244
pearlite, 54
Netherlands, 279
Peclet number, 46, 164
New Jersey, 305
peer, ix
New York, 74, 96, 97, 128, 153, 154, 185, 225, 267,
penalty, 257, 266
268
performance, viii, ix, 1, 6, 17, 25, 43, 54, 55, 58, 62,
Newtonian, 79, 105, 157
67, 69, 70, 73, 75, 76, 77, 103, 120, 122, 131,
nitrogen, 37, 225
132, 133, 147, 153, 155, 157, 179, 185, 187, 216,
nodes, 33, 65, 85, 89, 113, 178
217, 219, 226, 227, 231, 267, 269, 277, 279, 282,
non-linear, 65, 86, 245, 267, 268, 298
283, 284, 292, 296, 297, 298, 300, 305
non-linearity, 245
periodic, 237, 239, 250, 251, 252, 253, 254, 255,
non-metals, 127
257, 267
non-uniform, 17, 53
periodicity, 257, 259
non-uniformity, 17
permeability, 54, 129, 187, 188, 189, 192, 197, 215,
normal, 25, 32, 107, 113, 124, 206, 235
216, 217, 218, 219, 220, 222, 223, 224
normal distribution, 206
permeable membrane, viii, 187, 225
novel materials, viii, ix, 15
permeation, 15, 43, 53, 54, 101, 140, 144, 147, 150,
number of Transfer Units (NTU), 15, 49, 50, 53, 55,
182, 189, 190, 191, 193, 194, 195, 197, 202, 216,
69, 70, 73, 131, 132, 133, 135, 137, 140, 143,
217, 218, 219, 220, 222, 224, 226, 227
144, 145, 146, 147, 149, 152, 153, 155, 157, 171,
pharmaceutical, 2, 9, 271
188, 298
physical properties, 45, 46, 194, 197, 216
numerical analysis, 65, 77, 97
Index 313
physics, 244 pure water, 214

pitch, 30, 230, 236 purification, 185
planar, 27 PVA, 226
plastic(s), 180, 231, 274 PVC, 201
platinum, 17
plywood, 109
Poisson equation, 83, 110 R
polarization, 185
radius, 15, 31, 36, 37, 38, 40, 41, 47, 51, 52, 55, 61,
poly(phenylene oxide) (PPO), 226
78, 188, 204, 205, 216
poly(vinylchloride), 54
range, ix, 1, 20, 45, 119, 123, 135, 212, 244, 249,
polyether, 189, 226
251, 257, 281, 282, 297, 299
polymer(s), 17, 25, 26, 43, 45, 53, 54, 101, 103, 109,
raw material, 202
120, 127, 189, 190, 199, 216, 227, 266
reality, 4, 164, 272, 279
polymer film, 54
recovery, viii, ix, 1, 4, 5, 6, 7, 9, 10, 13, 15, 16, 55,
polymer membranes, 43, 45, 53, 54, 101, 189, 199,
57, 58, 62, 65, 66, 67, 68, 69, 70, 71, 72, 73, 76,
227
99, 103, 136, 269, 270, 272, 273, 275, 279, 290,
polynomial, 49, 125
292, 293, 296, 302, 303, 304
polypropylene, 190
recovery technology, viii, 269
polystyrene, 189
reduction, vii, 2, 190
polystyrenesulfonate, 128, 226
reference frame, 233
polyurethane, 189, 226
refrigerant, 13, 296, 297, 298, 299
polyvinyl alcohol, 43
refrigeration, viii, 13, 274, 292, 296, 300, 301, 305
pools, 2, 9, 271
regenerate, 57, 273, 277, 279, 281
poor, 4, 67, 272
regenerated cellulose, 189
poor performance, 67
regeneration, 19, 24, 26, 274, 279, 282, 284, 286,
pore(s), 60, 61, 71, 73, 187, 188, 200, 206, 207, 209,
291, 292
210, 211, 212, 217, 219
regular, 75, 78, 260
porosity, 54, 56, 61, 188, 207, 217, 218, 219, 220,
regulations, vii, 4, 6, 9, 270, 272
226
relationship(s), 143, 208, 209, 210, 282, 298
porous, 190, 192, 195, 199, 206, 207, 221, 222, 225
relaxation, 33, 84, 111, 250
porous media, 206, 221, 225
reliability, 101
positive reinforcement, ix
renewable energy, 285
powder, 190
research, vii, ix, 1, 231, 303
power, 19, 70, 180, 282
residential, 304
Prandtl, 75, 100, 106, 122, 160, 229, 244, 245, 247
residues, 33
prediction, 251, 260, 279
resistance, 15, 17, 18, 43, 50, 58, 69, 113, 132, 137,
preparation, 201, 202
138, 139, 140, 144, 146, 147, 153, 171, 172, 173,
pressure, 19, 22, 23, 29, 30, 33, 35, 40, 44, 61, 62,
182, 186, 188, 190, 192, 194, 205, 210, 213, 215,
70, 79, 96, 105, 137, 155, 179, 181, 188, 189,
216, 217, 219, 223, 225, 262, 265, 266
199, 208, 209, 210, 212, 213, 229, 233, 234, 235,
resolution, 259
236, 237, 249, 250, 253, 257, 259, 260, 266, 269,
resources, vii, 3, 270
273, 282, 286, 292, 297
respiratory, 2, 287
prevention, vii, 2
retail, 2, 3
prices, viii
returns, 275, 277, 292
probability density function, 206
Reynolds number, 15, 30, 31, 33, 37, 75, 76, 81, 96,
procedures, 18, 165, 179, 185
100, 106, 150, 156, 157, 160, 171, 181, 205, 229,
product design, 103
234, 239, 240, 243, 244, 245, 247, 253, 255, 257,
program, 103, 110, 113, 114, 279
259, 262, 263, 266, 267
promote, 240
Reynolds stress model, 249, 251, 257, 268
property, 279
room temperature, 46, 201, 208, 287, 288
protocols, 1
rotation axes, 40, 41
public, viii
pumps, 17, 35, 44, 203
314 Index
sorption process, 23, 67

S Southeast Asia, 2, 9, 271
spacers, 127, 230
salt(s), 190, 194, 197, 226
species, 207
sample, 9, 13, 18, 19
specific heat, 10, 59, 60, 79, 105, 122, 134, 136, 139,
SARS, 2, 4, 272
155, 175, 229, 234, 245
saturation, 19, 22, 62, 137, 281
speed, 19, 55, 63, 67, 68, 69, 73, 179, 181, 281, 283,
savings, 7, 9, 13, 185, 225, 269, 274, 302, 303, 304
297, 305
scalar, 233
sporadic, viii, 127
Scanning Electron Microscope (SEM), 194, 195,
square wave, 52
196, 197, 199, 200, 201
stability, 227
schema, 276, 277, 278
stabilize, 217
Schmidt number, 16, 30, 76, 100, 122, 156, 160, 205,
stages, 19
264
stainless steel, 34, 166, 167, 192
science, 153, 185
standard deviation, 188, 207
selecting, 45, 83, 110, 251
standardization, vii, 4, 270
selectivity, 188, 199, 222, 223, 224, 226
steady state, 43, 49, 52, 179, 223, 244
sensors, 17, 35, 44, 180, 181, 203
steel, 109
separation, 43, 199, 226, 267
stochastic, 245
series, 20, 142, 144, 214, 230, 275
storage, vii, 5, 272
Shanghai, 19
strain, 18, 247
shape, 15, 45, 58, 61, 83, 94, 127, 135, 155, 164, 259
strategies, 1, 266, 297, 302, 304
shear, 230, 245, 262
streams, 35, 43, 44, 58, 60, 64, 71, 103, 127, 131,
Sherwood number, 16, 31, 32, 35, 36, 37, 75, 76, 81,
133, 134, 135, 136, 137, 139, 149, 162, 163, 171,
82, 83, 89, 90, 91, 92, 94, 95, 96, 100, 123, 124,
173, 174, 178, 180, 182, 192, 194, 203, 222, 223,
126, 156, 157, 160, 161, 171, 172, 173, 176, 188,
226
205, 264
strength, 45, 101, 128, 190
signs, 78
stress, 230, 233, 245, 249, 257
silica, 18, 57, 279, 281, 283
students, ix
silicate, 127
summer, vii, 2, 4, 9, 10, 12, 17, 57, 270, 271, 275,
siloxane, 189
283, 284, 287, 288, 300
similarity, 189
sunlight, 10
simulation, 28, 73, 157, 249, 251, 267, 279, 283,
supply, 2, 4, 5, 6, 35, 44, 46, 78, 132, 203, 272, 273,
304, 305
274, 275, 277, 279, 280, 281, 286, 287, 292, 299
simulations, 173, 216, 259, 267, 279, 299
supported liquid membrane, viii, 6, 187, 189, 198,
sine, 77, 78, 90
199, 216, 223, 224, 226, 227, 266
Singapore, 54
surface area, 78, 207, 229, 234, 240
skin, 199
surface diffusion, 58, 60, 61
SO2, 199
surplus, 292
society, vii, 1, 2, 6, 12
sustainable development, vii, 2, 4, 270
software, 181
symmetry, 31, 32, 40, 58, 113, 124, 157, 238
solar, 4, 272, 274, 285, 304
synthesis, ix
solar energy, 274, 285
synthetic, 226
solid phase, 64
systematic, viii
solid surfaces, 157, 261
systems, vii, viii, ix, 1, 2, 4, 5, 6, 7, 9, 13, 17, 54, 76,
solubility, 54
86, 157, 161, 226, 270, 272, 273, 274, 276, 279,
solutions, 33, 165, 280
282, 283, 287, 288, 290, 292, 293, 301, 302, 303,
solvent, 190
304, 305
sorption, 15, 17, 18, 20, 21, 23, 25, 26, 45, 53, 54,
55, 61, 65, 67, 68, 73, 125, 131, 133, 135, 139,
147, 157, 164, 194, 197, 198, 275, 281 T
sorption curves, 20, 23, 25, 26, 55, 139
sorption experiments, 17, 54, 194 tanks, 267
sorption isotherms, 26, 53, 197, 198 Tc, 301
Index 315
technology, viii, 53, 58, 189 uniform, 32, 40, 41, 73, 76, 79, 82, 83, 84, 96, 97,
Teflon, 109 100, 103, 105, 111, 113, 114, 116, 120, 126, 128,
temperature gradient, 235, 240, 253, 261, 274 157, 161, 171, 172, 173, 186, 216, 236, 249, 250,
test procedure, 203 257, 262, 264, 266, 267
textbooks, 298
theoretical, ix, 20, 103, 206
theory, 136, 190, 206, 207, 212, 298 V
thermal, vii, ix, 2, 4, 5, 9, 58, 60, 69, 79, 94, 96, 105,
vacuum, 185, 190, 208, 227
117, 118, 119, 122, 132, 137, 156, 157, 161, 163,
validation, ix, 110, 251, 260
171, 172, 173, 190, 203, 233, 234, 240, 242, 253,
validity, 39
270, 271, 272, 274, 279, 286, 298
values, 4, 22, 23, 25, 37, 39, 43, 46, 50, 51, 65, 68,
thermal analysis, 279
84, 86, 89, 90, 93, 94, 111, 113, 114, 118, 120,
thermal energy, vii, 4, 270, 274
121, 126, 127, 137, 140, 146, 161, 164, 165, 171,
thermal properties, 79, 105, 157
172, 175, 178, 179, 192, 197, 216, 223, 235, 237,
thermal resistance, 58, 137, 172, 173
240, 241, 243, 245, 246, 250, 256, 259, 260, 272,
thermal stability, 190
281, 282, 302
thermodynamic, 213, 214, 234, 296
vapor, viii, 12, 15, 16, 19, 22, 23, 24, 30, 33, 46, 47,
thermodynamic calculations, 296
52, 53, 54, 55, 62, 79, 81, 100, 101, 123, 124,
thermodynamic equilibrium, 214
128, 129, 134, 135, 137, 156, 176, 187, 189, 190,
thermodynamics, ix
191, 192, 193, 194, 197, 198, 199, 203, 204, 208,
three-dimensional, 32, 268
209, 210, 213, 216, 222, 223, 224, 225, 226, 227,
Ti, 10, 106, 234, 281
229, 273, 297, 298, 304
time, 2, 4, 9, 16, 17, 19, 20, 21, 23, 24, 26, 47, 49,
variable(s), 4, 13, 29, 32, 38, 56, 62, 141, 164, 165,
51, 56, 59, 62, 65, 66, 67, 78, 83, 101, 110, 133,
171, 179, 181, 234, 243, 251, 272, 290, 305
201, 204, 229, 230, 232, 237, 244, 245, 251, 271,
variation, 37, 71, 94, 165
272, 279, 283, 284
vector, 237, 251, 260
time consuming, 133
velocity, 16, 17, 29, 31, 33, 39, 40, 41, 47, 59, 79,
TiO2, 267
80, 81, 82, 86, 92, 94, 96, 105, 106, 113, 114,
tissue, 201
115, 122, 150, 156, 158, 160, 162, 188, 194, 202,
total energy, vii, 1, 2, 5, 6, 12, 136, 271, 272, 286,
204, 230, 232, 234, 235, 236, 237, 238, 239, 241,
290, 302
245, 249, 251, 253, 257, 259, 260, 276
transformation(s), 21, 83, 86, 110, 112
ventilation, vii, 1, 2, 4, 5, 6, 9, 10, 11, 12, 13, 101,
transition(s), 211, 257
123, 135, 226, 270, 271, 272, 274, 276, 283, 286,
transmembrane, 189, 199, 209, 216
289, 290, 301
transmission, 4, 272, 282
ventilators, vii, 2
transport, viii, 6, 43, 45, 46, 49, 51, 52, 53, 54, 128,
versatility, 44
134, 155, 157, 164, 165, 190, 199, 203, 204, 206,
vinylidene fluoride, 43
209, 225, 226, 227, 231, 233, 234, 240, 244, 247,
virus(es), 4, 272, 275, 287
249, 267, 298
viscosity, 16, 30, 76, 100, 149, 188, 205, 209, 230,
transport phenomena, 157
233, 245, 264
trend, vii, 6, 253, 273, 290
voids, 21, 58
turbulence, 245, 246, 249, 250, 251, 252, 253, 255,
vortices, 40
256, 257, 260, 261, 264, 266, 267
turbulent, viii, 6, 229, 230, 244, 245, 246, 247, 249,
250, 251, 257, 262, 264, 267, 268 W
turbulent flows, 257
two-dimensional, 53, 58, 72, 79, 97, 128, 135, 174, wall temperature, 73, 76, 97, 103, 105, 106, 120,
182 234, 249, 267
waste, 274, 292
wastewater treatment, 199
U
water, vii, 2, 9, 10, 16, 17, 18, 19, 20, 21, 22, 23, 24,
25, 26, 29, 34, 35, 37, 43, 44, 45, 47, 52, 53, 54,
ultra-thin, 231
56, 60, 61, 71, 99, 101, 124, 127, 131, 132, 134,
uncertainty, 25, 44, 50, 114, 179, 181, 192
316 Index
135, 147, 148, 156, 157, 164, 166, 187, 190, 192, winter, vii, 10, 11, 12, 16, 57, 275, 283, 284, 287,
195, 196, 197, 202, 208, 209, 210, 213, 216, 224, 300
225, 226, 273, 274, 275, 279, 282, 283, 287, 289, working conditions, 192
290, 292, 298 working hours, 287, 289
water diffusion, 124 writing, ix
water evaporation, 10, 216
water sorption, 197
water vapor, 2, 9, 17, 19, 21, 22, 29, 43, 132, 134, Y
157, 190, 197, 208, 209, 210, 224, 225, 226, 273
yield, 140
wet, 4, 34, 54, 57, 58, 270, 281, 297
wet coating, 54
wind, 18, 19, 179, 181 Z
wind tunnels, 180
windows, 282 zeolites, 279

Total Heat Recovery Heat and Moisture Recovery From Ventilation Air

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Total Heat Recovery Heat and Moisture Recovery From Ventilation Air

Caricato da

Copyright:

Formati disponibili

TOTAL HEAT RECOVERY:

HEAT AND MOISTURE RECOVERY

Nova Science Publishers, Inc.

NOTICE TO THE READER

LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA

Published by Nova Science Publishers, Inc. New York

TOTAL HEAT RECOVERY IN AIR-CONDITIONING

1.2. OPPORTUNITIES AND CHALLENGES

Domestic loads 0.7

Figure 1.1. Sources of moisture loads in a medium size retail store.

Fresh air out

Exhaust air out

Figure 1.2. Schematic of an energy wheel.

Figure 1.3. Schematic of a stationary total heat exchanger.

The depleting nature of primary energy resources, negative environmental impact of

Humidity ratio (g/kg)

Figure 1.4. Comfort zone in a psychrometric chart.

ENERGY RECOVERY POTENTIALS

2.2. ENERGY CALCULATIONS

Figure 2.2. Schematic of a cross-flow parallel-plates total heat exchanger.

Sensors Cooling coils Heater Humidifier

ESTIMATION OF SORPTION AND DIFFUSION

ν kinematic viscosity of air (m2/s)

summer. In a membrane total heat exchanger system, moisture is adsorbed by membrane

3.2. SORPTION AND DIFFUSIVITY IN A THERMO-HYGROSTAT

HUTC-MEM02, a novel hydrophilic desiccant material sheet is considered for air

Figure 3.1. Photographic view of the thermo-hygrostat.

t=0 0<z<z0 w = win (3.2)

Dvm = [(d (ln Ω) / dt ) exp /(d (ln Ω) / dFo) th ]z 02 (3.7)

Another logarithmic transformation of the above equation gives

A differentiation of Eq. (3.9) suggests

The measured isotherms of the HUTC-MEM02 (a novel hydrophilic desiccant sheet

Water uptake (kg/kg)

Figure 3.3. Water uptake in relation to relative humidity of air.

Figure 3.4. Sorption curves of the desiccant sheet at two conditions.

Table 3.1. Values in the model of uptake curves

Figure 3.5. Regenerating curves for HUTC-MEM02 and vapor.

The constants in the equation are listed in Table 3.2.

Table 3.2. The constants in the model of regeneration curves

Temperature a b c Uncertainty (%)

Time 5min 10min 20min 1hr 1.5hr

The diffusivity of moisture in desiccant membrane is calculated with Eq.(3.7) as

w (kg H2 O/kg dry membrane)

Figure 3.6. Calculated diffusivity of moisture in desiccant membrane.

A method of directly measuring the sorption, drying, and diffusion characteristics of a

3.3. FLUID FLOW AND MASS TRANSFER IN A NOVEL EMISSION CELL

Figure 3.7. A schematic showing the flow geometry of the cell.

CFD simulation is employed to calculate the convective mass transfer coefficients

where p represents pressure (pa).

∂uφ* ∂uφ* uφ* ∂uφ* 1 ∂p * ⎡ 1 ∂ ⎛⎜ * ∂uφ ⎞ ∂ 2 uφ* ∂ 2 uφ* ⎤

Conservation of water vapor

where ρ is density (kg/m3). The dimensionless humidity ratio is given as

where Dva is vapor diffusivity in the air mixture (m2/s).

where k is the convective mass transfer coefficient (m/s).

where ωb is the bulk humidity ratio on a cross-section at radius r.

The cross-sectional area is

The local Sherwood number is

Similarly, the mean Sherwood number is

where the bulk dimensionless humidity ratio is

The inlet and boundary conditions for mass transfer are

r*=r0*: θ=1; (3.41)

z*=0: θ=0 (3.42)

where n is the normal direction.

r=r0: θ=1; (3.41)