Sei sulla pagina 1di 25

12

Flow Maldistribution and Header Design

One of the common assumptions in basic heat exchanger design theory is that fluid be distributed uniformly at the inlet of the exchanger on each fluid side and throughout the core. However, in practice, flow maldistribution' is more common and can significantly reduce the desired heat exchanger performance. Still, as we discuss in this chapter, this influence may be negligible in many cases, and the goal of uniform flow through the exchanger is met reasonably well for performance analysis and design purposes. Flow maldistribution can be induced by (1) heat exchanger geometry (mechanical design features such as the basic geometry, manufacturing imperfections, and tolerances), and (2) heat exchanger operating conditions (e.g., viscosity- or density-induced maldistribution, multiphase flow, and fouling phenomena). Geometry-induced flow maldistribution can be classified into (1) gross flow maldistribution, ( 2 ) passage-topassage flow maldistribution, and (3) manifold-induced flow maldistribution. The most important flow maldistribution caused by operating conditions is viscosityinduced maldistribution and associated flow instability. In this chapter, we consider geometry-induced flow maldistribution in Section 12.1 and operating condition-induced flow maldistribution in Section 12.2. Next, mitigation of flow maldistribution is discussed in Section 12.3. Finally, header design for compact heat exchangers is summarized in Section 12.4.

12.1 GEOMETRY-INDUCED FLOW MALDISTRIBUTION


One class of flow maldistribution, which is a result of geometrically nonideal fluid flow passages or nonideal exchanger inlet/outlet header/tank/manifold/nozzle design, is referred to as geometry-induced flow maldistribution. This type of maldistribution is closely related to heat exchanger construction and fabrication (e.g., header design, heat exchanger core fabrication including brazing in compact heat exchangers). This maldistribution is peculiar to a particular heat exchanger in question and cannot be influenced significantly by modifying operating conditions. Geometry-induced flow maldistribution is related to mechanical design-induced flow nonuniformities such as (1) entry conditions, (2) bypass and leakage streams, (3) fabrication tolerances,
F / o v muldisrriburio~iis defined as nonuniform distribution of the mass flow rate on one or both fluid sides in any of the heat exchanger ports andlor in the heat exchanger core. The term iden/ fluid flow passage/header/heat exchanger would, as a rule, denote conditions of uniform mass flow distribution through an exchanger core.

809

810

FLOW MALDISTRIBUTION A N D HEADER DESIGN

(4) shallow bundle effects,+ and (5) general equipment and exchanger system effects (Kitto and Robertson, 1989). The most important causes of flow nonuniformities can be divided roughly into three main groups of maldistribution effects: (1) gross flow maldistribution (at the inlet face of the exchanger), (2) passage-to-passage flow maldistribution (nonuniform flow in neighboring flow passages), and (3) manifold-induced flow maldistribution (due to inlet/outlet manifoldiheader design). First, we discuss gross flow maldistribution. Subsequently, the passage-to-passage flow maldistribution is addressed, followed by a few comments related to manifold-induced flow maldistribution.

12.1.1

Gross Flow Maldistribution

The major feature of gross flow maldistribution is that nonuniform flow occurs at the macroscopic level (due to poor header design or blockage of some flow passages during manufacturing, including brazing or operation). The gross flow maldistribution does not depend on the local heat transfer surface geometry. This class of flow maldistribution may cause (1) a significant increase in the exchanger pressure drop, and (2) some reduction in heat transfer rate. To predict the magnitude of these effects for some simple exchanger flow arrangements, the nonuniformity will be modeled as one- or two-dimensional as follows, with some specific results. Gross flow maldistribution can occur in one dimension across the free-flow area (perpendicular to the flow direction) as in single-pass counterflow and parallelflow exchangers, or it can occur in two or three dimensions as in single- and multipass crossflow and other exchangers. Let us first model a one-dimensional gross flow maldistribution with an N-step inlet velocity distribution function. The heat exchanger will be represented by an array of N subunits, called subexchangers, having uniform flow throughout each unit but with different mass flow rates from unit to unit. The number of subexchangers is arbitrary, but it will be determined to be in agreement with the imposed flow maldistribution. The set of standard assumptions of Section 3.2.1 is applicable to each subexchanger. The following additional idealizations are introduced to quantify the influence of flow nonuniformity caused by gross flow maldistribution on each subexchanger and the exchanger as a whole.
1. Total heat transfer rate in a real heat exchanger is equal to the sum of the heat transfer rates that would be exchanged in N subexchangers connected in parallel for an idealized N-step inlet velocity distribution function. 2. The sum of the heat capacity rates of the respective fluid streams for all subexchangers is equal to the total heat capacity rates of the fluids for the actual maldistributed heat exchanger.

With these auxiliary assumptions, the temperature effectiveness of a counterflow/ parallelflow heat exchanger can be calculated by modeling the heat exchanger as a parallel coupling of N subexchangers for a maldistributed fluid stream having N indiviUse of an axial tube-side nozzle (if the depth of the head is insufficient to allow the jet from the nozzle to expand to the tubesheet diameter) may result in larger flow in central tubes and lead to flow maldistribution (Mueller, 1987).

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

811

dual uniform fluid streams (i.e., having an N-step function velocity distribution). The other fluid side is considered as having uniform flow distribution for such an analysis. If flow nonuniformity occurs on both fluid sides of a counterflow or parallelflow exchanger, the exchanger is divided into a sufficient number of subexchangers such that the flow distributions at the inlet on both fluid sides are uniform for each subexchanger. For all other exchangers, the solution can only be determined numerically, and the solutions of Sections 12.1.1.1 and 12.1.1.2 are not valid in that case. 12.1.1.1 Counterflow and ParallelJpow Exchangers. In this section, we derive an expression for the exchanger effectiveness and hence heat transfer performance for counterflow and parallelflow exchangers having an N-step velocity distribution function on the fluid 1 side and perfectly uniform flow distribution on the fluid 2 side, as shown in Fig. 12.la. Here fluid 1 can be either the hot or cold fluid, and in that case, fluid 2 will be the cold or hot fluid. Subsequently, we apply this analysis to a heat exchanger having a two-step velocity distribution function at the inlet. Heat Transfer Analysis. Let us consider a counterflow exchanger with an N-step inlet distribution function of fluid 1, shown in Fig. 12.la. The same analysis would be valid for a parallelflow heat exchanger. Fluid 2 is considered uniform. We may model this exchanger as an array of N subexchangers, each obeying the standard assumptions of Section 3.2.1. Hence, (12.1)
i=A

where q represents the total heat transfer rate, and qj,j = A, B, ....N, the fractions of heat transfer rate in N hypothetical subexchangers, each having uniform mass flow rates on both sides, as shown in Fig. 12.lb. The assumptions invoked, including the auxiliary ones introduced above, lead to the following results:

L ( m x &
Fluid 2 side
c 2

.............

CIA

ClB

.....
C1.N

.-.......

(b)
FIGURE 12.1 Idealized two-step function flow nonuniformity on fluid I side and uniform flow on fluid 2 side of a counterflow exchanger.

812

FLOW MALDISTRIBUTION A N D HEADER DESIGN

Substituting Eqs. (12.2) and (12.3) into Eq. (12.1) and rearranging, the expression for fluid 1 temperature effectiveness becomes
(1 2.4)

where (12.5) Note that Eqs. (12.4) and (12.5) are valid for a maldistributed fluid regardless of whether it is hot or cold, C,,, or C,,,, or denoted as 1 or 2. The subscript 1 in these equations may be replaced by a designator of the maldistributed stream, say as in P,,,,\ and C,,,,. The temperature effectiveness of fluid 1 for each of the subexchangers of Eq. (12.4) is computed knowing individual NTU and heat capacity rate ratio:
PI I = PI , ( N T U ,
I.

">
c 2
/

j = A.B.. . . . N

(12.6)

on the right-hand side of Eq. (12.6) is computed for each exchanger using where the expression provided in Table 3.6 as follows:

Application of Eq. (12.6) requires the values of N T U I , , . CI /, and Cz for each subexchanger. To determine these variables. we should invoke the standard assumptions of Section 3.2.1 to get the free-flow area and heat capacity rate ratio as (12.8)

(12.9) Similarly, for fluid 2, we get (12.10) Note that the sets of relations given by Eq. (12.9) and (12.10) may be reduced by one equation each by utilizing Eq. (12.5) for fluid 1 (and similarly for fluid 2). The number of heat transfer units and the capacity rate ratios for Eq. (12.7) can be determined from their definitions as follows:

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

813

(12.11)
RI,,=%
2 '
J

j = A , B , . . . ,N

(12.12)

Note also that Dh = 4A,L/A

with Dll and L identical for all subexchangers

j = A, B , . . . , N for the counterflow/parallelflow heat exchanger. A I , , / A ~ of Eq. (12.1 1)

is obtained then from the definitions of Dh as

In addition,

(2) (?)I= (2)


A .
I
2

j = A , B , ...,N

(12.13)

j = A , B , . . . ,N

(12.14)

The reduction in the temperature effectiveness on the maldistributed side can be presented by the performance (effectiveness) deterioration factor as

APT = Pl.ideal - PI PI .ideal

or A&*=-&ideal - &
&ideal

(12.15)

where Pl,jdeal represents the temperature effectiveness for the case of having no flow maldistribution. The influence of gross flow maldistribution is shown in Fig. 12.2 for a balanced (C* = 1) counterflow heat exchanger in terms of A&* for a two-step inlet velocity distribution function (i.e., for two subexchangers). For a particular value of umax/~,,,

050

r*,

NTU

FIGURE 12.2 Performance deterioration factor A* for a balanced heat exchanger, C* = I , N = 2. (From Shah, 1981.)

814

FLOW MALDISTRIBUTION AND HEADER DESIGN

and given NTU, we find the greatest reduction in the heat exchanger effectiveness occurring when two-step function flow maldistribution occurs in equal flow areas (50 : 50%). The effect of flow maldistribution increases with NTU for a counterflow exchanger. Note that the reduction in the temperature effectiveness P I of Eq. (12.4), obtained using Eq. (12.15), is valid regardless of whether the maldistributed fluid is the hot, cold, C,,,, or Cminfluid. We can idealize an N-step function velocity distribution into an equivalent two-step function velocity distribution. Based on the analysis of passage-to-passage flow maldistribution presented in Section 12.1.2, it is conjectured that the deterioration in the exchanger effectiveness is worse for the two-step function velocity distribution. Hence, conservatively, any flow maldistribution can be reduced to a two-step function, and its effect can readily be evaluated on the exchanger effectiveness, which will represent the highest deterioration. As indicated above, the discussion of the effect of the gross flow maldistribution in this section refers to a heat exchanger with counterflow arrangement and balanced flow (C* = I). Hence, the increased effect of flow maldistribution with increasing NTU is valid only for this special situation. If either C* # 1 or if the flow arrangement is parallelflow, the influence of flow maldistribution may decrease with increasing NTU. This can be determined from Eq. (12.15), assuming the validity of the appropriate effectivenessNTU relationships for each subexchanger and for the heat exchanger as a whole. Also, the actual flow rate conditions in most practical cases would not correspond to a balanced heat exchanger case.
Pressure Drop Analysis. There is no rigorous theory available for predicting a change in the pressure drop due to flow maldistribution in the exchanger. This is because for nonuniform flow distribution, the static pressures at the core inlet and outlet faces will not be uniform, and hence, constant pressure drop across the core is not a valid assumption. The following is a suggested approximate procedure. This approach is not based on a rigorous modeling of the actual flow conditions and must be used very cautiously. Consider a two-step function velocity distribution at the core inlet on fluid I side as shown in Fig. 1 2 . 1 ~ for N = 2. Subexchangers in Fig. 12.lb for N = 2 are in parallel. Using Eq. (6.28), evaluate the pressure drop Apj for a specific subexchanger which has the highest fluid velocity in the flow passages. Also compute Apuniform for fluid 1 considering the flow as uniform at the core inlet in Fig. 12.1. Therefore, as a conservative approach, this largest Apj (i.e., Ap,,,) will be the pressure drop on the fluid 1 side having imposed flow nonuniformity. The increase in pressure drop due to flow nonuniformity is then

(12.16) It should be emphasized that the entrance and exit losses in addition to the core friction contribution will be higher (in the evaluation of Ap,,,) than those for uniform flow. If flow nonuniformity occurs on both sides of an exchanger, the procedure outlined above is applied to both sides, since the pressure drops on both sides of a two-fluid exchanger are relatively independent of each other, except for the changes in fluid density due to heat transfer in the core. Hence, the analysis above is applicable to any flow arrangement.
Example 12.1 A counterflow heat exchanger has a severe flow maldistribution due to poor header design. On the fluid 1 side, 25% of the total free-flow area has the flow

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

815

velocity 50% larger than the mean flow velocity through the core as a whole. The number of heat transfer units of the heat exchanger is N T U l = 3. The total heat capacity rates through the exchanger are nearly the same (i.e., the heat exchanger is balanced). Determine the reduction in the temperature effectiveness of fluid 1 and an approximate increase in the pressure drop due to flow maldistribution. Assume fully developed laminar flow on both fluid sides (i.e., U remains constant). SOLUTION
Problem Data and Schematic: A schematic of the heat exchanger under consideration is similar to that in Fig. 12.1 with only A and B subexchangers. The following data are known. NTUl = 3 U I , A = 1.5Ul C 1 = C2 A o ,= ~ 0.25A0,1

Determine: The temperature effectiveness PI of the maldistributed fluid Assumptions: All the assumptions listed in Section 3.2.1 are valid with the exception of nonuniformity of the mass flow rate of fluid 1, which is assumed to have a two-step function velocity distribution. Even with this flow maldistribution, U based on A l or A2 is assumed constant. Analysis: The effectiveness of the maldistributed heat exchanger is given by Eq. (12.4) w i t h j = A and B as

Let us determine all the parameters in this equation after determining the area ratios using Eqs. (12.13) and (12.14).

Also,

_ - 1.5
u1

where u1 is the mean fluid velocity on the fluid 1 side. The ratios of the heat capacity rates in the maldistributed subexchangers to the total capacity rate of fluid 1 are then given by Eq. (12.9) as

2 (2)
= -= C2,A
c 2
( A0.A z)2=

1.5 x 0.25 = 0.375

-= 1 - -= 1 - 0.375 = 0.625

CI,B

cl .A
CI

CI

Similarly, the heat capacity rate ratios of subexchangers to the total exchanger on fluid 2 side are given by Eq. (12.10) as

(%) =0.25
Ao,A
1

C2,B c2.A = 1 - -= 1 - 0.25 = 0.75


c 2

c 2

816

FLOW MALDISTRIBUTION AND HEADER DESIGN

We will determine temperature effectivenesses, using Eq. (l2.7), after calculating respective NTUs and R s . The NTUs from Eq. (12.11) are

The heat capacity rate ratios, required for effectiveness calculations, are computed using Eq. (12.12) as follows:

C1.B cI.B R 1 . B = -= --- = 0.625 X C2.B CI C 2 . B c 2

c 2 cl

1 -X I = 0.8333

0.75

Therefore, the temperature effectivenesses of subexchangers are given by Eq. (12.7) as

1 - ,-NTUi
P1.B =

ig(I-Ri

ill

1-R

~ , ~ ~ J -s (N 1-T R)~ H) ~

1 - exp[-3.60( 1 - 0.8333)] = 0.8315 1 - 0.8333exp[-3.60( 1 - 0.8333)]

The temperature effectiveness from Eq. (12.4) is CI .A PI = -P 1 . A CI

+-CI .B P 1 . B = 0.375 x 0.5584 +0.625


CI

x 0.8315 = 0.7291

The heat exchanger effectiveness of a balanced counterflow heat exchanger without any flow maldistribution on either fluid side would be
Pl.ideal =

3 NTU 1 -= 0.750 l+NTUl 1+3

Finally, the quantitative measure of the reduction in the effectiveness due to maldistribution is [see Eq. (12.15)]
Ans.

Discussion and Coninwnts: From the results, it becomes clear that a relatively large flow maldistribution on the fluid 1 side in this particular case causes a deterioration of the temperature effectiveness of approximately 2.8%. With all other parameters fixed, a heat exchanger with high NTU will suffer more pronounced effectiveness deterioration (see Fig. 12.2).

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

817

12.1.1.2 CrossfIow Exchangers. A direct extension of the approach used for a counterflow/parallelflow exchanger to that for a crossflow exchanger with different combinations of fluid mixing/unmixing on each fluid side is not necessarily straightforward. Only when flow nonuniformity is present on the unmixed fluid side with the other fluid side as mixed can a simple closed-form solution be obtained, as outlined next.

Mixed-Unmixed Crossflow Exchanger with Nonuniform Flow on the Unmixed Side. Let us consider a single-pass crossflow exchanger having the unmixed fluid (fluid 1) maldistributed. The inlet velocity distribution is represented with an N-step function (Fig. 12.3). Fluids 1 and 2 can be arbitrarily hot and cold, or vice versa. The total heat transfer rate in the exchanger is given by
N

q=):qj
i=A

(12.17)

where the qj represent individual heat transfer rates/enthalpy rate changes as followst:

(12.18)

( j = A, B, . . . , N) represent the mixed mean temperatures of fluid 2 In Eq. (12.18), TzM,j between the subexchangers. Note that the left-hand side of Eq. (12.17) can also be presented in the form

T2,o
c2

Fluid 2 side

FIGURE 12.3 Idealized two-step function flow nonuniformity on the unmixed fluid 1 side and uniform flow on the mixed fluid 2 side of a crossflow exchanger.

For the sake of simplified notation. the absolute value designators are omitted

818

FLOW MALDISTRIBUTION AND HEADER DESIGN

Our objective is to determine the relationship between the fluid 1 temperature effectiveness PI and the temperature effectiveness and heat capacity rates of the subexchangers of Fig. 12.36. From Eq. (12.19), we get

(12.20)
Replacing q in Eq. (12.20) with q from Eq. (12.17), and utilizing the relationships provided by Eq. (1 2.18), we get

Temperature difference ratios in Eq. (12.21) can be eliminated by manipulating relationships from Eq. (12.18) as follows:

( 1 2.22)

Combining Eqs. (12.22) and (12.21) and after rearrangement, we get

To reemphasize, the fluid 1 side is unmixed and the fluid 2 side is mixed for the expression of a crossflow exchanger above. The temperature effectiveness of the ideal heat exchanger of Fig. 12.3 as a whole, and those of the subexchangers, can be expressed in terms of corresponding heat capacity rate ratios and numbers of transfer units as follows (see Table 3.6):
(12.24)

(12.25)
where
UA NTU 1 -c

NTUl,j=-

UAj
ClJ

j = A , B ,..., N

(12.26)

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

819

When the flow maldistribution on the fluid 1 side is assumed to be a two-step velocity distribution function, Eq. (12.23) can be simplified to (12.27) Calculation of the temperature effectiveness of a maldistributed fluid stream using Eq. (12.23) or (12.27) is valid only if the maldistributed fluid side is the unmixed fluid side and the mixed side has uniform flow. However, if the maldistributed fluid is a mixed fluid and the unmixed fluid side has uniform flow, a closed-form solution cannot be obtained using this simplified approach. This is because T2Mtemperatures for this case will not be uniform, and hence we cannot determine the effectiveness of this exchanger using the formula of Table 3.6 when the flow at the inlet of a subsequent subexchanger is not uniform. The problem becomes inevitably nonlinear, and no closed-form solution is available for this case; only a numerical solution is the option. Note that Eqs. (12.6), (12.9), and (12.1 1) are also valid for this mixed-unmixed crossflow exchanger, while Eqs. (12.10), (12.12), and (12.14) are not valid and we don't need them to determine the temperature effectiveness of this flow maldistributed case.
Unmixed-Unmixed Crossflow Exchangers. The two-dimensional flow maldistribution has been analyzed numerically only for an unmixed-unmixed crossflow exchanger. In a series of publications as summarized by Chiou (1980) and Mueller and Chiou (1988), Chiou has studied the effects of flow maldistribution on an unmixed-unmixed crossflow single-pass heat exchanger with flow maldistribution on one and both fluid sides. When flow maldistribution is present on only one fluid side, the following general conclusions have been obtained.
0

For flow maldistribution on the C,,, fluid side, the exchanger thermal performance deterioration factor A&*approaches a single value of 0.06 for all C* < 1 when NTU approaches zero. The performance deterioration factor decreases as NTU increases. For a balanced heat exchanger (C* = l), the exchanger thermal performance deterioration factor increases continually with NTU. For flow maldistribution on the Cmin fluid side, the thermal performance deterioration factor first increases and then decreases as NTU increases. If flow nonuniformities are present on both sides, the performance deterioration factor can be either larger or smaller than that for the case where flow nonuniformity is present on only one side, and there are no general guidelines about the expected trends.

A study of the influence of two-dimensional nonuniformities in inlet fluid temperatures (Chiou, 1982) indicates that there is a smaller reduction in exchanger effectiveness for the nonuniform inlet temperature than that for the nonuniform inlet mass flow rate. For various nonuniform flow models studied, the inlet nonuniform flow case showed a decrease in effectiveness of up to 20%; whereas for the nonuniform inlet temperature case, a decrease in effectiveness of up to 12% occurred, with even an increase in effectiveness for some cases of nonuniform inlet temperature. This occurs when the hotter portion of the inlet temperature is near the exit end of the cold fluid, whose inlet temperature is uniform. In a recent study, Ranganayakulu et al. (1996) obtained numerical

820

FLOW MALDISTRIBUTION A N D HEADER DESIGN

solutions for the effects of two-dimensional flow nonuniformities on thermal performance and pressure drop in crossflow plate-fin compact heat exchangers. Example 12.2 Analyze a crossflow heat exchanger with fluid 1 unmixed and fluid 2 mixed having pronounced maldistribution on fluid 1 side and N T U l = 3. The total heat capacity rates of the two fluids are nearly the same. Determine the temperature effectiveness of the maldistributed fluid 1 if 25% of the total free-flow area has the flow velocity 50% larger than the mean velocity through the core on the fluid 1 side corresponding to the uniform flow case. SOLUTION
Problem Data and Sclienintic; A schematic of the heat exchanger under consideration is similar to that in Fig. 12.3 with only A and B subexchangers. The following data are known:

Determine; The temperature effectiveness of fluid I .


Assumptions; All the assumptions of Section 3.2.1 are valid here except for flow maldistribution on the fluid 1 side. Anal,vsis; Let us calculate the temperature effectiveness of fluid 1 under the idealized conditions of uniform mass flow rates on both fluid sides for a balanced unmixed-mixed crossflow heat exchanger using Eq. (12.24) as follows:
= 1 - exp[-( 1 - e-NTU1)] = 1 - exp[-( 1 - F 3 )= ] 0.6133

However, under given flow maldistribution conditions, this ideal effectiveness cannot be achieved. Thus, the temperature effectiveness should be calculated using Eq. (12.27). For this example, similar to Example 12. I , we can determine
-= 0.375 CI

cl A

C1.B -= 0.625 CI

NTUi,A = 2.00

NTUI,B = 3.60

Now PI,Aand PI,^ are computed using Eq. (12.25) as follows after incorporating

c z = c,:

=-{I 0.375

-exp[-0.375(1

-e-'"')]}

=0.7385

={ 1 - exp [-0.625( 1 - e-3.60)]} = 0.7288

0.625

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

821

Now, compute the temperature effectiveness of fluid 1 given by Eq. (12.27) as (with CI = C2)

= 0.7385 x 0.375

+ 0.7288 x 0.625 x (1 - 0.7385 x 0.375) = 0.6063

Ans.

This actual effectiveness is, indeed, smaller than the one calculated for an idealized situation, 0.6063 vs. 0.6133. Finally, the fractional deterioration in the temperature effectiveness is given by Eq. (12.15) as

AP* = Pl,ideal .ideal

- 0.6133 - 0.6063 = 0.0114

0.6133

Discussion and Comments: Deterioration in the temperature effectiveness caused by a relatively large flow maldistribution for this crossflow exchanger is 0.01 14, much smaller than 0.0279 for the counterflow exchanger (see Example 12.1), for the same operating conditions. The results of Examples 12.1 and 12.2 emphasize the fact that flow maldistribution has the highest effect on a counterflow exchanger (since it has the highest E for given N T U and C*) compared to exchangers with other flow arrangements for similar operating parameters.
12.1.1.3 Tube-Side Maldistribution and Other Heat Exchanger Types. Tube-side maldistribution in a 1-1 TEMA E shell-and-tube counterflow heat exchanger studied by Cichelli and Boucher (1956) led to the following major conclusions:
0

0 0

For C,/C, small, say C,/C, = 0.1, the performance loss is negligible for large flow nonuniformities for NTU, < 2. For C,/C, large, say C,/C, > 1, a loss can be noticed but diminishes for NTU,? > 2. C,7/C,= 1 is the worst case at large NTU,7as can be found from Fig. 12.2.

Fleming (1966) and Chowdhury and Sarangi (1985) have studied various models of flow maldistribution on the tube side of a counterflow shell-and-tube heat exchanger. It is concluded that high-NTU heat exchangers are more susceptible to maldistribution effects. According to Mueller (1977), the well-baffled 1-1 counterflow shell-and-tube heat exchanger (tube side nonuniform, shell side mixed) is affected the least by flow maldistribution. Shell-and-tube heat exchangers, which d o not have mixing of the uniform fluid [( 1) tube side nonuniform, shell side unmixed; or (2) tube side uniform, shell side nonuniform in crossflow], are affected more by flow maldistribution. According to Kutchey and Julien (1974), the radial flow variations of the mismatched air side and gas side reduce the regenerator effectiveness significantly. 12.1.2 Passage-to-Passage Flow Maldistribution Compact heat exchangers with uninterrupted (continuous) flow passages, while designed for nonfouling applications, are highly susceptible to passage-to-passage flow maldistribution. That is because the neighboring passages are geometrically never identical, due to imperfect manufacturing processes. It is especially difficult to control the passage size

822

FLOW MALDISTRIBUTION AND HEADER DESIGN

precisely when small dimensions are involved [e.g., a rotary regenerator with Dh = 0.5mm (0.020 in.)]. Since differently sized and shaped passages exhibit different flow resistances and the flow seeks a path of least resistance, a nonuniform flow through the matrix results. This phenomenon usually causes a slight reduction in pressure drop, while the reduction in heat transfer rate may be significant compared to that for nominal (average) size passages. The influence is of particular importance for continuous-flow passages at low Re (i.e., laminar flow) as found in compact rotary regenerators. For a theoretical analysis for passage-to-passage flow maldistribution, the actual nonuniform surface is idealized as containing large, small, and/or in-between size passages (in parallel) relative to the nominal passage dimensions. The models include (1) a two-passage model (London, 1970), (2) a three-passage model, and (3) an N-passage model (Shah and London, 1980). Although triangular and rectangular passage cross sections have been studied, similar analysis can be applied to any cross-sectional shapes of flow passages. The analysis to follow can also be utilized for analyzing flow maldistribution in viscous oil cooler with constant-wall-temperature boundary conditions (i.e., condensation or vaporization taking place on the other fluid side). See Section 12.2.1 for further details. Let us first define the two-passage-model flow nonuniformity. From the methodological point of view, this approach is the most transparent and offers a clear idea of how the modeling of flow nonuniformity can be conducted. Also, the two-passage model predicts a more detrimental effect on heat transfer and pressure drop than that of an N passage ( N > 2) model.
12.1.2.1

Models of Flow Nonuniformity

Two-Passage Model. Let us consider that a heat exchanger core characterizes flow nonuniformity due to two different flow cross sections differing in either ( I ) cross-section size of the same passage type, (2) different cross-sectional shapes of flow passages, or (3) a combination of both. The two most common types of idealized passage-to-passage nonuniformities are plate spacing and fin spacing, shown in Fig. 12.4a and b, respectively. For the analysis, the actual heat exchanger core will be assumed to be a collection of two (or more) distinct sets of uniform flow passages, passages 1 and passages 2
Geometry Idealized Nominal nonuniform uniform

Large passage

FIGURE 12.4 Two-passage nonuniformity model: (a) plate-spacing nonuniformity; (b) finspacing nonuniformity. Note that passages differ in size. The nominal size of the passage may be large, small, or in between, depending on how it is defined. (From London, 1968.)

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

823

(or N passages). Our objective here is to determine the reduction in heat transfer and pressure drop due to this passage-to-passage flow nonuniformity. The following assumptions are invoked for setting up the model.
0

Flow is hydrodynamically and thermally fully developed (Nu = constant,

f.Re = constant).
0 0

Thermophysical properties of the fluids are constant and uniform. Entrance and exit pressure losses are negligible (the core friction component is dominant). Static pressures are constant and uniform across the cross section at the entrance and exit of this multipassage exchanger. The total flow rate through all nonuniform flow passages is identical to that going through all nominal flow passages. The lengths of all flow passages are the same.

The pressure drop for all flow passages (regardless of the size, shape, and distribution of flow passages) will be the same in the core based on the fourth assumption above:

CAP), = (Pi where

~ o ) j

(12.28)

(12.29) wherej denotes the flow passage type. Invoking the definitions of the Reynolds number and mass flow rate, Eq. (12.29) is regrouped as (12.30) For a two-passage model, applying Eq. (12.30) for j = 1 and 2 and taking the ratio and rearranging, we get (12.31) since Apl = Ap2 from Eq. (12.28). Equation (12.31) provides the flow fraction distribu, , hydraulic diation in the two types of flow passages. Normalizing flow rates with h meters with Dh,n,and free-flow areas with A,,,, Eq. (12.31) becomes (12.32) where h, = hl h2and all variables with a subscript n denote nominal values (selected by the choice of an analyst), either the passage geometry 1, the passage geometry 2, or some nominal passage geometry in between (for normalization of Dh and A , used in the equation) for a two-passage nonuniformity.

824

FLOW MALDISTRIBUTION A N D HEADER DESIGN

To compute the flow area ratios in Eq. (12.32), we maintain approximately the same frontal area of the heat exchanger core with actual and nominal flow passages. There are two choices for selecting the nominal passage geometry, and accordingly, the values of A,,,,/Al1,, ( j = I . 2) will be different. They are as follows.

1. The number of flow passages for the nominal geometry is the sum of the number of flow passages for passage types 1 and 2 and the frontal area is the same. In this case,
(12.33)

A,,,2,and are the flow area for one passage of passage types 1, 2, where and 17, respectively, and x1and x2 are the corresponding fractions of the number of passages of types 1 and 2.' This case applies when comparing sharp and rounded triangular (or any two similar passages), where the frontal area remains constant for the same total number of flow passages, regardless of which is the nominal flow passage. However, the free-flow area will be different for the nominal flow passages since the flow areas of sharp and rounded corner passages are different (see Example 12.3). 2 . In the alternative case. the total number of flow passages for the nominal passages could be different from the actual number of flow passages for the same frontal area. This case applies when we compare the two-passage model (e.g., large and small rectangular or triangular passages with 50% : 50% or any other percent distribution) with the nominal passage geometry having approximately the same frontal area.: In this case, the number of flow passages for the nominal passage geometry will be different from the sum of the number of flow passages of passage types 1 and 2. The flow area ratios A,,,,/A,,,, and A(,,2/Al,,f,are given by
(12.34) where the definitions of I , if, 2, and x 2 are the same as defined above after Eq. (12.33). Note that we may use A(,, in Eq. (12.34) instead of j = I or 2, since the fraction is primarily known.

A,,,,

A,,,,,

A,

,/A,,,,

A ,,,

The pressure drop ratio (the ratio of the pressure drop for either of the two passage types, either 1 or 2, to the nominal passage pressure drop) can be calculated, using Eq. (12.30), as (12.35)

to the total number of passages. If more than two passages are involved, the following relation holds: C x, = I . Also note that ( A ( , , ,+ A o , 2 ) / A , , I# I in general. This is because we have presumed the same frontal area, and as a result, the wall thickness is different for differently shaped passages (see Example 12.3). The number of passages must be an integer, so the frontal area for a particular selection of passages may not necessarily be the same when compared to a two-passage model with a nominal passage model. However, in a compact heat exchanger with a very large number of flow passages. the difference will be negligible.

'Note that xI = 1 - x?.The parameter y , by definition. represents a ratio of the number of the ith shaped passage

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

825

Note that the flow area of two nominal passages is the same as the two (large and small) passages of the nonuniform core (see Fig. 12.4 for two examples). As we know, since the fluid seeks the path of least flow resistance, if we replace some nominal passages with different flow passages having larger and smaller flow areas, a larger fraction of the flow will go through the larger flow area passages. Then for a constant flow rate, the pressure drop (and hence heat transfer) will reduce for this exchanger with mixed passages.+ This means that A p , / A p , (= Ap2/Ap,) will be less than unity. This gain (reduction) in the pressure drop due to passage-to-passage nonuniformity is (12.36) Let us now determine a change in heat exchanger effectiveness due to passage-topassage nonuniformity. Heat transfer through differently shaped passages would be different, which in turn would produce different temperature differences between fluids 1 and 2. Hence, one cannot consider different passages, let us say two passages A and B, in parallel to arrive at an effective h as the average of conductances hA and h g . T o arrive properly at an effective value of h for a two-passage geometry heat exchanger, the passage geometrical properties, fluid physical properties, exchanger flow arrangement, and E-NTU relationship must be considered. A procedure is outlined in the following subsections for the two most important cases of a two-passage geometry for a counterflow exchanger with C* = 1 (a rotary regenerator case) and an exchanger with C* = 0 (an oil cooler case with constant wall temperature). Refer to Shah and London (1980) for an analysis of other flow arrangements. For both these cases, the heat transfer results are presented in terms of the number of transfer units ntuj for each type of passage on the maldistributed fluid side as follows:
/hA\

/ Nu 4L\

/ 4 k L NuA,\

. -

For the nominal passage, define ntu, by Eq. (12.37) with j = n. The normalized ntuj/ntu,, based on Eq. (12.37), is (12.38) Nuj and Nu, in this equation should be obtained from the results of Table 7.3 for the appropriate thermal boundary conditions for fully developed laminar flow. Note that for a counterflow exchanger with C* = 1, the boundary conditions are @ or @, while the boundary condition is @ for the C* = 0 case.
COUNTERFLOW HEAT EXCHANGER WITH

C* = 1. In this case,

E,

and ntu, are related as (12.39)

follows using Eq. (3.85):

J=iTqj =

ntuj

1,2,n

'This is what we have shown through Eqs. ( 1 2 . 2 S t ( 12.36) that the performance (y and Ap) of a continuous-flowpassage regenerator matrix will be lower when there are small and large passages in parallel compared to the performance of nominal (average)-size passages of the same shape.

826

FLOW MALDISTRIBUTION AND HEADER DESIGN

wherej depends on whether the heat exchanger unit considered has all uniform (nominal) passages, E, ( j = n ) , or it refers to a maldistributed heat exchanger that consists of two subexchangers (with the effectivenesses or c2 for the passage geometriesj = 1 or j = 2, respectively). Note that since ntuj is defined using a heat transfer coefficient (not the overall heat transfer coefficient U as in NTU), the heat exchanger effectiveness of Eq. (12.39) must be defined based on the passage wall temperature:
E J - . i F- O .l-p
I).

T . - T,
-

Ti 1 + ntuj

ntuj

j = 1,2,n

(12.40)

where T,, represents the mean wall temperature of the heat transfer surface, Ti is the inlet temperature of fluids in both subexchangers and nominal exchanger, and the distribution of T,,,jvs. x are parallel to the distribution of T/ vs. x as shown in Fig. 1 of Shah and is assumed to be the same for both passage geomeLondon (1980). The temperature TIC tries at the inlet (thus leading to the same inlet temperature difference for both passage types). The ntuj for Eq. (12.40) are computed from Eq. (12.38) for a specified value of ntu,, and known flow fraction distribution from Eq. (12.32). The average effectiveness of the maldistributed heat exchanger can be calculated from the effectiveness of two subexchangers using a simple energy balance and assuming constant specific heat of the fluids as follows:
rilEa"e = h,l

+ lj2*E2

(12.41)

It must be emphasized that the analysis presented here is for one fluid side of the exchanger (either the hot- or cold-fluid side of a rotary regenerator). To find the resultant effect on the exchanger performance, the effect of the other fluid side needs to be taken into account, as will be shown in Example 12.3. The effective ntu on one fluid side is then given by
( 12.42)

The "cost" of the influence of passage-to-passage nonuniformity on ntu is defined as ntueff ntu,,,, = 1 - ntu, (12.43)

We need to compute ntuefffor both fluid sides and subsequently calculate NTUeff for the exchanger to determine a reduction in the exchanger effectiveness due to passage-topassage nonuniformity, as shown for a specific exchanger in Example 12.3. London (1970) determined ntucostand Apgainfor plate-spacing and fin-spacing type nonuniformities and concluded that the deviation in passage size causes a more severe reduction in the number of transfer units than does the pressure drop gain. Specific results from the two-passage model for the passage-to-passage nonuniformity are presented in Fig. 12.5 for rectangular passages. This two-passage model consists of 50% of the flow passages large (c2 > c,) and 50% being small (cl < c,) compared to the nominal passages, and the nominal aspect ratios a,* = I , 0.5, 0.25, and 0.125. In Fig. 12.51, a reduction in ntu is presented for the @ and @boundary conditions and for a

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

827

47.5 45.0 40.0

30.0 ,f
2 ?

35.0

.C C C

5 25.0

g
C

20.0
15.0

8
10.0
5.0

0.0
Channel deviation parameter

8,

(a)

17

.6 15 rm
C .C

3
10e!

2 0
-0
L

a C

5-

000

005

01 0 015 020 Channel deviation parameter

025

030

6 ,

(b) FIGURE 12.5 Deterioration factors for two-passage nonuniformities in rectangular passages: (a) percentage loss in ntu as a function of 6 , , a,*, and thermal boundary conditions; (b) percentage reduction in Ap as a function of 6, and a,*.(From Shah, 1981.)

828

FLOW MALDISTRIBUTION A N D HEADER DESIGN

nominal (design) ntu,, of 5.0. Here, ntu,,,,, a percentage loss in ntu, and the channel deviation parameter 6,. are defined as

where ntueff is the effective ntu for the two-passage model passage-to-passage nonuniformity, and ntu,, is the ntu for nominal (or reference) passages. It can be seen from Fig. 12.50 that a 10% channel deviation (hC = 0.10, which is common for a highly compact surface) results in 10 and 21% reduction in ntuHl and ntuT, respectively, for a: = 0.125 and ntu,, = 5.0. In contrast, a gain in the pressure drop due to the passage-topassage nonuniformity is only 2.5% for 6,. = 0.10 and a,: = 0.125, as found from Fig. 12.91. Here ApEai, is defined as
( 12.45)

The following observation may be made from Fig. 12.5~1 and additional results presented by Shah and London (1980): ( I ) the loss in ntu is more significant for the @ boundary condition than for the @ boundary condition; (2) the loss in ntu increases with higher values of nominal ntu; and (3) the loss in ntu is much more significant than the gain in Ap at a given 6,. N-Passuge Model. The previous analysis was extended for an N-passage model by Shah and London (1980). In the N-passage model, there are N different-size passages of the same basic shape, either rectangular or triangular. The results of Fig. 1 2 . 5 and ~ h for rectangular passages are also applicable to an N-passage model in which there are N different-size passages in a normal distribution about the nominal passage size with a proper definition of the channel deviation parameter 6, as follows:

( 12.46) Here xi is the fractional distribution of the ith shaped passage. For N = 2 and x, = 0.5, Eq. (12.46) reduces to Eq. (12.44) for 6,. Similar results are summarized in Fig. 12.6 for the N-passage nonuniformity model associated with equilateral triangular passages. in this case, the definition of the channel deviation parameter 6,is modified to

6,. =

5I ? ) : i= 1

If?

( 12.47)

where r,,,,, is the hydraulic radius of the nominal passages, r/,,i is the hydraulic radius of the ith passage, and they are related for a two-passage model as follows: 2r/5,,,= r5,] + &, but this particular case corresponds to an equilateral triangular passage. Qualitative trends of the results in Fig. 12.6 are similar to those in Fig. 12.5 for rectangular flow passages.

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

829

45 c

ntu,
40 C

35 c

C .c 30C 0 .3 V '0

a
2

: I 3 C

25C

C .-

20c

0
C

a 2

150

100

50

0.0 0.00

0.05

0.10

0.1 5

0.20

0.25

0.30

Channel deviation parameter

6,

FIGURE 12.6 Percentage loss in ntu and percentage reduction in Ap as functions of 6, for N passage nonuniformities in equilateral triangular passages. (From Shah, 1985.)

Note that the percentage reduction in ntu and Ap vs. 6,. curves for N = 2 and N > 2 are identical (as shown in Figs. 12.5 and 12.6), except that the value of 6,. is higher for a two-passage model compared to the N-passage model for the same value of rmax/c,,. Hence, the two-passage model provides the highest deterioration in performance.
HEAT EXCHANGER W I T H

C* = 0. In this case,

E~

and ntu, are related as follows using

Eq. (3.84):
J

830

FLOW MALDISTRIBUTION AND HEADER DESIGN

wherej depends on whether the heat exchanger unit considered has all nominal passages, E, ( j = n), or it refers to a maldistributed heat exchanger that consists of two subexchangers (with the effectivenesses E ] or E~ for the passage geometries j = 1 or j = 2, respectively). The average effectiveness of the passage-to-passage maldistributed heat exchanger can be calculated using Eq. (12.41). Similar to the previous case, the cost of the influence of passage-to-passage nonuniformity on ntu is defined as follows: ntuefl ntu,,,, = 1 - ntu, where (12.50) Refer to Shah and London (1980) for further details.
Example 12.3 A vehicular gas turbine counterflow rotary regenerator is made up of triangular flow passages. Due to brazing of the core, some of the flow passages became triangles with rounded corners. Hence, idealize that the matrix is made up of 50% passages having all three corners rounded and 50% passages having all three sharp corners. Determine: ( 12.49)

(a) The flow fraction distribution in the two types of passages (b) The change in pressure drop due to passage-to-passage nonuniformity. Does it represent a loss or a gain in comparison to all passages of ideal sharp corners? (c) The change in the exchanger effectiveness due to the nonuniformity. Does it represent a loss or a gain? (d) The subsequent change in ntu The additional data are as follows:
Sharp Corner Passage
13.333 3.111
1

Characteristic f.Re
NUH I
A0I~o.A

Rounded Corner Passage


15.993 4.205

Dh/Dh.A

0.868 1.125

For this regenerator, the flow split gas :air = 50% : 50%. NTU,, = 2.5. Idealize negligible wall resistance, C* = 1 and C : -+ cm. SOLUTION
Problem Data a i d Scheivatic: The passage-to-passage flow nonuniformity is caused by differing flow resistance of sharp and rounded corner flow passages shown in Fig. E12.3. All data for thermal and hydraulic characteristics of the flow passages are provided in the table above, and xI = x2 = 0.5, NTU,, = 2.5, and C* = I .

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

831

50% 0-rounded corners 50% 3-rounded corners

FIGURE E12.3

Determine: The influence of passage-to-passage flow nonuniformity on heat transfer and pressure drop.
Assumptions: The pressure drop is uniform across all heat exchanger flow passages. All the assumptions of Section 3.2.1 are valid except for the nonuniform flow through the regenerator due to differently shaped flow passages.

Analysis: The flow fraction distribution can be determined from the assumption of uniform pressure drop distribution across the heat exchanger core. We consider the ideal sharp corner flow passage as a nominal passage for this example. Therefore, using Eq. (l2.33),
Ao.l _
'0.n '%,sharp A

-XI--

- 0.5 x 1 = 0.5

-- x2 Ao,round A = 0.5 x 0.868 = 0.434


AO,n

2
Ao,sharp A

&,sharp A

The ratio of mass flow rates through these passages, using Eq. (12.31), is given by

From Eq. (1) we get


hl =
h l b 2

riz,

I+mjll/hZ

1.092 1+1.092

ml =0.522

and

h 2 -=0.478

Ans.

mn

mn

The ratio of the pressure drop for the sharp corner triangular passages to the nominal pressure drop can be determined as follows using Eq. (12.35):

Ans.

Similarly, ratio of the pressure drop within the rounded triangular passages to the pressure drop through the nominal passages should, ideally, be 1.044. We could have calculated this ratio using an analogous relationship to the one given by Eq. (12.35) by replacing the subscript 1 with 2.

832

FLOW MALDISTRIBUTION A N D HEADER DESIGN

The heat exchanger effectiveness due to the nonuniformity depends on the number of transfer units for the respective passages. The number of transfer units for sharp and rounded corner passages, normalized with respect to nominal passages, are determined using Eq. (12.38) as

ntu,,

Nu,,

*,

where j = 1 or 2 (i.e., sharp-corner triangular passages or rounded-corner triangular passages, respectively). Utilizing the given and calculated data, we get

For the given NTU,, = 2.5, we obtain ntu,, = ntq, = ntu,. = 5.0 for C* = 1 [see, for example, Eq. (9.23)]. With ntu,, = 5, we get ntul = 5 x 0.9579 = 4.7893 ntu2 = 5 x 0.9697 = 4.8483

Consequently, the heat exchanger effectivenesses for the two types of passages for C* = 1 would be
E,

+ ntu,

ntu,

j = 1,2
E~

Therefore, for sharp-corner triangular passages and passages are

for rounded-corner triangular

The average heat exchanger effectiveness can be calculated from a simple energy balance, using Eq. (l2.41), as
E,,

m1 = -1

m,,

+ m2 - 2 = 0.522 x 0.8277 + 0.478 x 0.8290 = 0.8283 m,


Eave

Therefore, ntu,f from Eq. (1 2.43) is ntu,ff = 1 - E,,


-

0.8283
1 - 0.8283

= 4.8241 = ntueff,/, = ntueff,<

Next Page

GEOMETRY-INDUCED FLOW MALDISTRIBUTION

833

The effective NTU for this regenerator, using Eq. (5.54), is given by NTUeu =
1 1 = 2.412 l/ntueff,h l/ntu,u,c 1/4.8241 + 1/4.8241

Subsequently, the regenerator effectiveness with the passage-to-passage flow nonuniformity is given by
Eeff

NTUefi - 2.412 -= 0.7069 1 + NTUen 1 + 2.412

In contrast, the effectiveness of the heat exchanger with nominal uniform passages is
E n

NTU,

- l +NTU,

2.5 l f2.5

Thus, the loss in the regenerator effectiveness is


AEloss

En
~

- Eave
En

x loo=

0.7143 - 0.7069 0.7069

Ans.

Discussion and Comments: It should be noted that in pressure drop analysis, sharp-corner passages are considered as nominal passages. Thus, the pressure drop of the matrix with nonuniform flow passages is 4.4% larger than that for the ideal matrix with uniform sharp corners. This is because the sharp-corner triangular passage (used for the comparison) has a lowerf Re than that for the rounded-corner triangular passage. In contrast, there is a reduction (1.0%) in regenerator effectiveness due to nonuniform flow because of the poor performance of the rounded-corner flow passages. The comparison was performed by comparing the performance of nonuniform passages with nominal sharp-corner passages which have a lower heat transfer coefficient and a lower friction factor. If we would have considered the rounded-corner triangular passages as nominal passages, there would have been a reduction in the pressure drop and a slight gain in heat exchanger effectiveness. 12.1.2.2 Passage-to-Passage Flow Nonuniformity Due to Other Effects. Finally, passage-to-passage flow nonuniformity for very compact surfaces may be induced by brazing and/or fouling in addition to manufacturing imperfection. Both controlled atmosphere brazing and vacuum brazing have a negligible effect on j and f data if the plates/tubes/primary surface is clad and fins are unclad, and the ratio of the joint area to free-flow area is less than 10%. For ultracompact surfaces/flow passages, this ratio may not be small (i.e., flow area blockage and brazing-induced surface roughness may not be negligible, and accurate experimentalj andfdata are essential in this case). Gross blockage due to brazing may increase the pressure drop substantially. The influence of surface roughness induced by salt dip brazing (currently an outdated technology due to environmental concerns) is generally nonnegligibk (i.e., can increase Ap considerably with only a slight increase in h or j ) in highly compact surfaces (Shah and London, 1971). Controlled atmosphere brazing, a state-of-the-art manufacturing process for compact heat exchangers (Sekulik, 1999), provides a very uniform flow passage

Potrebbero piacerti anche