State of The Art Review of Erosion Modeling in Fluid Solid Systems PDF

Progress in Energy and Combustion Science 28 (2002) 543–602
www.elsevier.com/locate/pecs
State-of-the-art review of erosion modeling in fluid/solids systems

Robert W. Lyczkowskia,*, Jacques X. Bouillardb
a
Argonne National Laboratory, Energy Systems Division, Building 362, Room C348D, 9700 South Cass Avenue, Argonne, IL 60439-4815, USA
b
INERIS, Parc Technolgique ALATA, BP2, 60550 Verneuil en Halate, France
Received 11 June 2001; accepted 5 September 2002
Abstract
Erosion in fluidized-bed combustors, commercial process units used to burn coal cleanly, has surfaced as a serious issue that
may have adverse economic effects. The evidence suggests that the key to understanding this erosion is detailed knowledge of
the coupled and complex phenomena of solids circulation and bubble motion. The FLUFIX computer code has been developed
for this purpose. Computed hydrodynamic results compare well with experimental data (including the bubble frequency and
size and the time-averaged porosity and pressure distributions) taken in a thin ‘two-dimensional’ rectangular fluidized beds
containing a rectangular obstacle and a few-tube approximation of the International Energy Agency Grimethorpe tube bank
‘C1’ configuration.
Six representative erosion models selected from the literature, comprising both single-particle and fluidized-bed models are
critiqued. A methodology is described whereby the computed hydrodynamic results can be used with such erosion models.
Previous attempts (none involving fluidized beds) to couple fluid mechanics and erosion models are reviewed. The energy
dissipation models are developed, and are shown to generalize the so-called power dissipation model used to analyze slurry jet
pump erosion. It is demonstrated, by explicitly introducing the force of the particle on the eroding material surface, that
impaction and abrasive erosion mechanisms are basically the same. In doing so, it has been possible to unify the entire erosion
literature developed for over a century.
Linkage is made to two previously developed single-particle erosion models: Finnie’s and Neilson and Gilchrist’s. The
implementation methodology, which can be applied to any erosion model, be it single-particle or fluidized bed, is summarized.
The monolayer energy dissipation (MED) erosion model is developed. The erosion rates computed from the EROSION code
are compared with each other and for the cold few-tube approximation of the IEA Grimethorpe tube bank ‘C1’ fluidized-bed
experiment, and with other available erosion data literature to validate the calculations. The simplified closed form MED
(SCFMED) erosion models and erosion guidelines are developed using semi-empirical correlations in order to allow quick
engineering estimates of erosion. Alternative methodologies to couple hydrodynamics and erosion using the kinetic theory of
granular flow and discrete element method (DEM) models are briefly reviewed.
Finally, a critical review of the integrated experimental and computational fluid dynamics (CFD) pressurized fluidized-bed
hydrodynamics and erosion research ongoing at Chalmers University is presented. This body of work has been influenced by the
research at Argonne National Laboratory (ANL) and Illinois Institute of Technology (IIT) and reinforces the trends and
conclusions reported in this review. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Fluidized beds; Hydrodynamics; Erosion; Computational fluid dynamics; Kinetic theory; Discrete element method
Contents
1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
1.1. Fluidized-bed combustors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
1.2. The erosion issue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
* Corresponding author. Tel.: þ 1-630-252-5923; fax: þ1-630-252-9281.

E-mail address: rlyczkowski@anl.gov (R.W. Lyczkowski).
0360-1285/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 6 0 - 1 2 8 5 ( 0 2 ) 0 0 0 2 2 - 9
544 R.W. Lyczkowski, J.X. Bouillard / Progress in Energy and Combustion Science 28 (2002) 543–602
2. Modeling of fluidized-bed solids motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548

2.1. Hydrodynamic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 549
2.2. Representative computational results and comparisons with data . . . . . . . . . . . . . . . . . . . . . . . . 550
3. Overview of erosion modeling and mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 556
3.1. Single-particle erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
3.1.1. Finnie’s ductile erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
3.1.2. Bitter’s combined ductile and brittle erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
3.1.3. Neilson and Gilchrist’s combined ductile and brittle erosion model . . . . . . . . . . . . . . . . 559
3.1.4. Sheldon and Finnie’s 908 brittle erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 560
3.2. Fluidized-bed erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 560
3.2.1. Soo’s ductile and brittle erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 560
3.2.2. Wood and Woodford’s fatigue erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
4. Previous efforts to couple fluid mechanics and erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
4.1. Erosion in turbomachinery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 562
4.2. Erosion in horizontal tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
4.3. Erosion in curved elbows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
5. Power and energy dissipation erosion models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
5.1. Power dissipation erosion model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
5.1.1. Motivation and origins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564
5.1.2. Linkage to algebraic erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
5.1.3. Extension to account for particle size and threshold energy dependence . . . . . . . . . . . . . 566
5.1.4. Extension to two dimensions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
5.2. Energy dissipation models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569
5.2.1. Derivation of the general transient energy dissipation model . . . . . . . . . . . . . . . . . . . . . 569
5.2.2. Coupling of hydrodynamic and energy dissipation models . . . . . . . . . . . . . . . . . . . . . . . 570
6. Implementation of erosion models in EROSION/MOD1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
6.1. Monolayer energy dissipation erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
6.2. Neilson– Gilchrist and Finnie erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572
7. Comparison of erosion predictions for two cold few-tube FBC geometries . . . . . . . . . . . . . . . . . . . . . 574
7.1. Hydrodynamic results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575
7.2. Bed dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575
7.3. Erosion predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
7.4. Simplified closed form MED erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
7.5. Erosion guidelines using semi-empirical erosion correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . 587
7.6. Simplified mechanistic erosion guidelines procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
8. Kinetic theory and discrete element method erosion models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
8.1. Kinetic theory erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
8.1.1. Three-dimensional computer model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
8.1.2. Computed erosion results and comparison with data. . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
8.2. Discrete element method erosion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
9. Summary of the Chalmers University of Technology research program. . . . . . . . . . . . . . . . . . . . . . . . 594
9.1. Fluidized-bed hydrodynamics experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595
9.2. Erosion experiments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595
9.3. Fluidized-bed hydrodynamics CFD simulations and validation . . . . . . . . . . . . . . . . . . . . . . . . . . 596
9.4. Erosion simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 597
10. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 597
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 598
1. Background terms which came into use in the petroleum industry in

1940s during the development of fluid catalytic crackers
1.1. Fluidized-bed combustors (FCCs) [1] and have been virtually synonymous with
operation in the bubbling regime, corresponding to
Fluidized-bed combustion (FBC) is an established fluidizing gas velocities in the order of 1 m/s. As the
means of burning high-sulfur coal in an environmentally gas is increased through a fixed or packed bed, the
acceptable manner. Fluidization and fluidized beds are pressure gradient across the bed rises sharply until
R.W. Lyczkowski, J.X. Bouillard / Progress in Energy and Combustion Science 28 (2002) 543–602 545
Nomenclature K1 constant given by Eq. (3.10b), (m/s)1/2

kp, kt elastic constants defined by Eq.
A area, m2
(3.7b) ¼ ð1 2 g2p Þ=ðpEp Þ and ð1 2 g2t Þ=ðpEt Þ;
Ao area of the distributor per orifice, m2
respectively, Pa21
Ar Archimedes number given by Eq. (7.11)
Ls bed depth before fluidization (static bed height)
BF defined by Eq. (8.2)
in Eq. (7.15), m
C fluctuating particle velocity, m/s
M mass of abrasive particles, kg
C c=ðckÞ; factor in the power and energy dissipa-
M mass of a particle ¼ pdp3 rp =3; kg
tion models
m_s mass flux of solids ¼ 1s rs l~vs l; kg/(m2 s)
C0 constant given by Eq. (3.10a) (s5/2 m1/2)/kg
Nf number of cycles to failure
Cb, Cd correction factors for non-sphericity in Eqs.
N_ impaction frequency, Hz
(3.16) and (3.19)
NIM impact number given by Eq. (3.17b)
Ce erosion constant defined by Eq. (3.15c)
Npw collision frequency of particle with a stationary
c instantaneous particle velocity, m/s
surface, Hz/m2
c compaction modulus; non-dimensional constant
n outward drawn normal from a stationary surface
in Finnie’s erosion model, Eqs. (3.1a) and
P pressure, Pa; tube inclination in Eq. (7.15);
(3.1b), to allow for non-ideal model particle
power dissipation defined by Eq. (5.7), W/m2
behavior ð0 , c # 1Þ
p eroding surface flow stress related to hardness,
cp particle concentration, kg/m3
Pa
D tube diameter in Eq. (7.15), mm
q~s particle mass flux ¼ ð1 2 1Þrs ~vs D; kg/(m s)
de equivalent sphere bubble diameter
Re fluidization Reynolds number given by Eq.
dp particle diameter, m
(7.10)
dKEs/dt energy dissipation rate, W/m3
Rp particle radius, m
E rate of energy dissipation per unit volume,
r space vector, m
W/m3
_E rp ratio of rebound to approach velocity, V2/V1
erosion rate, m/s
S specific erosion rate given by Eq. (3.27),
E_ b ; E_ d brittle and ductile erosion rates, m/s
dimensionless
E_ EDCF erosion rate given by Eq. (7.1), m/s
Sp Moh’s scale hardness of particles, Pa
E_ FM maximum erosion rate from Finnie’s model
St Brinell hardness of target, Pa
given by Eq. (3.5), m/s
T fluctuating kinetic energy, usually referred to as
E_ 0 erosion rate group given by Eq. (7.2), m/s
the granular temperature, m2/s2
Ep, Et Young’s modulus of particle and target,
t time, s
respectively, Pa
U energy, J in Eqs. (5.1) and (5.3a); fluidizing
Er reduced Young’s modulus of elasticity given by
velocity ¼ 1Vg, m/s; energy dissipation rate,
Eq. (3.7b), Pa
W/m2
Esp specific energy of eroding material (related to
Ug, Us gas and solid phase velocities, respectively, in
hardness), Pa
the x direction, m/s
Ewp kinetic energy of the wake particles given by
Up velocity of abrasive particles, m/s
Eq. (7.6)
uv void (bubble or slug) velocity ,U, m/s
e coefficient of restitution, ratio of particle
V velocity, m/s; volume, m3
rebound velocity, Vp2, and particle approach
Vc Crater volume, m3
velocity, Vp1, striking a solid surface
Vg, Vs gas- and solids-phase velocities, respectively, in
F force, N
the y direction, m/s
Fr Froude number given by Eq. (7.12)
Vel threshold velocity defined by Eq. (3.8), m/s
f coefficient of friction; single particle velocity
Vp threshold velocity defined by Eq. (3.12)
distribution function
Vr relative velocity ¼ vf 2 vp, m/s
fv void frequency, Hz
V1, V2 approach and rebound velocities, respectively,
fw fraction of bubbles occupied by wakes
m/s
G solids elastic modulus, Pa
~v velocity vector, m/s
g acceleration due to gravity, m/s2
l~vl magnitude of velocity vector, m/s
H hardness, Pa; fluidized-bed height, m
W volume of eroding surface, removed, m3
K ratio of vertical to horizontal forces in Finnie’s
Wb volume of eroding surface removed by defor-
erosion model, Eqs. (3.1a) and (3.1b)
mation (brittle) erosion, m3
Kdp dimensionless erosion resistance given by Eq.
Wd volume of eroding surface removed by cutting
(3.17a)
(ductile) erosion, m3
x lateral coordinate, m sf tensile strength at failure, Pa

xd characteristic acceleration distance, m sy elastic load limit, Pa
y axial coordinate, m t solids stress, related to particle-to-particle
pressure, Pa
Greek letters t SV solids viscous stress, Pa
a impingement angle, degrees fc cutting wear factor, Pa
a0 transition impingement angle, degrees fs particle sphericity
bx ; by fluid particle friction coefficient in the x- and C ratio of depth of contact to depth of cut in
y-directions, respectively, kg/(m3 s) Finnie’s erosion model, Eqs. (3.1a) and (3.1b)
gp, gt Poisson’s ratio of particle and target, respect-
ively Subscripts
D layer thickness ¼ (Vf/At) b brittle
d viscous boundary layer thickness; thickness of d ductile
layer removed, m f fluid
1 void fraction or porosity (fluid-phase volume ED energy dissipation
fraction); wave voidage in Eq. (7.6) F Finnie
1b deformation wear factor, Pa FM Finnie maximum
1f tensile stress at failure g gas
1gd porosity in the densest region of the fluidized mf minimum fluidization
bed p particle
1s solids-phase volume fraction ¼ 1 2 1 PD power dissipation
1p compaction gas volume fraction s solids
1b, 1d energy to remove a given volume of target t target
material for brittle and ductile erosion, respect- x x-direction coordinate
ively, J/m3 or Pa y y-direction coordinate
hb, hd mechanical efficiency of impact for brittle and
ductile erosion, respectively Superscripts
uc ¼ 908 2 18.438 ¼ 71.578, degrees · denotes time rate of change
mf fluid viscosity, Pa s ! denotes a vector quantity
r density, kg/m3 ¼ denotes a tensor quantity
rf, rp fluid and particle densities, respectively, kg/m3 n, n þ 1 time levels nDt and (n þ 1)Dt, respectively
rs, rg solids- and gas-phase densities, respectively,
kg/m3 Operators
sb, sd yield strength for brittle and ductile erosion, d=dts total derivative following the solids ¼ ð›=
respectively, Pa ›tÞ þ ~vs ·7
at the minimum fluidization velocity, above which the at 10– 16 atm (1.01– 1.62 MPa) for use in combined-cycle
bed begins to expand and the pressure gradient becomes electricity generation have progressed to pilot-plant scale.
essentially the density of the bed solids per unit area. Because PFBCs are more compact, provide better environ-
Above the minimum fluidization velocity, bubbles mental performances, and have higher thermal efficiencies
develop as the gas velocity is increased. A fast or than AFBCs, the former may be used more successfully for
circularing fluidized bed results when the gas velocity is power generation [2].
increased sufficiently to entrain all or most of the solids Beer [2] and Erlich [3] have reviewed the substantial
out of the bed while simultaneously feeding solids at a progress made in FBC technology since 1970, when the
sufficiently high rate. It is possible to maintain solid Clean Air Act was passed by the US Congress. FBC had
concentrations approaching those of the bubbling been the subject of research on a small scale from the early
fluidized bed. The reader is referred to Kunii and 1970s in US as well as in England [2,4,5]. The concept and
Levenspiel for a more detailed discussion of fluidization usage of fluidized beds, however, go back to 1920– 1930s [1,
phenomena [1]. 4,5]. The environmental benefits issued from this technol-
Atmospheric fluidized-bed combustors (AFBCs) are ogy that attracted interest in 1970s were (1) the reduced
already enjoying some success in the industrial marketplace nitrogen oxides (NOX) emissions made possible by lower
as a highly competitive technology for producing heat and (800– 900 8C) coal combustion temperatures and (2) up to
process steam while meeting stringent pollutant emission 90% SO2 capture using limestone directly in the fluidized
regulations. Pressurized fluidized-bed combustors (PFBCs) bed [6]. Because of the lower combustion temperatures, the
ash is below its sintering or melting point and little or no Most materials work on FBC systems has been
alkali metal vaporization occurs. Thus, buildup of clinkers concerned with corrosion of in-bed components or with
and undesirable deposits on the tubes immersed within the combined erosion – corrosion of gas turbine blading
fluidized bed, on the waterwalls containing the bed, and materials, because the latter problems were thought to be
downstream in the convection pass tube banks is minimized. more severe than erosion of in-bed components. Stringer
The US oil embargo in 1973 further stimulated interest in and Wright [8] argued that, given the existing erosion
FBC technology development [1,5,7]. With the effects of criteria, one would conclude that erosion should not be a
acid rain apparently accelerating—including the death of problem in fluidized beds under ‘normal’ conditions.
large portions of forests and of all animal life in many lakes As more FBC units of large scale accumulated
and ponds in Europe and now in the US—FBC technology substantial numbers of operating hours, more instances of
seems certain to attract increased interest. general, rather than localized erosion have been encoun-
In 1969, the US Environmental Protection Agency tered. For instance, severe wear on in-bed tubes was
(EPA) began exchanging information with the National reported by the International Energy Agency/Grimethorpe
Coal Board (NCB) in England [4]. Close cooperation (IEA/G) PFBC facility in 1982 [9]. The maximum tube wall
ensued at the Grimethorpe Experimental PFBC Facility in metal wastage after 468 h of operation was 1.2 mm, which is
England [2]. This project, which started in 1975, was funded more than 50% of the 2-mm minimum thickness considered
equally by England, the Federal Republic of Germany, and acceptable for safe operation. This translates into an erosion
the United States under the auspices of the International rate of 2.5 mm/1000 h. Because the tube walls are only
Energy Agency (IEA). 4.5 ^ 0.5 mm thick, failures would begin to occur in less
The Tennessee Valley Authority (TVA) and the Electric than 2000 h (about two months) of continuous operation.
Power Research Institute (EPRI) operated a 20-MW(e) Intensive short-term investigations at a number of insti-
AFBC and built a 160-MW(e) demonstration plant. Krishnan tutions under contract to IEA/G resulted in the recommen-
et al. [7] have reviewed US FBC technology. Erlich [3] dation of design changes (lower fluidizing velocity, fins on
pointed out that some 50 companies in 25 countries offered tubes, etc.) that achieved partial mitigation of the wear and
fluidized-bed boilers for sale in 1984. These boilers ranged in allowed the test program to be continued.
size from 1 MW(t) to more than 100 MW(t), and units up to Several tube failures actually occurred in a 16-MW(t)
490 MW(t) in size were being designed. FBC appears to be AFBC after approximately 4000 h of service [10]. The initial
destined for commercial success for both small and large tube wall thickness was 4.2 mm; hence, the average erosion
units in the range of 2.5– 1000 MW operating at atmospheric rate was 1 mm/1000 h. In China, where more than 2000 FBC
and high pressure [1,5,6]. boilers were in operation in 1980, FBC boiler developers
report excessive erosion rates, 1.3 mm/1000 h, of in-bed tube
1.2. The erosion issue surfaces, probably resulting from the use of low grade
Chinese coals of high ash content [11]. In-bed tubes in China
Fluidized-bed combustors apparently work better than are partially protected from excessive erosion through the use
could have been anticipated, a rather unusual situation for a of fins and studs, which break up solids flow patterns around
new technology. However, the issue of erosion has surfaced the tubes. These measures have not been completely
as a serious problem. Some thought was given to the erosion successful in eliminating tube wear, but they may extend
of tubes in FBCs in 1970 [3]; however, the general attitude tube life by a factor of almost 20. Tube bank lifetimes of
held then was that erosion was not a problem [8]. In the few 100 000– 160 000 h (10 – 15 y) are not yet possible.
instances where erosion was noticed, it was localized and Other FBCs, such as the 1.81 m £ 1.81 m, approxi-
appeared to be the result of unusual conditions that could be mately 2-MW(e) Babcock and Wilcox (B&W)/EPRI
resolved by changes in operating conditions and design research AFBC at Alliance, OH, and the overbed feed
modifications. The units were small and operated for short 3.6 m £ 5.5 m, 20-MW(e) TVA/EPRI pilot plant AFBC at
periods of time, typically no more than 1000 h. Paducah, KY, apparently operated for thousands of hours
Erosion wear received little attention in very early without showing any appreciable evidence of erosion [7,8].
investigations of FBCs, because fluidizing velocities were The B&W facility did experience severe bundle damage
typically between 1 and 2 m/s which is generally considered when a deflector plate was lost [8], and the TVA/EPRI
to be too low to produce significant wear. Consequently, facility experienced erosion in the underbed feed system [7].
early materials testing concentrated on understanding Circulating fluidized-bed combustors (CFBCs) eliminate
corrosion and corrosion/erosion processes that could result in-bed tubes and rely on vertical water walls and heat
in heat exchanger and gas turbine component failures. exchanger surfaces hung in the gas flow path high in the
Results of these and other studies led to the belief that, with combustor to remove heat. Refractory lining is required on
selection of alloys appropriate for their intended service and the walls of the hot cyclones to prevent erosion. Bubbling
with careful design and operation of FBC systems, the FBCs (BFBCs), rather than CFBCs, are of primary interest
corrosion problem was manageable. Subsequent experience in this review. However, the methodologies and models
has shown this belief to be generally true. developed are general and would be applicable to CFBCs, as
well as to both AFBCs and PFBCs and other in-bed and out- 2. Modeling of fluidized-bed solids motion
of-bed components (such as fuel feed-nozzles, waterwalls,
and the gas-pass boiling bank, where erosion has also been The key to understanding erosion in fluidized beds is a
found to occur) [6,9,12]. The Georgetown AFBC had 452 of detailed knowledge of the coupled and complex solids
its 980 gas pass boiling bank tubes replaced after 13 636 h circulation and bubble motion. Much experimental work in
of operation, as a result of severe erosion. Hereafter, where gas– solids [13] and gas– liquid – solids [14] fluidized beds
reference is made to FBCs, it is to be understood that they has been carried out during the last 30 þ years. The aim of
are BFBCs. that work has been to understand bubble formation,
Erosive wear of materials is known to be controlled by frequency, size, and velocity and the manner in which the
several complex phenomena that influence the erosion bubbles affect the mechanisms of mixing, heat transfer,
behavior of heat-exchanger tubes and support hangers in gasification, and combustion. Not until relatively recently,
FBCs. Among these phenomena are (1) the feedstock with the emergence of the erosion issue, has equivalent
characteristics, such as chemical composition, particle size experimental work been done to investigate the effects of
and size distribution, and hardness; (2) the operating these phenomena on erosion [15]. Therefore, much less is
conditions, such as fluidizing velocity, temperature, and known about how solid particles actually move and interact
pressure; (3) the mechanical design of the combustor, with system components in FBC bubbling beds [16] than in
including the diameter and pitch of tubes within tube beds that do not contain such components.
bundles, distance from the air distributor and coal and Because of experimental difficulties, solids motion
limestone introduction points, and materials of construction studies are not common. Motion-picture and photographic
for tubes and support hangers. The complex interaction of techniques have been used in conjunction with thin ‘two-
these variables determines the nature and quality of dimensional’ fluidized beds [17] that are much less thick
fluidization (slugging vs. smooth), the combustion gas than wide. This approach is both tedious and time-
composition, the possible formation of protective deposits consuming. The uncertainties in determining the particle
or excessive corrosion, and, finally, the rate of erosion. velocity have been reduced by the ‘quasi-stereoscopic’
Stringer and Wright [8,12] have indicated that a number technique, which has a claimed accuracy within ^ 2.4 mm/s
of additional erosion mechanisms may operate in FBCs, [18]. Miniature, transistorized ‘radio pills’ with in-bed
pickups have been used [19], but the size and density of the
although not at all times or in all units. These mechanisms
pill are not the same as those of the bed material, and the
include:
pickups may alter the flow field. Fiber-optic probes have
been used, but they are too intrusive and may significantly
† Particles ‘loaded’ onto the surface by a block of other alter the flow field [20]. A novel non-intrusive, radioactively
particles. tagging facility, the Computer Aided Particle Tracking
† Fast-moving particles in the wake of rising bubbles. Facility, has been used to obtain time-averaged solids
† Large, but slow-moving, particles with high kinetic circulation data in fluidized beds, both without [21] and with
energy. immersed tubes [22].
† Particles thrown onto metal surfaces by collapsing of Understanding of how bubbles interact with solid
bubbles. surfaces within fluidized beds is almost non-existent.
† Particles accelerated onto in-bed components under These surfaces may take the form of instrument probes
the influence of in-bed jets associated with coal used to detect the bubbles themselves [23] and, in the case of
and limestone feedports, limestone recirculation ports, fluidized-bed combustors, heat exchanger tubes and baffles
and air streams used to keep bed drains clear of placed in the bed, in the free-board, or both [24,25].
obstructions. State-of-the-art computational fluid dynamics (CFD)
† Particles trapped within and moving with large-scale techniques and improved computing capability have made it
flow patterns (gulf streams) in the fluidized bed. possible to model the movement of solids and bubbles in
FBCs. Significant progress has been demonstrated in the
ability of the hydrodynamic model of fluidization to predict
In summary, it is still not possible at present to explain bubble formation (including frequency of formation, growth
why some fluidized beds or bed regions experience rapid rate, size, trajectory, rise velocity, and conditions giving rise
erosion and others do not. Consequently, it is not possible to to bubble splitting). Just as importantly, solids volume fraction
suggest completely satisfactory remedies or design a and gas velocity computations have also been performed.
fluidized bed for which it is known that erosion would not The hydrodynamic approach to fluidization, which started
be a problem. The ad hoc approaches mentioned earlier are with Davidson in 1961 [26], serves as the basis of the three
unlikely to succeed in the long-term without a better two-dimensional fluidized-bed codes: CHEMFLUB (devel-
understanding of the underlying processes [8]. One of the oped by Systems, Science and Software), FLAG (developed
major objectives of this review is to improve that by JAYCOR), and IIT. The capabilities of these codes were
understanding. reviewed by Smoot [27] and Gidaspow [28]. The progress
made in the last 20 years of modeling small-scale, highly Gas-phase momentum in the x-direction
instrumented, two-dimensional fluidized-bed experiments at › › ›
the Illinois Institute of Technology (ITT) and the University ðrg 1Ug Þ þ ðrg 1Ug Ug Þ þ ðrg 1Vg Ug Þ
›t ›x ›y
of Champaign at Urbana, using both the IIT computer code ›P
and the FLUFIX/MOD2 computer code developed at ¼ 21 þ bx ðUs 2 Ug Þ þ SUG ð2:3Þ
›x
Argonne National Laboratory (ANL) [29], has served in
Solids-phase momentum in the x-direction
great part to validate the hydrodynamic models [22,28,
30 – 33]. This progress up to approximately 1993 has been › › ›
½r ð121ÞUs þ ½rs ð121ÞUs Us þ ½rs ð121ÞVs Us
reviewed by Gidaspow [34]. ›t s ›x ›y
Enwald et al. [35] performed a review in 1996 of ›P ›1
¼2ð121Þ þ bx ðUg 2Us ÞþGð1Þ þSUL ð2:4Þ
publications on simulations of bubbling and circulating ›x ›x
fluidized beds. They summarized the works in tabular form Gas-phase momentum in the y-direction
and concluded that ‘The members of the IIT/ANL group are › › ›
pioneers in the field and have published the greatest amount ðr 1V Þþ ðr 1U V Þþ ðr 1V V Þ
›t g g ›x g g g ›y g g g
of articles, using a code which is based on the K-FIX code, ›P
originally developed by Rivard and Torrey’ [36]. Enwald ¼21 þ by ðVs 2Vg Þ2 rg 1g þSVL ð2:5Þ
›y
et al. [35] also reviewed the formulation of two-fluid
hydrodynamic models and closure relationships applied to Solids-phase momentum in the y-direction
fluidization and made an excellent attempt to categorize › › ›
½r ð121ÞVs þ ½rs ð121ÞUs Vs þ ½rs ð121ÞVs Vs
these models systematically. ›t s ›x ›y
This section presents a summary of the basic hydrodyn- ›P ›1
amic models used by us, and representative computations ¼2ð121Þ þ by ðVg 2Vs Þ2 rs ð121ÞgþGð1Þ
›y ›y
and comparisons with experimental results for the IIT thin, þSVL ð2:6Þ
two-dimensional, rectangular bed having a central jet and
containing a rectangular obstacle [31]. This experiment,
along with the one without an obstacle [30], has become an Hydrodynamic Model B differs from Hydrodynamic
ad hoc standard problem. Model A in the treatment of the pressure gradient term, all of
The purpose of summarizing the basic hydrodynamic which is contained in the gas phase [29,31].
models in this section is to give the reader some appreciation The solid elastic modulus, Gð1Þ; is used to calculate the
of their complexity and the computed hydrodynamic normal component of the solids stress through the following
quantities that can be predicted, which, once validated, relationship
serve as inputs to the various erosion models described in
Sections 3 and 5. For other models in the literature, the ›t
›t ¼ ›1 ¼ Gð1Þ›1 ð2:7Þ
reader is referred to the review by Enwald et al. [35]. ›1
which is similar to what is done in solid mechanics. Eqs.
2.1. Hydrodynamic model (2.1) – (2.6) are used in Section 5.2 to derive energy
dissipation models.
The hydrodynamic model of fluidization uses the There are six non-linear, coupled, partial differential
principles of conservation of mass, momentum, and energy. equations for the six dependent variables to be computed: the
The continuity equations and the separate phase momentum void fraction, 1; the pressure, P; the gas velocity components
equations for two-dimensional, transient, isothermal two- Ug and Vg and the solids velocity components Us and Vs in the
phase flow in Cartesian coordinates, which form the basis of x- and y-directions, respectively. The equations are written in
the FLUFIX/MOD2 [29] code, are given below for hydro- a form similar to that used in the K-FIX computer code, from
dynamic Model A (terms are defined in the Nomenclature): which FLUFIX/MOD2 has been developed [36]. The terms
Hydrodynamic model A: SUG, SUL, SVG, and SVL represent the viscous stress terms
for the gas and solids phases, respectively, in symbolic form.
Gas-phase continuity For details of their form, the reader is referred to K-FIX [36]
› › › and Bouillard et al. [37]. Bouillard et al. [37] assessed the
ðr 1Þ þ ð1rg Ug Þ þ ð1rg Vg Þ ¼ 0 ð2:1Þ effect of the solids-phase stresses (the gas-phase stresses are
›t g ›x ›y
negligible in comparison) and found negligible effect up to
0.1 Pa s (1.0 Poise). The conservation equations in the
Solids-phase continuity FLUFIX code are solved in conservation-law form in two-
dimensional Cartesian and axisymmetric cylindrical coordi-
› › › nates using the implicit multifield (IMF) numerical technique
½r ð1 2 1Þ þ ½rs Us ð1 2 1Þ þ ½rs ð1 2 1ÞVs ¼ 0
›t s ›x ›y [38]. Cartesian coordinates were used in the present
ð2:2Þ computations.
The treatment of the pressure gradient term in the gas Table 1

and solids-phase momentum equations above (Hydrodyn- Solids elastic modulus parametersa
amic Model A) results in an initial value problem that is
Reference c 1p Model
illposed, as is discussed in detail by Lyczkowski et al. [39].
This situation leads to a conditionally stable numerical
[31] 600 0.376 1
solution. One way to overcome the problem is to retain a
[43] 500 0.422 2
normal component of solids-phase stress, t, sometimes [30] 20 0.62 3
associated with solids-phase pressure or particle-to-particle
a
interactions, in the solids-phase momentum (Eqs. (2.4) and G in Eq. (2.8) is in Pa.
(2.6)). Another way to overcome the problem is to retain the
viscous stresses. Bouillard et al. [32] compared calculations In the above analysis, Gð1Þ ¼ 1:0 Pa when 1 ¼ 1p : The
using Hydrodynamic Model A and Hydrodynamic Model B Rietma and Mutsers model is inappropriate to keep the bed
(which is well posed) and found little differences using the from compacting, because the model was developed for data
same solids elastic modulus and no solids viscous stresses. taken at a much higher compaction porosity (0.62), and the
In addition to increasing stability, the primary compu- compaction modulus, c, is very low. Porosities below 0.2,
tational function of the solids elastic modulus term is to keep which is possible only with great compressive force, have
the bed from compacting below the defluidized or packed- been computed with this model. The Gidaspow and Syamlal
bed state of approximately 0.38– 0.4 gas phase volume model, developed from solids – gas flow data through
fraction, 1, usually called the porosity in the fluidization aerated hoppers, is appropriate at a compaction porosity
community. Any solids stress model that accomplishes this near minimum fluidization (0.422) [44]. The compaction
is adequate as long as the solids pressure equals the weight modulus is high because the particles are being compacted
of the solids per unit area at equilibrium [34]. The Rietma as they flow down through the hopper. Our model has a
and Mutsers data [40] used earlier by Gidaspow and somewhat higher compaction modulus and a lower
Ettehadieh [30] was found by us to be computationally compaction porosity, more appropriate for a packed bed
inadequate for cases involving obstacles, because it resulted (0.376). This is necessary because of the greater compaction
in overcompaction, and was modified to overcome this
resulting from solids striking the obstacle.
shortcoming. From computational experience, this solids
The other major empirical input in this cold bed model is
stress term has been shown to affect the instantaneous bed
the fluid-particle drag coefficient, b. This coefficient,
interface and bubble shapes [31].
obtained from standard correlations as originally used by
To place this term in perspective, we must consider the
Gidaspow and Ettehadieh [30], presented in Gidaspow’s
mechanisms of the compaction of powders [41,42]. The
book [34], was used unchanged for calculations involving
motivation for the most generally satisfactory expression is
an obstacle. Below a porosity of 0.8, b is given by the Ergun
the experimental observation that plotting the logarithm of
equation and above 0.8, it is given by Wen and Yu’s
consolidating pressure versus volume yields a substantially
expression. Other drag functions are discussed by Enwald
straight line for both metallic and non-metallic powders
et al. [35] who show that they all produce essentially the
undergoing compaction [41]. Bouillard et al. [31] used this
same values.
simple theory to derive a generalized solids elastic modulus
coefficient, Gð1Þ; of the form
2.2. Representative computational results and comparisons
Gð1Þ ¼ G0 exp½2cð1 2 1p Þ ð2:8Þ with data
where c (the compaction modulus) is the slope of ln(G )
versus 1, and 1 p is the compaction gas volume fraction. The As mentioned at the beginning of Section 2, the IIT
normalized units factor, G0, has been taken to be 1.0 for experiments for a thin rectangular two-dimensional flui-
convenience. Considerable disagreement exists over the dized bed with a central jet with [31] and without [30] a
exact form of this relationship. Shinohara has summarized rectangular obstacle have become ad hoc standard problems
15 different expressions [42]. to test the validity of the models, closure laws (also called
Eq. (2.8) is a convenient and consistent expression with constitutive relationships), and numerical solution methods
which to interpret the physical significance of solids for fluidized-bed hydrodynamics CFD codes [45,46]. For
pressure data. We have converted the Rietma and Mutsers that reason, a description of the simulation is included in this
data [40], as curve fit and modified by Gidaspow and section together with a comparison with data. The validated
Ettehadieh [30], the expression used by Gidaspow and output from the FLUFIX code is used as input to the erosion
Syamlal for gas – solids critical flow [43], and the expression models described in Sections 3 and 6.
we have been using for hydrodynamics and erosion The configuration being modeled is the same as that in
calculations for a fluidized bed containing an obstacle [31, the publication by Bouillard et al. [31] (Fig. 1). The
32,34] into the form of Eq. (2.8), as shown in Table 1. computational region is 19.685 cm wide (x direction) by
Enwald et al. [35] present a plot of Table 1 in their Fig. 8. 58.44 cm high (y direction). The cell dimensions are
velocity, 26.0 cm/s. There are no solids entering, hence, the

inlet porosities are set to 1.0. The pressures in the dummy
cells at the top ðJ ¼ 14Þ are set equal to atmospheric
pressure (101.3 kPa), and the axial solids velocities, Vs are
set to zero at the exit ðJ ¼ 13Þ; that is, a wire mesh is
simulated to prevent solids carry-over. The pressures in the
bottom row of dummy cells ðJ ¼ 1Þ are set equal to
atmospheric pressure plus 1.2 times the total bed weight of
solids per unit area (105.5 kPa), to simulate the distributor
plate pressure drop measured in the experiment. On all solid
surfaces except the inlet, outlet, and line of symmetry,
no-slip boundary conditions are used (i.e. normal and
tangential velocities for each phase are set equal to zero).
Initially, all of the lateral gas velocities, Ug, are set to
zero, the axial gas velocities, Vg, are set equal to the
interstitial gas velocity at minimum fluidization (26.0 cm/s
in the freeboard and 61.9 cm/s in the bed having a uniform
porosity of 0.42) and all the solids lateral and axial
velocities, Us and Vs are set to zero. The initial pressure
distribution corresponds to the static bed height per unit
area. At time, t, greater than zero (0þ), the gas flow through
the jet into cell (2,2) is increased to 578 cm/s. A fixed time
step of 0.1 ms was used. Typical running times on an IBM
3033 mainframe computer were about 1 h for each second of
transient time in the simulation. Somewhat shorter running
times are experienced on present day personal computers
and workstations.
The computations were performed before high-speed
(800 frames/s) motion pictures were taken of a flow
visualization experiment modified to include an obstacle.
Fig. 1. FLUFIX computational mesh for coarse mesh, showing
This experiment is described in more detail by Gidaspow
obstacle location. et al. [47]. A detailed schematic diagram of the plastic two-
dimensional fluidized bed is shown in Fig. 2. Fig. 3 shows a
Dx ¼ 0.635 cm and Dy ¼ 4.87 cm, so that the number of still frame from this motion picture at about 0.25 s into the
computational cells is 31 in the x-direction and 12 in the y- transient (there is an uncertainty of about 0.02 s in the time
direction, for a total of 372. In Fig. 1, the numbers in the jet was turned on), illustrating the formation of the first
parentheses refer to key cell numbers ðI; JÞ: Symmetry about bubble. The numbers above the white horizontal lines are
the central jet is assumed; hence the actual bed width is the bed height in inches. The bed height has increased from
39.37 cm. In the previous modeling work without an its initial value of 29.22 cm (11.5 in.) to 35.6 cm (14 in.)
obstacle [28], symmetry was assumed, and agreement with near the bed center and to 33 cm (13 in.) near the bed edge.
data was good. The jet half-width is 0.635 cm (one cell The bubble height is between 25.4 and 30.5 cm (10– 12 in.),
width). The jet velocity is 578 cm/s, and the secondary air and the bubble width is between 5.08 and 10.16 cm (2–
velocity of 26.0 cm/s maintains the bed without a jet at 4 in.). Fig. 4 shows a dot plot representation of the computed
minimum fluidization. The particle diameter is 503 mm, and porosity distribution at 0.255 s. The dots are distributed
the density is 2.61 g/cm3. The obstacle is placed two nodes randomly throughout each computational cell. The densest
above the jet and is two nodes wide by two nodes high white regions represent a packed-bed state ð1 < 0:4Þ; and
(1.27 cm wide by 9.74 cm high). Because the initial bed black represents all gas ð1 ¼ 1:0Þ: The right side of Fig. 4 is
height is 29.22 cm (six cells high), the obstacle lies a contour plot representation of the computed porosity
completely within the bed. Although this configuration is distribution at the same time. Comparing the dot and
not typical of FBC geometries, it was selected because (1) it contour plot representations shows that the perceived edge
is similar to the model without the obstacle, so that prior of the bubble is a contour of porosity, 1, of about 0.7– 0.8.
experience is relevant and (2) it serves to further validate the Comparison of Figs. 3 and 4 reveals generally good
hydrodynamic model. agreement. The predicted size and location of this first
The boundary conditions are described here. At the inlet bubble agree well with the experimental results, and the
ðJ ¼ 2Þ; the axial gas velocities, Vg, are set equal to the expanded bed height and shape are approximately correct.
experimentally determined minimum-fluidization superficial The slight asymmetries present in the experiment were not
Fig. 2. Detailed schematic of the two-dimensional cold fluidized bed with a rectangular central jet and rectangular obstacle.
accounted for in the model. Hence, the formation of a vortex

street above the obstacle (which appears to sweep particles
back and forth, keeping them from piling up) is absent, and
the computations show a solids buildup.
A computation utilizing a finer mesh was performed to
obtain better resolution. The number of computational cells
in the x-direction was the same, but the number in the
y-direction was increased to 48 (Dy ¼ 1.217 cm). This
change makes the cell aspect ratio much closer to unity
(Dy/x ¼ 1.92) than is the case for the 31 £ 12 node
computational mesh (Dy/Dx ¼ 7.67). The agreement
between the 31 £ 12 node mesh and the 31 £ 48 node
mesh computations is good, as can be seen by comparing
Figs. 4 and 5. The expanded bed shape is in better agreement
with the experiment and is also sharper, indicating that the
bed interface is resolved better with the finer mesh. Fine
details, such as the splash of solids against the bottom of the
obstacle, are also resolved. The price paid for the finer
resolution is considerably longer computing time (approxi-
mately 10 times as much as required for the coarser mesh).
Fig. 3. High-speed motion picture still from a two-dimensional flow The solids velocity and porosity patterns at 0.255 s,
visualization experiment of a fluidized bed with an obstacle, central plotted in Fig. 4 indicated the existence of a vortex pattern
jet, and secondary air flow at minimum fluidization (time ~ 0.255 s). in the wake of the rising bubble near the lower sides of
Fig. 4. Computer-generated porosity distributions for a two-dimensional fluidized bed with an obstacle, central jet, and secondary air at
minimum fluidization at 0.255 s, 31 £ 12 nodes nodes.
the obstacle. Also revealed is a larger general solids

concentration pattern, induced by the rising bubble. Such
solids motions give rise to the erosion of immersed heat
exchanger tubes in fluidized-bed combustors.
With the 31 £ 48 node mesh, it is possible to resolve the
angularity of the bubble bottom. Fig. 6 shows a still frame
from the high-speed motion picture at about 0.105 s. At this
Fig. 5. Computer-generated porosity distribution for a two- Fig. 6. High-speed motion pictures still from a two-dimensional
dimensional fluidized bed with an obstacle, central jet, and flow visualization experiment of a fluidized bed with an obstacle,
secondary air flow at minimum fluidization at 0.255 s, 31 £ 48 central jet, and secondary air at minimum fluidization (time ~
nodes. 0.105 s).
Fig. 7. Computer-generated porosity distributions for a two-dimensional fluidized bed with an obstacle, central jet, and secondary air at
minimum fluidization at 0.13 s, 31 £ 48 nodes.
time the bed has barely expanded, and the size of the bubble almost identical to those using Hydrodynamic Model A
is significantly less than at about 0.255 s, as shown in Fig. 3. [31]. This agreement implies that the solids flow patterns are
The bed height is just over 30.5 cm (12 in.). Fig. 7 shows dot also reasonably correct, although they were not measured at
and contour plot representations of the computed porosity this time. Multiple experimental runs should be performed,
distributions at 0.13 s. Comparison of Fig. 7 with Fig. 6 so that the slightly different bubble patterns that result from
reveals good agreement in terms of the bubble size and the random variations in initial conditions from run to run can
angle at the bottom of the bubble. The experimental angle is be averaged out.
428; the computed angle (solid line in Fig. 7), 368, does not As mentioned earlier, it was assumed for the computer
vary significantly over a ^ 0.02-s time span. The computed simulation that the initial solids velocity is zero, that the bed
bed expansion is also in good agreement with experiment. porosity is uniform at 0.42, and that the gas mass flux
Comparative analysis of a computer-generated motion corresponds to minimum fluidization conditions. The high-
picture of the contour plot representation of the computed speed motion picture study revealed that very small bubbles
porosity distribution and the high-speed motion picture were observed to originate from the side of the obstacle
study reveals good agreement for the frequency of the before the jet was turned on. This same phenomenon was
bubble around the obstacle (, 4 Hz) and the higher- observed by Loew et al. [18] and Buyevich et al. [48]. The
frequency, small bubble that forms under the obstacle latter investigators who studied flow around a rectangular
(,10 Hz). This smaller bubble can be seen under the obstacle in two-dimensional fixed and fluidized beds,
obstacle in Figs. 3 –5. explained the phenomenon of bubble formation as arising
Fig. 8 illustrates the computed porosity fluctuations from the low gas flow resistance next to the obstacle (due to
around the obstacle. Locations 1 and 2 are below the high local porosity). The resultant excess gas flow, which
obstacle, locations 3 and 4 are on the side of the obstacle, was measured in a fixed bed to be 5 – 10 times that in the
and locations 5 and 6 are on the top of the obstacle. The bulk of the bed, produces the bubbles.
porosity amplitudes were not measured and, hence, are not Computations were performed with the central jet off
validated. We suspect that the computed porosity and the secondary air at minimum fluidization. For these
amplitudes may be somewhat low, because the porosities computations, symmetry was not assumed. The number of
do not approach 1.0 on the bottom and side of the obstacle nodes in the transverse direction was increased to 64 (from
after the passage of the first bubble. As can be seen in Fig. 8, 33 for the coarse mesh nodalization). The inlet velocity was
the porosity only approaches 0.7– 0.85 after the first bubble maintained at 26.0 cm/s. As can be seen in Fig. 9, the results
has passed (after 0.3 s). are not quite symmetrical. This asymmetry is probably
Additional studies that modified the form of the triggered by perturbations introduced into the uniform flow
hydrodynamic model momentum equations and the solids as a result of asymmetric sweeping through the compu-
stress [31]. The conclusion reached is that the computed tational cell during the iteration process. A weak bubble
bubbles do vary, but all are in generally good agreement forms under the obstacle and moves upward, splitting in the
with the high-speed motion picture study. Surprisingly, process; split bubbles then move up the sides of the
computed results from using Hydrodynamic Model B are obstacle and eventually out of the bed. Examination of
Fig. 8. Computed porosity fluctuations around a rectangular obstacle immersed in a two-dimensional fluidized bed with a central jet and
secondary air at minimum fluidization (Hydrodynamic Model A).
Fig. 9. Computed porosity distribution in a two-dimensional fluidized bed at minimum fluidization with an immersed rectangular obstacle at
(a) 0.1, (b) 0.2, and (c) 0.3 s.
the computer-generated motion picture showed that bubbles an obstacle. These distributions are shown in Fig. 10, where
formed below the obstacle and moved upward at a they are compared with the time-averaged FLUFIX
frequency of approximately 3 Hz. It is clear from Fig. 9 computations using approximately the same scale. The
that the experimental fluidized bed was probably not in a agreement is generally very good; in fact, it is better than
completely uniform initial condition when the jet was turned that previously reported for the case without an obstacle
on. This could explain some differences between the [30]. The angle and extent of the slumped zone, 1 ¼ 0:4; are
experiment and the calculated results. in reasonable agreement with the experimental results, as
The time-averaged porosity distributions were also are the bed height and shape. The presence of a layer of gas
measured in the two-dimensional fluidized bed that included on the side of the obstacle is also predicted. The shape and
Fig. 10. Comparison of experiment (right) and computed (left) (Hydrodynamic Model A) time-averaged porosity distribution in a two-
dimensional fluidized bed with an obstacle, central jet, and secondary air at minimum fluidization (I: error range; 1: as indicated: …, 0.4; – -, 0.5;
—, 0.6; †-†-, 0.7; ___, 0.8).
extent of this layer is in good agreement with the data. The 2. Dense phase fluidized-bed models. The consideration for
vertical bars on the time-averaged computations are the development of this type of model is the repeated
estimated error ranges obtained by time-averaging over interaction of many particles with the wall. Depending
different time intervals between 1.5 and 2.5 s. on the model, the wall material is assumed to be removed
in a (a) purely ductile mode, (b) purely brittle mode, or
(c) low-cycle fatigue-failure mode.
3. Overview of erosion modeling and mechanisms 3. Power and energy dissipation models. These models
are based on the very general consideration of erosion
This section presents an overview of the erosion models resulting from energy transfer from the solids in a
that we consider reasonable candidates for inclusion in a two-phase mixture to the eroding surfaces.
preliminary consolidation, i.e. a collection of various types
of erosion models that would be driven—for the first time to All but the power and energy dissipation models are
our knowledge—by the types of hydrodynamic fluidized- algebraic in nature.
bed calculations presented in Section 2. A detailed literature review of erosion models is not
The marriage of hydrodynamic and erosion models has attempted here. Engel has provided a good literature review
been fairly recent. This may account for the general attitude up to 1978 [50], Sarkar has provided one up to 1980 [51],
that present-day erosion models do not yet satisfactorily and the exhaustive review by Meng and Ludema [52]
explain erosion patterns in FBCs. To be successful, the fluid provided one up to 1995, found 182 equations for wear, and
mechanics and erosion models must be used together. This selected 28 for special study. A review done at the
viewpoint was pointed out in the review by Humphrey [49] Morgantown Energy Technology Center (METC), now the
who concluded that the interpretation and understanding of National Energy Technology Laboratory (NETL) [53], deals
particle impact erosion in terms of material properties alone specifically with erosion in FBCs. Humphrey [49] has
has been hampered by paying little or no attention to reviewed the importance of various fundamental consider-
clarifying the influence of fluid motion, especially in the ations relating to the motion of dilute suspensions of solids
turbulent regime. affected by the carrier fluid and a constraining surface.
The erosion models that form the preliminary consolida- In this review, we summarize the models we have
tion may be organized as follows: selected from the literature, explain the criteria for their
selection, and point out their various shortcomings. We also
1. Single-particle (dilute phase) models. The consideration explain why we were led to the more fundamental energy
for the development of this type of model is the interaction dissipation approach to modeling erosion. Because we
of a single particle with a planar wall. Depending on the cannot claim that this approach will ultimately be more
model, the wall material is assumed to be removed in a (a) successful than the others, the other models selected for the
purely ductile mode, (b) purely brittle mode, or (c) consolidation are also considered. At this time, most of the
combination of ductile and brittle modes. calculations have been done with the monolayer energy
dissipation (MED) erosion model discussed in Section 6, as the idealized model particle; Finnie arbitrarily took c ¼
and limited comparisons have been made with the others. 1=2: With M, the total mass of abrasive particles, V; ~ the
Although the results are promising, very few directly abrasive-particle velocity, and p, the eroding surface ‘flow
appropriate erosion experiments in fluidized beds are stress’, W is the total volume of target material removed. To
available to validate the models. Dellenback and Johansen obtain the mass of eroded material removed, W is multiplied
[15] and the body of work ongoing at the Chalmers by rt, the target density.
University of Technology critiqued in Section 9, have The erosion rate (or erosion velocity because it has
recently performed such experiments. The better than order- dimensions of length/time), E;_ is usually given as
of-magnitude agreement with published erosion rate data,
E_ ¼ W=A
_ t ð3:2aÞ
taken at comparable jet and/or superficial fluidizing
velocities, as well as correct trends leads us to conclude where W _ is the volumetric rate of target material removal,
that the energy dissipation approach is predictive as shown and At is the average area of the target. The erosion rate (or
in Sections 7.5 and 7.6. erosion velocity) of the target itself, Et, is given by
3.1. Single-particle erosion models E_ t ¼ 2E_ ¼ 2W=A

_ t ð3:2bÞ
because the volumetric rate of change of the target itself is
Single-particle erosion models treat the erosion process negative.
in terms of the interaction of a single particle with the In order to apply the single particle Finnie erosion model
eroding surface. Such models may be more appropriate for (or for that matter, any other single particle erosion model)
the case of erosion due to dilute solid suspensions, as in to compute erosion rates for a continuum of solids, such as
pneumatic conveying pipelines and elbows or in turbine in fluidized beds, using the FLUFIX/MOD2 code [29]
blade cascades. In these cases, the particle – particle computed hydrodynamic results, Eqs. (3.1a) and (3.1b)
interactions are negligible, and the erosion process may be must be modified. First, the total particle mass, M, is
thought of as removal of surface material by the cumulative replaced by the mass flux of solids, m _ s ¼ ð1 2 1Þrs l~vs l;
action of the individual particles. In this sense, these models where ð1 2 1Þ ¼ 1s is the solids volume fraction, rs is the
are non-continuum models. particle density, and l~vs l is the magnitude of the velocity of
the solids phase. The mass flux, m _ s ; is assumed to be
3.1.1. Finnie’s ductile erosion model positive toward the eroding surface. The particle velocity,
According to Engel [50], Finnie was the first (in 1958) to ~ is replaced by ~vs to obtain
V;
derive a single-particle erosive cutting model [54]. In 1960, 8 2
Finnie discussed the assumptions of the model, quoted the >
< C ð1 2 1Þrs l~vs lð~vs Þ f ðaÞ m
_s . 0
results, and compared the results with experimental data E_ ¼ p ð3:3aÞ
>
:
[55]. This model set the basic pattern and tone for all single- 0 _s # 0
m
particle models, so it is discussed here in some detail. The
major assumption is that a particle, approaching the eroding where E_ is the erosion rate (in m/s) and C ¼ c=CK: The
surface (or target) at angle a as measured from the surface erosion rate is positive if the solids velocity vector points
(called the impingement angle), will remove material in toward the eroding surface; otherwise, it is zero. An
much the same way as a machine tool would. The particle is alternative approach using the kinetic theory of granular
assumed to be much harder than the surface and does not flow is summarized in Section 8.1.
break up. The surface material is assumed to deform With K ¼ 2; the angular dependency function f ðaÞ is
plastically during the cutting process; hence, the material is given by
ductile. Ductile materials, such as aluminum or structural
f ðaÞ ¼ sinð2aÞ 2 3 sin2 ðaÞ a # 18:438 ð3:3bÞ
steel, can develop a relatively large tensile strength before
they rupture. f ðaÞ ¼ cos2 ðaÞ=3 a . 18:438 ð3:3cÞ
The final expression for the volume of target material, W,
removed obtained by Finnie is as follows [55]: Below 18.438, the surface is cut until the particle leaves the
surface; above 18.438, cutting ceases before the particle
M V~ 2 6 K leaves the surface. The transition angle given in Eq. (3.3a) is
W ¼c sinð2aÞ 2 sin2 ðaÞ tan a # ð3:1aÞ
CpK K 6 close to the angle of maximum erosion, amax, given by
" # Engel [50] as
M V~ 2 K cos2 ðaÞ K
W ¼c tan a . ð3:1bÞ
CpK 6 6 amax ¼ 1
2 tan21 ðK=3Þ ð3:3dÞ
Finnie [55] took K ¼ 2; where K is the ratio of vertical to With K ¼ 2; amax ¼ 16.858. Eq. (3.3a) is in the same form
horizontal (frictional) force, and C ¼ 1; where C is the ratio as that used by Pourahmadi and Humphrey [56] in their
of the depth of contact to the depth of the cut. The constant c erosion modeling studies. They defined C as the fraction of
allows for the fact that many particles will not be as effective particles cutting in an idealized manner, consistent with
Finnie’s 1972 modification of his model [57]. (If one uses 3.1.2. Bitter’s combined ductile and brittle erosion model
Finnie’s values of c, C, and K, then C ¼ 1=8:) The basic assumptions in Bitter’s analysis are that
Pourahmadi and Humphrey [56] used the following deformation and cutting erosion occur simultaneously and
expression for f ðaÞ : that the two effects can be linearly superimposed [60,61].
Thus, Bitter’s work extends Finnie’s model and corrects it
f ðaÞ ¼ sinð2aÞ 2 4 sin2 ðaÞ a # 14:048 ð3:4aÞ by bringing in the concepts of a threshold erosion rate and
f ðaÞ ¼ cos2 ðaÞ=4 a . 14:048 ð3:4bÞ energy dissipation. Physically, the impinging particle cannot
erode the target material if its impacting velocity is smaller
It may be that they used the original Finnie model with than a threshold velocity, Vel.
K ¼ 1:5; in which case a value of C ¼ 1=6 is obtained. Brittle erosion model. In Bitter’s model, the brittle
Alternatively, they may have used Finnie’s modified model erosion rate is postulated to be equal to the energy
[57], which accounted for particle inertia, with K ¼ 2; in dissipation of an elastic sphere deforming the planar target
this case, C ¼ 1=8: The value of C used by Pourahmadi and material surface elastically and plastically, divided by an
Humphrey [56] is never mentioned, so either case is energy (itself dependent on material properties) needed to
possible. remove material. This brittle erosion model is given by
At 18.438, Finnie’s model ðK ¼ 2; C ¼ 1=8Þ yields an
erosion rate for ductile target materials given by M½V~ sin a 2 Vel 2
Wb ¼ V~ sinðaÞ . Vel ð3:6aÞ
21b
_ ð~v Þ2
m
E_ FM ¼ 0:075 s s ð3:5Þ
2p Wb ¼ 0 V~ sinðaÞ , Vel ð3:6bÞ
which is, in effect, the maximum value, E_ FM : Eq. (3.5)
where 1b is the material-dependent deformation wear factor.
predicts that only 7.5% of the particle’s kinetic energy goes
The threshold velocity, Vel, is the velocity of collision at
into erosion for a given hardness (or flow stress), p. The
which the elastic limit of the eroding surface is just reached,
corresponding percentage for Pourahmadi and Humphrey’s
given theoretically from the Hertz contact theory [50] by
expression ðC ¼ 1=6Þ is 7.8%. Using Eq. (3.5), Finnie [55]
analyzed some data taken for silicon carbide eroding SAE Vel ¼ 15:4sy5=2 rp21=2 Er22 ð3:7aÞ
1020 low carbon steel and found that the value of p
exceeded the ‘true stress at fracture in a tension test’ by a where sy is the plastic load limit, rp is the particle density,
factor of almost three. However, if Finnie’s 1972 modifi- and Er is the reduced Young’s modulus of elasticity. The
cation of his model (which accounts for particle inertia) is value of Er is given by
used together with his recommended values of c, C, and K,
then 5.9% of the particle’s kinetic energy would go into 1
Er ¼
erosion, instead of 7.5%. ½ð1 2 g2p Þ=ðpEp Þ þ ½ð1 2 g2t Þ=ðpEt Þ
Single particle erosion models in general, and Finnie’s
1
erosion model in particular, cannot be used to calculate ; ð3:7bÞ
absolute erosion rates a priori. Finnie’s erosion model can be ðkp þ kt Þ
used, together with estimates of target material flow stress
where gp and gt are the Poisson’s ratios, and Ep and Et are
(or hardness), to back out a value for parameter C to match
the Young’s moduli of elasticity, of the particle and target,
the data. On the other hand, if a value for C is assumed, then
respectively. The threshold velocity, Vel, which can be
the value of the flow stress or hardness appropriate for
computed from Eq. (3.7a), can be determined from particle
erosion is backed out of the data.
rebound data using the following relation [60]
The angular dependence of the erosion rate predicted by
using the original Finnie model and its first modifications is V2 ¼ ð2V1 Vel2 Þ1=2 V2 ¼ ð2V1 Vel2 Þ1=2 ð3:8Þ
quite good up to 458. Above 458, the Finnie model
underpredicts the erosion rate. At 908, it predicts no erosion where V1 and V2 are the velocities at the beginning
rate at all, whereas the analyzed data clearly indicate that (approach) and end (rebound) of collision, respectively.
this prediction is not correct. Further reworking of Finnie’s Ductile erosion model. The portion of Bitter’s model
theory did not resolve this problem [58]. The Finnie model devoted to ductile erosion consists of the following
also predicts no erosion at 08 (scouring erosion). Shewmon
and Sundararajan [59], who reviewed the literature on 2MC 0 ½V~ sinðaÞ 2 Vel 2 ~
Wd ¼ £ V cosðaÞ
erosion in 1983, concluded from scanning electron micro- ½V~ sinðaÞ1=2
scope (SEM) examinations of erosion surfaces that the
cutting tool analogy is not valid. They regarded the Finnie C0 ½V~ sinðaÞ 2 Vel 2 ð3:9aÞ
2 fc
model as being of historical interest only and suggested ½V~ sinðaÞ1=2
other mechanisms, such as shear localization leading to lip
formation and fracture. a # a0
M{V~ 2 cos2 ðaÞ 2 K1 ½V~ sinðaÞ 2 Vel 3=2 }

Wd ¼
2f ð3:9bÞ
a . a0
where fc is the material-dependent cutting wear factor. The

constants C0 and K1 are given by [44]
C0 ¼ 0:288ðrp =sy Þ1=4 =sy ð3:10aÞ
and
K1 ¼ 8:036s2y Er22 ðsy =rp Þ1=4 ð3:10bÞ
The angle a0 may be estimated from Finnie’s model (Eqs.

(3.1a) and (3.1b)), or by equating Eqs. (3.2a) and (3.2b) and
solving for a. The total erosion rate, W, is then given by the
sum of Wb and Wd.
The effect of K1 is negligible in Eq. (3.9b) (where
a . a0 ) and can safely be dropped [61]. If fc is associated
with 3Cp then (except for the factor c ) Eq. (3.1b) Fig. 11. Erosion mechanisms show wear trends as a function of
(tan a . K=6) of Finnie’s erosion model and Eq. (3.9b) impact angle (a ¼ impingement angle).
ða . a0 Þ of Bitter’s erosion model are the same.
The interesting features of Bitter’s model are that (1) for
soft ductile materials, it produces wear curves similar to W ¼ Wd þ Wb
Finnie’s, but with non-zero wear at 908 and (2) for hard
brittle materials, it produces wear curves that reach a M V~ 2 cos2 ðaÞ M½V~ sinðaÞ 2 Vel 2
¼ þ ð3:11bÞ
maximum at 908. The shapes of these curves are shown 2fc 21b
schematically in Fig. 11.
Bitter’s model involves more material properties than a . a0
Finnie’s and includes those of the particle. However,
Bitter’s model still predicts zero wear at a zero impingement with the provision that Wb ¼ 0 when V~ sinðaÞ , Vel :
angle. In fact, the erosion rate is zero when a , In addition to a threshold velocity normal to the eroding
~ which is greater than zero. Bitter’s model,
sin21 ðVel =lVlÞ; surface, Vn ¼ Vel there is a threshold velocity parallel to the
like Finnie’s, also assumes that the particles do not erode. eroding surface, Vp given by
The application of Bitter’s model to calculation of wear in
fluidized beds, using hydrodynamic results computed by the Vp2 ¼ V~ 2 cos2 ðaÞ½sinðnaÞ 2 1 ð3:12Þ
FLUFIX/MOD2 computer code [29], would be essentially
the same as that for Finnie’s model (Section 3.1.1). where n is an empirical constant and a0 ¼ p=2n: Substi-
tution of Eq. (3.12) into Eq. (3.11a) results in
3.1.3. Neilson and Gilchrist’s combined ductile and brittle
W ¼ Wd þ Wb
erosion model
Neilson and Gilchrist [62] simplified Bitter’s combined M V~ 2 cos2 ðaÞsinðnaÞ M½V~ sinðaÞ 2 Vel 2
¼ þ ð3:13Þ
model by postulating a simplified ductile erosion model 2fc 21b
while retaining Bitter’s brittle erosion model (Eqs. (3.6a), a # a0
(3.6b), (3.7a) and (3.7b)). The result is given by
Comparison of the first terms of Eqs. (3.13) and (3.11b) with
W ¼ Wd þ Wb Bitter’s ductile erosion model, Eqs. (3.9a) and (3.9b),
reveals the extent of Nielson and Gilchrist’s simplifications.
M½V~ 2 cos2 ðaÞ 2 Vp2 M½V~ sinðaÞ 2 Vel 2 The first term of Eq. (3.11b) is the same as Bitter’s ductile
¼ þ ð3:11aÞ
2fc 21b erosion model (Eq. (3.9b)) with K1 ¼ 0: The second terms of
Eqs. (3.13) and (3.11b) are the same as Bitter’s brittle
a # a0 erosion model, Eqs. (3.6a) and (3.6b).
3.1.4. Sheldon and Finnie’s 908 brittle erosion model where NIM is the impact number given by:
Brittle materials, such as ceramics or glass, cannot
deform plastically; instead, they crack and fracture when NIM ¼ ð5p2 =2Þrp V~ 2 ðkp kt Þ1=2 ½ðkp kt Þ1=2 þ ðkt =kp Þ1=2 4 ð3:17bÞ
subjected to tensile stress. The angle of maximum erosion
On the assumption that particle motion in a fluidized bed is
for brittle materials is near 908.
random, Soo averages Eq. (3.16) over all directions and
Sheldon and Finnie [63] analyzed brittle erosion
magnitudes to obtain
occurring at 908. Their final result for spherical particles is
given by E_ dFB ¼ ð1 2 0:9586K pd Þrp ðV 2 Þ3=2 Cd f ð1 þ rp Þð2:94Þ
W ¼ Ce Rjp V h ð3:14Þ pffiffi
ð5=16Þ½2=ð3 pÞhd =1d ð3:18Þ
where
where the overscore denotes averaging. This erosion rate
j ¼ 3f =ðf 2 2Þ ð3:15aÞ given by Eq. (3.18) is no longer a function of angle, and V 2
(the intensity of random motion) replaces V~ 2 in the
h ¼ 2:4f =ðf 2 2Þ ð3:15bÞ
impaction number, Eq. (3.17b). Because particle motion in
and a fluidized bed is not completely random due to the presence
of dead zones and jetting regions, for example, Eq. (3.16)
Ce ¼ Et0:8 s2b ð3:15cÞ may be preferred over Eq. (3.18).
In the above relations, Rp is the particle radius, f is the Brittle wear. Soo’s brittle wear model is expressed in
coefficient of friction, and sb is the flexural strength. This terms of
model gives velocity exponents of 3.2, 2.72, and 2.66 for E_ b ¼ sinðaÞ½1 2 Kbp sinðaÞ21=5 rp V 3 Cb ð1 þ r p Þð2:94Þ
f ¼ 8; 16.9, and 20 (glass, graphite, and hardened steel),
respectively. ð5=16Þhb =1b ð3:19Þ
3.2. Fluidized-bed erosion models where Cb, 1b, and hb are the analogous terms for Cd, 1d, and
hd in the ductile erosion model and Kbp is given by
Single-particle erosion models have been the subject of
much more research than fluidized-bed erosion models. Kbp ¼ 6p2 sb ðkp kt Þ1=2 =½22=5 Cb ð1 þ r p Þ1=5 NIM
1=5
ð3:20Þ
Only two fluidized-bed erosion models have been found in By averaging overall directions and magnitudes, as in the
the literature; they are summarized in this section. ductile erosion model, Soo obtains the following fluidized-
bed brittle erosion model:
3.2.1. Soo’s ductile and brittle erosion models
Soo [64] extended his treatment of heat transfer and E_ bFB ¼ ð1 2 0:8981K pB Þrp ðV 2 Þ3=2 Cb ð1 þ rp Þð2:94Þ
charge transfer by impact to a treatment of material removal
pffiffi
in the case of small deviations from elastic impact. ð5=16Þ½2=ð3 pÞhB =1B ð3:21Þ
Conceptually, Soo’s models resemble Bitter’s models in
that the energy expended to remove material must exceed Clearly, Soo’s erosion models resemble those that have been
the yield stress in order for ductile or brittle failure to discussed previously. However, there are differences in the
produce wear. Soo’s erosion models treat ductile wear and details of the exact angular dependence and the way in
brittle wear separately. which the material properties enter.
Ductile wear. Soo’s ductile wear model is expressed in Results from Soo’s erosion models have not been
terms of compared with experimental data. Soo estimates that, in
the case of 700-mm dolomite particles eroding 316 stainless
E_ d ¼ cosðaÞ½1 2 Kdp sinðaÞ21=5 rp V~ 3 Cd f ð1 þ rp Þ steel, 1 2 Kdp . 102 ; f . 0:1; and hd Cd . 1024 : Assuming
that 1 þ r p < 1; the ductile erosion estimated using Soo’s
ð2:94Þð5=16Þhd =1d ð3:16Þ model (Eq. (3.18)) is as follows:
where V~ is the particle velocity; Cd is a correction factor for 0:69 £ 1027 rp ðV 2 Þ3=2
non-sphericity (< 1); r p is the ratio of particle rebound to E_ dFB ¼ ð3:22Þ
21d
approach velocities, V2 =V1 ; 1d is the energy required to
remove a unit volume of ductile material; and hd is This result is five orders of magnitude lower than E_ FM ; the
the mechanical efficiency of impact (, 1024). (Soo’s 1d maximum erosion rate given by Finnie’s model (Eq. (3.5)).
plays the same role as that of fc in Bitter’s ductile erosion Models of ductile erosion of tubes in FBCs are probably
model.) The dimensionless resistance parameter, Kd, is more appropriate than brittle erosion models.
given by Soo’s model can be used with FLUFIX/MOD2 [29]-
computed hydrodynamic results by replacing rp with
Kdp ¼ 6p2 sd ðkp kt Þ1=2 =½22=5 fCd ð1 þ r p Þ1=5 NIM
1=5
ð3:17aÞ (1 2 1 )rs and V~ 3 with l~vs lð~vs Þ2 :
3.2.2. Wood and Woodford’s fatigue erosion model or 6 £ 10220 m3 for 15 £ 1026 kg of particles. Using
From their studies of tube erosion in fluidized beds, rt ¼ 8.2 £ 103 kg/m3, Eq. (3.27) yields 1.3 £ 10210. This
Wood and Woodford [65] concluded that damage is not a value is in the same order of magnitude as the measured
cutting phenomenon but is more like a local fatigue specific erosion rate (deduced from independent measure-
phenomenon. In basic studies of abrasion [66], a similar ment of N_ ¼ 500 s21 ):
mechanism has been identified.
1:3 £ 10212
The starting point of Wood and Woodford’s model is Sexp ¼ ¼ 2 £ 10210 ð3:29Þ
Hutching’s model [67] involving constant indentation 500 £ 15 £ 1026
pressure acting on a rigid particle, impacting a plastically
deforming material that has no elastic recovery. The Wood
and Woodford model is given by 4. Previous efforts to couple fluid mechanics and erosion
models
bMp N_ Up2
E_ ¼ ð3:23Þ
Nf p A 2
The three major categories of single-particle erosion
where b is a fraction of the indented volume, V; Mp and Up models discussed in Section 3—ductile (cutting), brittle
are the mass and velocity of the particle, respectively; Nf is (deformation), and fatigue (repeated low-frequency particle-
_ is the impact
the number of cycles to fatigue failure; and N=A surface interaction)—all establish a relationship between
rate per unit area. This impact rate is given by: the erosion rate and the particle kinetic energy. Beyond this
similarity, the models in different categories differ in terms
N_ 1 6ð1 2 1Þ of the detailed relationships (involving material properties)
¼ Up ð3:24Þ
A 2 pdp3 that enter into the erosion models’ use of all or a portion of
the kinetic energy of the particles.
where dp is the particle diameter. According to Wood and
Portions of other models may be useful in modifying or
Woodford [65], the factor of 1/2 in Eq. (3.24) accounts
extending the single-particle models, as well as the power
approximately for the fact that the expected impact
and energy dissipation models discussed in Section 5. This
velocity is a fraction of the mean velocity of all particles.
is particularly true in regard to material properties and
As in Hutchings’ model, the indentation pressure, p, in
relationships between them and the particle energy. For
Eq. (3.23) could be given by p ¼ C sy ; where sy is the
example, under certain circumstances particle impact and
uniaxial yield stress. Because particle velocities were not
rebound velocity data may be used to estimate the portion of
measured in the Wood and Woodford study, they used a
the particle’s kinetic energy transferred to the surface;
simple (and possibly incorrect) model for the solids
which, in turn, determines the material removal [69,70].
velocity and generated curves of normalized erosion rates,
_ b=Nf p: Such information is needed directly in several of the
E=½
models.
In order to justify their model, Wood and Woodford
The major shortcoming of all the models discussed thus
assumed the Coffin –Manson equation
far is that they incorporate the macroscopic or nominal fluid
!2
1f mechanical properties of particle velocity and impact angle,
Nf ¼ ð3:25Þ rather than local (or differential) properties. They also fail to
2D1p
account directly for particle fragmentation (or attrition) and
where 1f is the tensile stress at failure and D1p is the plastic concentration of particles in the fluid stream. An exception
strain range. They set 1f/D1p ¼ 1 and b ¼ 1: They then is the model developed by Gansley and O’Brien [69] which
substituted for p from uses Davidson’s bubble model [26] in conjunction with the
Finnie erosion model [55]. Finnie stated in 1972 [57] that no
p¼ 1
2 Mp Up2 =Vc ð3:26Þ satisfactory explanation of concentration dependence
existed, and that very little had been published concerning
where Vc is the crater volume used to obtain the specific
the effect of the carrier fluid itself. In 1990, Humphrey [49]
erosion rate, S, equal to the ratio of mass loss per unit area,
_ : addressed some of these problems in his review article
m=A;
_ to mass flux of incident particles, Mp N=A
where he also reviewed his group’s research on coupling
m=A
_ E_ rt 4r V fluid mechanics and single particle erosion models for dilute
S¼ _ ¼ _ ¼ t c ð3:27Þ flows.
Mp N=A Mp N=A Mp
Knowing that the local erosion rate depends mostly on
Then they calculated Vc from an independent experiment the local velocity, Wolak et al. [71] measured the velocities
[68], in which they observed a crater diameter, d, of 6.4 mm of 60-mesh (250-mm) SiC particles exiting nozzles located
for 1.9-mm silica sand eroding A286 steel as at various distances from a planar target for different loading
! ratios. They found that (1) the particles accelerated as they
p 3 d
Vc ¼ d ð3:28Þ left the nozzle, (2) the particles reached a maximum velocity
48 dp some distance from the nozzle, and (3) the particle velocity
also varied in the radial direction. The implication is that the Meng and Ludema [52] is not recommended, because they
macroscopic erosion models may predict erosion patterns may be highly system-dependent and not mechanistic. We
incorrectly because of local variations in the particle energy. believe that a more fruitful avenue is to couple the fluid
Using dimensional analysis, Tsai et al. [72] identified mechanics with erosion models used locally.
seven dimensionless groups, listed in Table 2, considered to
be important in excessive wear. Several of the groups 4.1. Erosion in turbomachinery
include information about fluid mechanical variables, such
as the boundary layer thickness, d; carrier fluid viscosity, mf; The coupling between fluid mechanics (also termed
concentration of particles, cp; and relative velocity between hydrodynamic modeling) and erosion modeling is fairly
particles and carrier fluid, Vr. None of these variables is recent and coincides closely with the increasing interest in
included in the erosion models discussed in Section 3. burning coal to reduce the US dependence on imported oil.
The last group in Table 2 accounts explicitly for the Long before the FBC erosion issue developed (Section 1.2),
hardness of the particles. The work of Tsai et al. [72] erosion problems were encountered in turbomachinery
suggests that at least for slurries, over a fairly wide range of components. The success achieved in coupling fluid
variables, the overall dependence of the particle and target mechanics and erosion models to predict erosion in this
hardnesses on erosion is approximately E_ / S1=2 p =St ; where and other areas is described in this section.
Sp is the particle hardness and St is the target hardness. The Tabakoff and co-workers at the University of Cincinnati
dependence on particle hardness appears to disappear when have been engaged in turbine blade erosion and particulate
Sp . St [51, p. 105]. flow research since 1971 [70,73– 75]. It is their hope that the
The use of purely empirical erosion correlations, such as incorporation of erosion into the engine design as a
those in Tsai et al. [72] and those chosen for special study by parameter could lead to the production of an erosion-
tolerant engine. They have developed a steady-state
computer program capable of describing three-dimensional
Table 2 particle trajectories through turboaxial or radially rotating
Fluid-mechanical parameters important in erosiona
turbomachinery. Tabakoff has reviewed this group’s work
Term Name Definition up to 1990 [70], the same year Humphrey reviewed the
subject of erosion in turbomachinery and related phenomena
dp rf Vr =mf Particle Reynolds Relative measure of inertial-to- [49].
number viscous effects for the particles In his earlier review of the erosion literature, Tabakoff
in suspension concluded [75]
d/dp ‘Impact-cushioning’ Argued to be a measure of
number the extent to which viscous
effects cushion particle impact …one may conclude that in the investigation of
on the eroded surface erosion there have been three predominant approaches
rp/rf ‘Particle inertia’ Relative measure of particle taken. The first method involves making assumptions
number phase to fluid phase density about the erosive process and the introduction of
(characterizes relative inertial parameters which make the proposed theory conform
effects of the two phases) to experimental results. The second method is one in
Vr/Vp ‘Slip’ number Ratio of relative particle
which the dynamic forces acting between the particle
velocities (characterizes slip
and the material surface are considered, along with
velocity effects)
cp/rp ‘Energy dilution’ Argued to be a measure of the well-known material properties. In this scheme
number extent to which the fluid phase assumptions are made as to the condition of the
dilutes the kinetic energy con- material during the impacts. The third method is to
tent of the particle phase in assume that there is no common material property that
suspension can be used to describe erosion. In this method a
St/rpV2p ‘Barrier’ Relative measure of the energy hypothetical erosion resistance property is invented
number barrier that suspended particles and different materials are related by this property.
must overcome on impact to
create conditions that are
propitious for erosion The group’s own experimental data on the coefficient of
Sp/St ‘Disorder’ Postulated as an indication of restitution (the ratio of rebound to approach velocities) and
number the tendency for erosion to the ratio of rebound to impingement angle are obtained
occur on the impact of a particle experimentally by use of high-speed photography or laser
with sufficient energy to cause
doppler anemometry and are expressed in terms of
erosion
impingement angle. In the computer model first developed
a
Source: Tsai et al. [72]. See Nomenclature for symbol in 1973 – 1974, these ratios are used to account for
definitions. momentum loss of the particles caused by collision with
the turbine blades or channel walls in the Lagrangian energy in the gas phase. The momentum equations for the
trajectory calculations [73]. incompressible, two-dimensional viscous Eulerian gas
The major force on the particles is given by a drag phase include a momentum sink term due to particle drag
expression similar to the one used in the FLUFIX/MOD2 and, thus would appear to be an improvement over
computer code [29]. The added mass, Basset, and Bagnold Tabakoff’s equation; however, the model does not contain
forces on the particles are neglected. The compressible, non- the void fraction as a variable. Hence, the assumption of low
viscous, steady-state gas-phase momentum equations are particle loadings is again implicit, and the approach is
assumed to be unaffected by the particles. This implies low similar to the ‘dusty gas’ model of Rudinger and Chang
particle loadings, an assumption that allows the use of [79]. The Eulerian gas flow equations and Lagrangian
existing single-phase computer codes for turbomachinery. particle trajectory equations are solved iteratively, using an
The three-dimensional Eulerian gas flow and Lagrangian extension of the TEACH program developed at Imperial
particle trajectory equations are solved on a square grid with College, London [80]. Included in the PSI-Cell technique is
the coordinates fixed on the rotating blade (one row of a two-parameter turbulence model and heat transfer; but it
blades is solved at a time). appears that they were not used in this erosion prediction.
This same group has also simulated erosion with a Monte The particle trajectory calculations use empirical
Carlo technique [74]. The erosion model used to compute rebound angle and coefficient of restitution data in a manner
the erosion rates also incorporates the particle rebound data; similar to that of Tabakoff [70]. The erosion model is very
it is given by a semi-empirical equation that relates the mass simple and is given by
of material removed (in mg) to the mass of particles (in g),

as follows mass removed
. f ðaÞV 2:35 ð4:3Þ
~2 mass abrasive
S ¼ K1 f ðaÞV cos2að1 2 R2T Þ þ K3 ðV~ sinaÞ 4
ð4:1Þ
where where f ðaÞ is given by an experimental curve. The exponent
on the particle velocity is taken from Sheldon and Finnie’s
RT ¼ 1 2 0:0016V~ sin a ð4:2aÞ 908 brittle erosion model [63]. The erosion curve for f ðaÞ
and was determined for impingement angles as low as 48 by
using hardened steel shot (Ro45) of 270-mm average
f ðaÞ ¼ 1 þ K12 sin½ð90=a0 Þa 0 , a , 2a0 ð4:2bÞ diameter striking 6061-T6 aluminum alloy, a ductile
f ðaÞ ¼ 1 a . 2a0 ð4:2cÞ material. The rebound data were obtained from multiple
flash exposures of 3.175-mm ball bearings striking the same
where a0 is the angle corresponding to maximum erosion. aluminum alloy. Erosion data taken by blasting the steel
The parameters 10, K1, K12, and K3 are material-dependent shot through a horizontal tube 4.95 mm in diameter and
empirical constants. Comparison of Eq. (4.1) with Eqs. 30.5 cm long, made of the same aluminum alloy, were in
(3.11a) and (3.11b) show that this model is a variant of reasonable agreement, considering that the majority of
Neilson and Gilchrist’s combined ductile and brittle erosion particles were incident at less than 58 (probably deduced
model. The approach velocity, V; ~ and impact angle, a, are from the trajectory computations), which corresponds to
computed from the particle trajectory computer program. almost pure abrasion (scouring). Whether a coefficient of
Private discussions with John Stringer of the Electric proportionality was introduced into Eq. (4.3) is uncertain.
Power Research Institute (EPRI) indicate that Tabakoff’s
methodology for predicting erosion was considered for FBC 4.3. Erosion in curved elbows
applications [76]. Because the particle loading is so much
higher in an FBC than in turbomachinery, the particle In 1983, Pourahmadi and Humphrey [56] published a
trajectory calculations would have to include particle – more sophisticated steady-state, two-dimensional hydro-
particle interactions and erosion simulation using the Monte dynamic model to predict erosive wear in straight channels
Carlo technique; these calculations would be very time- and curved ducts. In that article, the dilute particle loading
consuming, even if they were valid. In addition, the gas assumption is again made but, this time, the particle
phase fluid mechanics code would not be appropriate, momentum equation is formulated following the Eulerian
because the effect of the particles on the gas would be approach (treating the particles as a continuum). Thus, the
significant. void fraction becomes an explicit variable in this model. The
standard two-parameter turbulence model is used for the
4.2. Erosion in horizontal tubes turbulent viscosity of the gas phase. A turbulent-particle
diffusion correlation is used, as well as a single-parameter
Sheldon et al. [77] developed a numerical method to turbulent-particle viscosity model. Stokes’s drag law is
predict erosion on a horizontal, round tube wall. It is known employed in both the gas and particle momentum equations.
as particle-source-in-cell (PSI-Cell), [78] in which the Both phases are assumed to be two-dimensional, steady-
particles are treated as sources of mass, momentum, and state, incompressible, and isothermal. The solution
procedure is also given by an extension of the TEACH Therefore, a new model—termed the power dissipation
code [80]. model—was developed.
The erosion model used is given by the single-particle
Finnie model, expressed by Eq. (3.2a), with f ðaÞ given by 5.1.1. Motivation and origins
Eqs. (3.4a) and (3.4b). The flow pressure, p, is replaced by The particle grinding (comminution) and abrasive
the Vickers hardness. The square-cross-section, curved erosion models provide the genesis of the so-called power
perspex-plastic-duct erosion experiments of Mason and energy dissipation erosion model. As long ago as 1885, Kick
Smith [81] were analyzed for two cases: radius of curvature [83] postulated that the energy, U, required to fracture
to channel widths of 5 (strong curvature) and 12 (mild particles in ball mills is directly proportional to the particle
curvature). The particles used were 55-mm diameter volume and independent of the number and size of the
alumina. Although the relative erosion pattern predictions particles. This postulate, known as Kick’s law, can be
_ E_ max ; where E_ max is the maximum erosion rate, were in
of E= expressed as
generally good agreement, the absolute magnitudes of E_ U ¼ K1 Vp ð5:1Þ
were described as inaccurate by Pourahmadi and Humphrey
[56]. They admitted that the very low values of erosion that where Vp is the volume of particles being ground in the ball
were measured experimentally had been predicted as higher mill and K1 is a constant of proportionality having units of
by their model. However, they believed that the relative pressure [84].
comparisons did have value for design purposes and that the Rabinowicz’s much later expression [66,85] for two-
model is useful for understanding the controlling par- body abrasive (sandpaper or scouring) erosion of a surface
ameters. The factor c in the erosion model, described as the by particles is given as
‘fraction of particles cutting in an idealized manner,’ is Vt
never quantified, nor is the Vickers hardness of perspex ¼ kF=ð3HÞ ð5:2Þ
L
plastic.
where Vt is the volume of target removed, L is the distance
traveled by the particle, F is the applied load (force), H is the
hardness of the surface, and k is a dimensionless adjustable
5. Power and energy dissipation erosion models constant (sometimes called the abrasive wear coefficient
[66]) related to the average angularity of the abrasive
The origins, derivation, and numerical evaluation of the particles. The equivalence of Eqs. (5.1) and (5.2) is easily
power dissipation and energy dissipation erosion models are seen by rewriting them in terms of Vp and Vt as
described in this section. We show that the conceptual bases Vp ¼ U=K1 ð5:3aÞ
for this approach can be traced back to empirical laws of size
reduction, grinding and comminution using a consistent and
nomenclature. We attempt to unify the concepts of ductile, Vt ¼ kFL=ð3HÞ ð5:3bÞ
brittle, abrasive, and impaction erosion. In effect we extend,
in a rational manner, single particle erosion models to The factor of 3 in Eq. (5.3b) relates the yield strength, sb, to
continuum erosion models. The power dissipation model is hardness, H, by sy ¼ 3H: FL is an energy (force times
shown to be a special case of what we call the energy distance). Comparison of Eqs. (5.3a) and (5.3b) reveals that
dissipation model. The MED erosion model, which is the FL is equivalent to U, K1 is equivalent to ð3=kÞH; and Vt is
culmination of our study of power and energy dissipation equivalent to Vp. Thus, the two seemingly dissimilar
erosion models is described. processes of grinding of particles in ball mills and abrasive
wear of a surface by particles are describable by the same
relationship.
5.1. Power dissipation erosion model In their high-speed-photography slurry studies, Shook
et al. [86] observed that the particle motion was parallel to
The concept that energy dissipation of the particles the pipe wall in the region of high concentration where the
impinging on a surface gives rise to deformation wear was erosion rates were high. Hence, the simple Rabinowicz
postulated as long ago as 1963 by Bitter [60]. It was, abrasion relationship holds. A heuristic generalization of
however, the slurry erosion literature of Usimaru et al. [82] this relationship is made in order to compute the volumetric
which provided us with the beginning point of our research _ caused by the flow of a
loss-per-unit-time per unit area, E;
into developing a mechanistic erosion model which _
slurry. Using Eq. (5.3b), E is given by the following for a
culminated with the monolayer energy dissipation (MED) constant applied force, F:
(Section 6.1) and simplified closed form MED (SCFMED)
E_ ¼ V_ t =At ¼ kF L=ð3A
_ t HÞ ð5:4aÞ
erosion models. Because the angle of approach of the
particles eroding a slurry pipeline is essentially zero, none of where V_ t plays the same role as W
_ in Eqs. (3.2a) and (3.2b)
the models described thus far appeared to be useful. in the Finnie ductile erosion model described in Section
3.1.1. We replace the time rate of change of length, L; _ by the the energy dissipation to the eroding target material
velocity of the solids, ~vs ; and the applied load, F, by the force pressure, Pt, through
of the solids on the material surface, F ~ s ; in Eq. (5.4a) to obtain
dKEsPD dV
·Vf ¼ Pt t ¼ Pt At E_ t ð5:10bÞ
~ s ·~vs =ðAt HÞ
E_ ¼ V_ t =At ¼ CF ð5:4bÞ dt dt
where C is a factor that includes k=3 (to account for the fact where dVt/dt is the volumetric time rate of change of the
that not all the volume, Vt, is removed) [66]. The force of the eroding target material. Eq. (5.10b) states that the total
solids on the surface can be computed from solids kinetic energy dissipated equals a solid pressure times
a displacement. The erosion rate of the target, E_ t ; is then
~ s ¼ 2ð1 2 1Þrs d~vs Vf
F obtained from
dts 2 3
dKEsPD
›~vs 6 7 V
¼ 2ð1 2 1Þrs þ ~vs ·7~vs Vf ð5:5Þ 4 dt 5 f
›t E_ t ¼ ¼ 2E_ PD ð5:11Þ
Pt At
where Vf is the volume of the fluid and d/dt s is the total
derivative following the solids velocity, ~vs : Combining Eqs. where dKEsPD/dt, the solids kinetic energy, plays the role of
(5.4b) and (5.5), we obtain U in Kick’s law (Eq. (5.1)). Comparison of Eqs. (5.11) and
(5.9) shows that Pt plays the role of Esp, and the constant
d~v V C ¼ 1: The power dissipation erosion model has been
E_ ¼ 2Cð1 2 1Þrs ss ·~vs f H ð5:6Þ
dt At derived heuristically in this section by generalizing the
Eq. (5.6) is the basic form of the ‘power dissipation’ erosion empirical abrasive erosion relationships. The more general
model proposed in 1984 by Ushimaru et al. [82]. The factor C energy dissipation model will be derived by more funda-
was not explicitly noted but was used, as will be discussed mental means in Section 5.2. The power dissipation model
shortly. Ushimaru et al. used a ‘layer thickness’ D, which can will be shown to be a special case of the energy dissipation
be associated with Vf/At. Defining the solids mass flux to be model in Section 5.2.
q~s ¼ ð1 2 1Þrs ~vs D; the power dissipation, P, was defined by Ushimaru et al. [82] implemented this apparently
Usimaru et al. [82] as: fundamental power dissipation erosion model to analyze
steady-state scoring erosion in a slurry jet pump and
d~vs obtained reasonable comparisons with data. They used a
P¼ ·~q ð5:7Þ
dt s hydrodynamic code similar to that used by Sheldon et al.
Replacing the hardness, H, by a ‘specific energy,’ Esp, the [77] (a variant of the TEACH code [80]).
erosion rate for the power dissipation model, E_ PD ; becomes The fluid flow equations used earlier [77] were extended
to handle dense solid – liquid slurries. The Eulerian particle
d~v
E_ PD ¼ 2CP=Esp ¼ 2C ss ·~qs =Esp ð5:8Þ momentum equation was extended to treat the particles as a
dt continuum of solids. The standard two-parameter turbulence
E_ PD is positive, because the particles must decelerate in order model was used for the liquid phase. A boundary-fitted
to the force of the solids on the surface, F ~ s to be positive Eq. coordinate transformation technique was used to generate
(5.5). The sign differs from that of Ushimaru et al. [82] the computational mesh. The other features of the model
because they obviously considered their symbol, e, for the closely resemble those of Pourahmadi and Humphrey’s [56].
erosion rate, to stand for E_ t ¼ E,
_ the erosion rate of the target The computed results of the two-dimensional, incom-
itself (Eq. (3.2b)). pressible, steady-state code were used as inputs to the power
Eq. (5.6) states that the erosion rate is given by the total dissipation erosion model. This erosion model was
differential of the particle kinetic energy, multiplied by the implemented in one dimension next to the boundary, to
mass flow of particles and divided by some material compute the erosion rate of the slurry jet pump walls.
property resembling or related to hardness. Hence, we will The steady-state one-dimensional form of the power
write Eq. (5.6) symbolically as follows dissipation model is given by

C dKEsPD Vf ›U
E_ PD ¼ 2 ð5:9Þ E_ PD ¼ 2Cð1 2 1Þrs Us Us s D=Esp ð5:12aÞ
Esp dt At ›x
where the rate of kinetic energy dissipation per unit volume With reference to Fig. 12, the finite-difference approxi-
of the solids (which for brevity’s sake we will call the mation to Eq. (5.12a) for erosion of a horizontal surface is
energy dissipation rate) is defined as given by

dKEsPD ›~vs Us2 2 Usl
¼ ð1 2 1Þrs ~vs · þ ð~vs ·7~vs Þ ¼ EPD ð5:10aÞ E_ PD ¼ 2Cð1 2 10 Þrs Usl USl Dy=Esp ð5:12bÞ
dt ›t Dx
An alternative and shorter derivation, using an energy where Dy has been used for D, the layer thickness, and the
balance similar to that used by Bouillard et al. [87], relates donor cell formulation has been used for Uso.
Finnie’s erosion model can be written as

C
E_ F ¼ ð1 2 1Þrs l~vs l~v2s Dc f ðaÞ=p ð5:14Þ
Dc
where Dc is the depth of the cut. Now the linkage can be
made explicit. In Eqs. (5.15a) –(5.15c), the left-hand side
represents Finnie Model and the right-hand side represents
Power Dissipation Model.
ð1 2 1Þr~s l~vs l~v2s f ðaÞ dKEsPD
$ 2ð1 2 1Þrs ~v2s ·7ð~vs Þ ¼ 2
Dc dt
Fig. 12. Finite-difference region for power dissipation erosion
ð5:15aÞ
model.
Dc $ Vf =At ð5:15bÞ
and
Ushimaru et al. [82] applied an empiricism that indicates p $ Esp ð5:15cÞ
that, when metal is removed with sandpaper, less than 10% of
the grains in contact with the surface actually remove metal The differential form of the solids kinetic energy replaces
[88]. The remaining particles cause only elastic deformation, the magnitude of the algebraic solids kinetic energy times an
which does not result in material wear. Hence, the factor C angular dependence, divided by a depth of cut. The ratio of
was taken to be 0.1. As discussed above and in Section 3, such fluid volume to target area replaces the depth of cut, and the
factors are influenced by the hardness of the erodent relative specific energy replaces the flow pressure. There is an
to the eroding surface and the erosion mechanism, cutting implicit fluid-mechanical angular effect in the differential
tool (impaction) or sandpaper (abrasion). form of the kinetic energy. Function f ðaÞ may be viewed as
We consider the results of Usimaru et al.’s [82] a part of the material-property relationships (as well as Dc),
computations fortuitous since the kinetic energy dissipation because it results from integrating the equations of motion
increased through the slurry jet pump as the solids were for the particle gouging the eroding wall.
accelerated. Had the solids decelerated, the kinetic energy
dissipation would decrease and the erosion rate would have 5.1.3. Extension to account for particle size and threshold
the wrong sign. Erosion can occur for both accelerating and energy dependence
decelerating solids flows and therefore the power dissipation As was shown in Sections 5.1.1 and 5.1.2, the power
model derived in this section and used by Usimaru et al. dissipation model evolves from the empirical laws of
[82] is not sufficiently general. Humphrey [49] discusses comminution and abrasion, and can be linked to algebraic
additional particle – surface interactions and related single-particle and fluidized-bed impaction erosion models.
phenomena which could enter into the factor C. However, the former laws hold only for very small particles,
typically less than 1 mm. Another relationship, Rittinger’s
5.1.2. Linkage to algebraic erosion models law, also goes back more than 100 years [84,89]. It states that
It is possible to develop a link to the algebraic single- the energy required to crush a substance is proportional to the
particle and fluidized-bed impaction erosion models dis- production of new area according to the following relation-
cussed in Section 3. These models rely on particle energy ship
dissipation and energy transfer to the eroding surface. On U ¼ K2 ðA 2 A1 Þ ð5:16aÞ
the other hand, the abrasion erosion models utilize the force
of the particles on the eroding material surface. By explicitly where K2 has the units of N/m. Walker and Shaw [84] showed
computing this force via Eq. (5.5), we showed the basic that Eq. (5.16a) is equivalent to the much more meaningful
equivalence of these two types of erosion models. This form,
equivalence, which arises because both erosion models rely U 2 U0
on the concept of cutting of the eroding surface, has not been ¼ K2 =d ð5:16bÞ
Vp
discussed before in the erosion literature. The only
difference is in the view of how the cutting takes place— where d ¼ A=Vp is the thickness of the layer removed,
by the action of the energy of the particle or by the action of assumed to be the same as the particle. Eq. (5.16b) states that
the force of the particle. We will use the Finnie erosion the energy needed to crush a volume of particles actually
model as an example of the linkage. decreases for larger particles. Rewriting Eq. (5.16b) in terms
The steady-state form of the power dissipation erosion of the volume of particles, Vp results in
model can be written as: U 2 U0
Vp ¼ ð5:17aÞ
C V K2 =d
E_ PD ¼ 2 ð1 2 1Þrs ~v2s ·7ð~vs Þ f ð5:13Þ
Esp At Using the analogy between Kick’s law and Rabinowicz’s
abrasion expression (Eqs. (5.3a) and (5.3b)), we replace flow to occur before a flaw is encountered. Thus, materials
U 2 U0 by FL 2 (FL)0 to obtain like marble, which are brittle in the bulk, are ductile in the
FL 2 ðFLÞ0 size range of ball mill grinding (1 –1000 mm), and follow
Vp ¼ ð5:17bÞ the same laws that apply to the grinding of steel by belts
K2 =d
and wheels.
In this form, Rittinger’s law looks like Rabinowicz’s abrasion Such behavior has also been observed in later,
model with a hardness that depends on particle size, ð3=kÞH ¼ independent erosion tests by Sheldon and Finnie [63].
K2 =s; where K2 is a material-property resembling surface They found that, at the same nominal velocity of 152 m/s, 9-
tension and (FL)0 is a sort of threshold energy, which may mm SiC particles eroded glass in a ductile manner, whereas
itself depend on particle size. Expressing Eq. (5.17b) in terms 21 and 127 mm SiC particles exhibited brittle wear. Hence,
of the volumetric loss per unit time per unit area, we obtain the concept of the explicit angular dependence describing
~ s ·~vs 2 ðF ~ s ·~vs Þ0 dp single-particle ductile and brittle erosion may not be
V_ p C½F
E_ ¼ ¼ ð5:18Þ applicable to multiple-particle erosion. What may actually
At A t K2 be occurring is a continuous energy transfer to the surface by
where dp has replaced d. Thus, the erosion rate depends the particles, producing pressures that may be in excess of
directly on the particle size. Substitution of Eq. (5.5) into Eq. the plastic flow limits. This pressure distribution is
(5.18) produces determined by the fluid mechanics of the continuum of
particles, and could be computed from the force of the
d~vs d~vs
Cð1 2 1Þrs ·~
v s 2 ·~
v s dp particles on the surface.
dts dts 0
E_ ¼ 2 0 ð5:19Þ It is possible that particle surface roughness acts very
Esp
much like very small particles; in this case, the erosion
where E0sp ¼ K2 ðAt =Vp Þ is a material-property resembling behavior would follow a ductile model. Most of the surface
hardness and possibly dependent on particle size. damage to the eroding surface is caused by the roughness,
Thus, in this extension of the power dissipation model, much as in a ball mill. This may explain why Wood and
the erosion rate depends on a threshold energy and the Woodford [68] found that dolomite and alundum, which
particle diameter. This threshold energy similar to the differ greatly in hardness, erode tubes at very much the same
threshold energies used by Bitter [60,61] and by Neilson and rate. The erosion experiments with dolomite had to be
Gilchrist [62] in their single particle-impaction erosion terminated after only 20 h, because the filtering system
models. could not handle the fine dust created by the comminution of
We have thus demonstrated, for the first time to our the particles.
knowledge, that there is a strong connection between cutting
or impaction erosion and abrasion erosion. This association 5.1.4. Extension to two dimensions
was made possible by introducing the force of the particles As discussed in Section 5.1.1, the power dissipation
on the eroding surface and by associating this force times a model used by Ushimaru et al. [82] was used in one
displacement with the particles’ kinetic energy. Conse- dimension (finite-differenced next to the eroding wall), even
quently, much of the confusion arising because of the though the solids phase flow is two-dimensional. Hence, the
artificial distinction between impaction and abrasive erosion velocity in the y direction, Vs, shown in Fig. 12 was not used.
is swept away and thereby allows for a rational generaliz- This is justifiable if the solids flow next to the wall in the
ation using the energy dissipation principle. axial direction is much greater than in the transverse
Shook et al. [86] found slurry erosion rates to increase direction (i.e. Us q Vs). Examination of their computed
with particle diameter. The dependence was found to be velocity fields showed this to be a good assumption.
linear for particle diameters over a range of 100– 400 mm. The solids (and gas) velocity fields corresponding to the
Above this diameter, the erosion rate increased hardly at all. two-dimensional fluidized-bed computations reported in
Below 100 mm, the erosion rate fell to zero, which, we infer, Section 2.2 were time-averaged over 2.5 s. This has been
supports the existence of a threshold energy. determined to be long enough for good time-averaging,
Walker and Shaw [84] argue that the important aspect because the gas mass flux exiting the top of the bed differs
of particle size dependence for the grinding energy by less than 1% from that entering the bottom of the bed.
requirements has to do with the probability of finding The solids velocity vector plot in Fig. 13 shows three
flaws in the particles. Very small particles, because they regions of recirculation: one on each side of the obstacle,
have correspondingly fewer flaws, are much more difficult indicating a toroidal vortex (sometimes referred to as the
to break up, and follow Kick’s law. Larger particles have a ‘gulf stream’ pattern), and one above the obstacle with
larger number of flaws, and break up below their solids flowing downward in a nearly stagnant region. The
theoretical strength; so they follow Rittinger’s law. The gas flows upward everywhere. The time-averaged porosity
fact that normally brittle and ductile materials also follow distribution was shown in Fig. 10, where it compared well
Kick’s and Rittinger’s laws is also explained by the size with the data. Examination of the computed time-averaged
effect. Very high-pressure, small particles cause plastic solids phase velocity field next to the obstacle leads to the
Fig. 13. Computed solids velocities time averaged over 2.5 s.
conclusion that it is highly two-dimensional. Hence, the

one-dimensional power dissipation model given by Eq.
(5.12a) is deemed to be inadequate.
The extension of the power dissipation model to two
dimensions is given by

V
E_ PD ¼2C f rs ð121Þ
At

›U ›U ›V ›V
Us Us s þVs s þVs Us s þVs s Esp
›x ›y ›x ›y
ð5:20Þ
For erosion on a horizontal surface (shown in Figs. 12 and

14), Vf =At ¼Dx·Dy=Dx¼Dy: For erosion on a vertical
surface, Vf =At ¼Dx·Dy=Dy¼Dx:
It may be interesting to interpret a somewhat simplified
two-dimensional power dissipation model in terms of
the concepts of impaction and scouring or abrasive wear. Fig. 14. Finite-difference region for two-dimensional power
If the cross-product terms Us(›Vs/›x ) and Vs(›Us/›y ) are dissipation model.
neglected, Eq. (5.20) becomes 5.2.1. Derivation of the general transient energy dissipation
model
V ›U
E_ PD ¼ 2C f rs ð1 2 1Þ Us2 s Esp The solids phase momentum equations given by Eqs.
At ›x
(2.4) and (2.6) with the viscous shear, t SV ; retained may be
written in vector form as
V ›V
2 C f rs ð1 2 1Þ Vs2 s Esp ð5:21aÞ
At ›y ›
½r ð1 2 1Þ~vs þ 7·½rs ð1 2 1Þ~vs ~vs
The first term of Eq. (5.21a) is the model Ushimaru et al. ›t s
[82] used for scouring erosion. We will refer to it as the ¼ 2ð1 2 1Þ7P þ b ð~vg 2 ~vs Þ þ Gð1Þ71
scouring erosion model, E_ SPD : The second term of Eq. þ rs ð1 2 1Þ~g þ 7½ð1 2 1Þt SV ð5:23aÞ
(5.21a) can be interpreted as an impaction erosion model,
E_ IPD : Thus, Eq. (5.21a) can be written as Eq. (5.23a) may be written in non-conservation law form as
E_ PD ¼ E_ SPD þ E_ IPD ð5:21bÞ ›~vs

rs ð1 2 1Þ þ rs ð1 2 1Þ~vs ·7~vs ¼ 2ð1 2 1Þ7P
›t
where
þ bð~vg 2 ~vs Þ þ Gð71Þ þ rs ð1 2 1Þ~g þ 7½ð1 2 1Þt SV

V ›Us ð5:23bÞ
E_ SPD ¼ 2C f rs ð1 2 1ÞUs2 Esp ð5:21cÞ
At ›x
Following Bird et al. [90], the equation of mechanical
and energy, extended to two-phase flow, is obtained by taking
the scalar product of the solids velocity, ~vs ; with Eq.
V ›Vs
E_ IPD ¼ 2C f rs ð1 2 1ÞVs2 Esp ð5:21dÞ (5.23b) to obtain:
At ›y
›~vs
The finite-difference approximation to Eq. (5.21c) is given rs ð1 2 1Þ~vs · þ rs ð1 2 1Þ~vs ·ð~vs ·7~vs Þ
by Eq. (5.12b) for a horizontal wall, and the finite-difference ›t
approximation to Eq. (5.21d) for a horizontal wall can be ¼ 2ð1 2 1Þ~vs ·7P þ ~vs ·b ·ð~vg ·~vs Þ þ vs ·Gð1Þ71
given, by reference to Fig. 14, as
þ rs ð1 2 1Þ~vs ·~g þ ~vs ·½7·ð1 2 1Þt SV ð5:23cÞ
V 2 Vsl
2
E_ IPD ¼ 2Cð1 2 10 Þrs Vs2 Dy s2 Esp ð5:22aÞ
Dy Note that the sign convention of the viscous terms used is
Because Vs1 ¼ 0; Eq. (5.22a) becomes simply opposite to that of Bird et al., and agrees with that used in
the K-FIX computer code [36] and is used in the FLUFIX/
3
E_ IPD ¼ 2Cð1 2 10 Þrs Vs2 =Esp ð5:22bÞ MOD2 computer code [29].
The very last term in Eq. (5.23c) may be split into two
which bears a strong resemblance to Finnie’s erosion model
terms as:
as extended to a continuum of solids, Eq. (3.3a) for m _s . 0
and f ðaÞ ¼ 1: In order to obtain the correct sign for E_ IPD ; Eq. ~vs ·½7·ð1s t SV Þ ¼ 7·½ð1s t SV Þ·~vs 2 ð1s t SV Þ : 7~vs ð5:24Þ
(5.21d) would have to be evaluated right at the eroding
The first term in Eq. (5.24) represents the rate of reversible
surface. Then Dy would become 2 Dy, from the outward
work done by the solids viscous forces, while the second
drawn normal to the surface.
term represents the rate of irreversible conversion to internal
The full two-dimensional power dissipation model given
energy and can be shown to be always by positive [90].
by Eq. (5.20) was differenced, and the erosion rates around
Substitution of Eq. (5.24) into Eq. (5.23c) and solving for
the rectangular obstacle described in Section 2 were
the rate of irreversible conversion to internal energy, EED,
calculated, using the time-averaged porosity and solids
results in
phase velocities. Because both positive and negative values
were calculated for the power dissipation erosion model for dKEED
EED ¼ 2 ¼ ð1s t SV Þ : 7vs
locations around the obstacle, it was concluded that the two- dt
8 ›v
dimensional power dissipation model was incomplete, and >
< 1s rs v s · s
hence inadequate. ¼2 ›t þ 1s rs v s ·vs ·7vs þ ð1s v s Þ·7P
>
: 1 2 3
5.2. Energy dissipation models
~vs ·b ·ð~vs 2 ~vg Þ Gð1s Þ~vs ·I·71 1 r ~v ·~g
þ 2 2 s s s
The power dissipation erosion model was discussed in 4 9 5 6
Section 5.1.1. It is necessary to extend that model to a >
=
two-phase continuum, consisting of both solids and fluid 7·½ð1s t SV Þ·~vs
2 ð5:25Þ
phases. The power dissipation model will be shown to be a 7 >
;
special case of the more general energy dissipation model
developed in this section. where 1s is 1 2 1.
Eq. (5.25) is simply that portion of the solids phase Cartesian coordinates, with Term 7 zero:

› 1 2 › 1 2
mechanical energy equation resulting in the rate of
dKEs
irreversible conversion to internal energy, a portion of ¼ 1s rs U s þ V s
dt ›t 2 ›t 2
which is available for energy transfer to the solid surface
›Us ›U
(i.e. the rate of kinetic energy dissipation per unit volume) to þ 1s rs Us Us þ Vs s
›x ›y
produce erosion. It is a direct extension of the well-accepted
single-phase expression. ›Vs ›Vs
þ Vs Us þ Vs
The seven terms on the right-hand side of Eq. (5.25) ›x ›y
represent the following: ›P ›P
þ 1s Us þ 1s Vs þ Us bx ðUs 2 Ug Þ
›x ›y
Term 1 ¼ rate of increase in kinetic energy; ›1
þ Vs by ðVs 2 Vg Þ 2 Us Gð1Þ
Term 2 ¼ net rate of input of kinetic energy by bulk ›x
flow; ›1
2 Vs Gð1Þ þ 1s rs Vs g ð5:27Þ
›y
Term 3 ¼ rate of reversible conversion to internal
energy plus rate of work done by pressure of surround- where g~ ¼ 2g (gravity negative downward) and ¼
ings on volume element; dKEs =dt ¼ 2EEDKE : The subscript ‘s’ stands for simpli-
Term 4 ¼ rate of work done by drag between the gas and fied kinetic energy dissipation rate per unit volume.
solids on volume element;
5.2.2. Coupling of hydrodynamic and energy dissipation
Term 5 ¼ rate of work done by solids stress on volume
models
element;
A simple methodology has been developed to compare,
Term 6 ¼ rate of work done by gravitational force on on a consistent basis, the energy dissipation rate computed
volume element; from the energy dissipation model and the kinetic energy
Term 7 ¼ rate of work done by viscous forces on volume available for single particle erosion models using as inputs
element. hydrodynamics from the FLUFIX Computer programs [29,
92]. The Finnie erosion model is chosen to illustrate the
Clearly, each term must be present, because each has a methodology. The total available kinetic energy KETOT, is
physical interpretation contributing to the energy dissipa- given by:
tion. Eq. (5.25) constitutes the rational extension of the
kinetic energy dissipation rate for the power dissipation _ s ~v2s =2
KETOT ¼ m ð5:28Þ
erosion model discussed in Section 5.1. where KETOT has dimensions of W/m2.
To see this, Eq. (5.25) is rewritten as In order to compare the energy dissipation rate with the
total kinetic energy defined by Eq. (5.28), the energy
dissipation rate given by Eq. (5.27) is multiplied by the
2EED ¼ EPD þ 1s ~vs ·7P þ ~vs ·b ð~vs 2 ~vg Þ 2 ~vs ·Gð1Þ71 particle diameter to obtain:
2 1s rs ~vs g~ 2 7·½ð1s t SV Þ·~vs ð5:26Þ UEDKE ¼ ð2dKEs =dtÞdp ¼ EEDKE dp : ð5:29Þ

2
UEDKE has also the dimensions of W/m . The particle
where EPD is the rate of kinetic energy dissipation per unit diameter, dp, was argued by Bouillard et al. [37] to arise
volume for the power dissipation erosion model given by from the observation that most of the energy dissipation
Eq. (5.10a). Eq. (5.26) also shows that the sign for the power occurs in a monolayer of particles in the vicinity of the
dissipation rate used by Usimaru et al. [82] appears to be eroding surface as shown in Fig. 15. Hence the ratio Vf/At
incorrect. which arises in the power dissipation model becomes
The model for the energy dissipation used to Vf =At ¼ dp :
compute the erosion rate in the prototype EROSION The maximum amount of the total kinetic energy
computer code described by Lyczkowski et al. [91] available for erosion in the Finnie erosion model is obtained
contains the terms (1 – 6) in braces in Eq. (5.25). The from Eq. (3.5) as
seventh term was not included since the hydrodynamic _ s ~v2s =2Þ ¼ 0:075KETOT
KEFM ¼ 0:075ðm ð5:30Þ
input was from the FLUFIX/MOD1 computer program
[92] which contained neither the fluid nor the solids at an impingement angle of 18.438. The kinetic energy
viscous terms. The prototype EROSION code also available for erosion at any other angle is given by
contained the implementation of the Finnie erosion
_ 2s f ðaÞ
KEF ¼ ð1=8Þm ð5:31Þ
model given by Eq. (3.3a) – (3.3d) with f ðaÞ given by
Eqs. (3.3b) and (3.3c), and the Finnie maximum erosion where f ðaÞ is given by Eq. (3.3a).
model given by Eq. (3.5). Eq. (5.25) becomes, in The energy dissipation model was also time-averaged
and
UEDvREL ¼ ½ð1s t SV Þ : 7~vs þ 1s b B ðvf 2 ~vs Þ2 dp
¼ EEDvREL dp ð5:32dÞ
It has been our experience that the evaluation of Eq. (5.32a)

is subject to more discretisation errors than the evaluation of
Eq. (5.32b), which forms a portion of Eqs. (5.32c) and
(5.32d). This results in both positive and negative values of
the energy dissipation rates as shown by Lyczkowski et al.
[91]. The additional difference equations are described in
the EROSION/MOD1 report [93]. A discussion of the two
ways to evaluate the rate of irreversible conversion to
internal energy, ð1t SV Þ : 7~vs ; is also presented there.
6. Implementation of erosion models in

EROSION/MOD1
The implementation of single particle and differential

erosion models for fluid– solids systems is described in this
section using outputs from the hydrodynamic models
contained in the FLUFIX/MOD2 computer [29].
Fig. 15. Conceptual picture of the MED model.
6.1. Monolayer energy dissipation erosion models
by Lyczkowski et al. [90] to accept time averaged
hydrodynamic inputs from the FLUFIX/MOD1 computer A typical FLUFIX/MOD2 computational cell next to a
code [92]. Both time-averaged and time-averaged transient tube surface is shown in Fig. 16. If the surface is curved, it
energy dissipation rates were analyzed by Lyczkowski et al. must be approximated with a ‘staircase’ approximation, as
[91] for the case of the rectangular obstacle described in indicated because Cartesian coordinates are used. The
Section 2.2. normal components of the fluid- and solids-phase velocities
Eq. (5.29) was the prototype MED model. The MED are set equal to zero on all stationary surfaces. The
model requires temporal and spatial derivatives of the solids tangential component of the fluid-phase velocities are set
phase velocities to compute the energy dissipation rates at to zero and the tangential component of the solids velocities
the cell centers around obstacles. In addition, the gas and are either set to zero or computed from the partial slip
solids phase velocities are required to compute the energy boundary condition [29]. If the stationary surface is aligned
dissipation rate caused by the drag, and gradients of pressure along the coordinate direction, no approximation is
and porosity are needed to evaluate the pressure and solids necessary.
stress, Terms 4 and 5 in Eq. (5.25). A detailed description of The erosion rates for the MED erosion model are given
the finite differencing for the prototype MED model was by:
given in the EROSION/MOD1 report [93].
From Eq. (5.25), it is clear that the terms representing the E_ ED ¼ CUED =Esp ð6:1aÞ
irreversible rate of internal energy conversion, ð1t SV Þ : 7~vs ; E_ EDv ¼ CUEDv =Esp ð6:1bÞ
are equivalent to all seven of the terms in the braces.
Boundary condition ambiquities arise in evaluating these, as E_ EDvCF ¼ CUEDvCF =Esp ð6:1cÞ
discussed by Lyczkowski et al. [91, pages 52 – 55].
The MED model was therefore refined by Bouillard et al. and
[37]. These energy dissipation rate models which are E_ EDvREL ¼ CUEDvREL =Esp ð6:1dÞ
programmed into the EROSION/MOD1 computer program
[93] are given by Eqs. (6.1a)– (6.1d) may be written succinctly as:
UED ¼ ðTerms1 – 7Þ in Eq: ð5:25Þ ¼ EED dp ð5:32aÞ E_ EDa ¼ CUEDa =Esp ¼ CEEDa dp =Esp ð6:2Þ
where UEDa and EEDa are given by Eqs. (5.32a) – (5.32d)

UEDv ¼ ½ð1s t SV Þ : 7~vs dp ¼ EEDv dp ð5:32bÞ and Eqs. (6.1a) – (6.1d), respectively, with a ¼ 1, v, vCF,
and vREL, respectively. Details of the derivation of Eqs.
UEDvCF ¼ ½ð1s t SV Þ : 7~vs þ b B ~v2s =2dp ¼ EEDvCF dp ð5:32cÞ (6.1b)– (6.1d) are given by Bouillard et al. [87]. Esp is
Fig. 16. Coupling of hydrodynamics and erosion models.
the specific energy, which is a material property related to of solids. Thus the rate of energy dissipated during erosion
hardness or flow stress, p. Some typical values of hardnesses represents only a fraction of the total energy dissipation
are given in Table 3 for the materials used by Wood and (which is related to the total entropy production).
Woodford [65] and in Table 4 for data from Usimaru et al. The presently recommended MED erosion model, which
[82] who claim that Esp < 2 £ p. is essentially the one derived by Bouillard et al. [37], as
The MED erosion models given by Eqs (6.1a) – (6.1d) refined by Bouillard and Lyczkowski [87] is Eq. (6.5) with
can be interpreted in either of two ways. They can be UEDa given by Eq. (5.32c) and is given by:
thought of as extensions of the power dissipation erosion
model given by Eq. (5.11) with Vf =At ¼ dp ; or as a E_ MED ¼ ð1 2 e2 Þ½ð1s t SV Þ : 7~vs þ b B ~v2s =2dp =p
generalization of Rittinger’s law given by Eq. (5.19) but
with no threshold energy. We chose Esp ¼ p. Bouillard et al. ¼ E_ EDvCF ð6:6Þ
[87] showed that the factor C is in fact related to the target- In their cold fluidized-bed erosion experiments, Wood and
particle coefficient of restitution, e, given by Woodford [68] found that the erosion rate increased with
2 2 increasing particle diameter. For example, they obtained
Vs1 2 Vs2 ¼ ð1 2 e2 ÞVs1
2
ð6:3Þ
erosion rates of 0.036, 0.48, and 1.16 mm/1000 h for
where Vs1 and Vs2 are the approach and rebound solids aluminum tubes for 100-, 930-, and 1900-mm silica sand;
velocities, respectively, by these values suggest a very nearly linear dependence.
Basically, linear dependence was found for all the other
C ¼ 1 2 e2 ð6:4Þ materials tested. Therefore, the choice of the particle size
2 dependence in the MED erosion models given by Eqs (6.5)
Typically, e < 0.9 [94], which, fortuitously agrees with
and (6.6) appears justified.
the value used in the power dissipation erosion model by
Usimaru et al. [82] which was based on the data of Mulhern
6.2. Neilson – Gilchrist and Finnie erosion models
and Samuels [88]. This is the default value used in
the EROSION/MOD1 computer code [93]. Therefore,
The Neilson– Gilchrist (N– G) erosion model [62] was
the definitive form of the MED erosion model becomes:
chosen to be implemented into the EROSION/MOD1
E_ EDa ¼ ð1 2 e2 ÞUEDa =p ¼ ð1 2 e2 ÞEEDa dp =p ð6:5Þ computer program [93] because it has the capability of
computing erosion at a 908 angle of solids incidence relative
These MED erosion models are based on the premise that to the eroding surface. It has also normal and parallel
the mechanical energy of the solids is irreversibly dissipated velocity threshold components below which erosion does
in the neighborhood of stationary surfaces by three not take place. However, in place of a single value for
competitive mechanisms: (1) heat transfer between the hardness, two material properties are needed to characterize
fluid-and-solids phase, between the fluid-phase and station- cutting and deformation energy absorption.
ary surfaces, and between the solids-phase and stationary As discussed in Section 3.1.3, the N– G erosion model
surfaces; (2) erosion of stationary surfaces; and (3) attrition [62] is a simplification of Bitter’s earlier model [60,61]. It
Table 3
Hardness of materials tested by Wood and Woodforda
Material Crystal structure Diamond pyramid hardness Composition

(kgf/mm2)b
Aluminum fcc 18.5 99.99% Al

Iron bcc 90 Electrolytic Fe (99.95%)
Copper fcc 104 OFHC Cu (99.99%)
Nickel fcc 131 Ni 270 (99.95%)
Cobalt fcc 210 Electrolytic Co (99.92%)
SA213-T11 bcc 0.1 C; 0.44 Mn; 0.65 Si;
1.22 Cr; 0.48 Mo
SS304 fcc 177 18 Cr; 8 Ni; 1.7 Mn;
0.5 Si; 0.05 C
SS316 fcc 171 17 Cr; 12 Ni; 1.7 Mn;
2.0 Mo; 0.5 Si; 0.05 C
A 286 fcc 393 53 Fe; 26 Ni; 15 Cr;
1.3 Mo; 0.2 Al; 2.0 Ti;
1.35 Mn; 0.5 Si; 0.05 C;
0.01 SB
Stellite-6B fcc 377 Co; 29.9 Cr; 2.2 Ni; 1.3
Fe; 1.0 Mo; 4.3 W; 0.48
Si; 1.61 Mn; 0.99 C
High-speed steel bcc 1010 1.2 C; 3.75 Cr; 1.6 V;
2.75 W; 8 Mo; 8.25 Co
Limestone (Cao) 134
Silica sand 766
Alundum (A12O3) 1890
a
Source: Wood and Woodford [65].
b
0.5 kg load.
consists of the sum of a ductile erosion rate, E_ d ; and a brittle with the provision that E_ b ¼ 0 when lV~ el lsinðaÞ , Vel and a0 ¼
erosion rate, E_ b : The N –G erosion model was extended in a p=2n; where n is an empirical constant.
manner consistent with the extension of the Finnie erosion The erosion rates for the Finnie erosion model are
model (Section 3.1.1) by introducing the mass flux of computed from:
particles and is given by
E_ FM ¼ KEFM =p ð6:8Þ
E_ N–G ¼ E_ d þ E_ b
8 and
>
> m_ s l~vs l2 cos2 ðaÞsinðaÞ m _ ½l~v lsinðaÞ2 Vel 2
>
> þ s s a # a0
< 2f c 2jb E_ F ¼ KEF =p ð6:9Þ
¼
>
> _ s l~vs l2 cos2 ðaÞ m _ ½l~v lsinðaÞ 2Vel 2
>m
> For the extended N – G and Finnie erosion models given
: þ s s a . a0
2f c 2 jb above and implemented in the EROSION/MOD1 computer
ð6:7Þ program [93], the solids-phase velocities are resolved in
each coordinate direction at the cell center as shown in
Table 4
Fig. 16 and the magnitude of the solids velocity is obtained.
Typical values of material hardnessa The angle of the resolved velocity vector at the cell center
with respect to each horizontal and vertical approximation
Metal Hardness, P (kgf/mm2) to a curved surface is then obtained. The porosity is known
at the cell center, so the solids mass flux is computed from
Lead 5 the magnitude of the solids velocity resolved at the cell
Aluminum 22 –35 center and the porosity times the solids density (which is
Copper 42 –120 constant). Given eroding material hardness, Esp, or flow
Brass 42 –180 stress, p, and some estimate for the restitution coefficient, e,
Nickel 115 –350
the erosion rate is then computed for the MED, N – G, and
Hardened steel 900
Finnie erosion models. We should note that b B used in the
a
Source: Ushimaru et al. [82]. MED erosion models, Eqs. (6.1a) –(6.1d) or (6.6) are the
drag expression computed for recommended Hydrodynamic contact theory given by Eqs. (3.7a) and (3.7b). Using the
Model B [29,32]. values from Ref. [95], sy ¼ 0:414 £ 109 Pa; rp ¼ 0:244 £
In addition, the ductile (cutting) and brittle (deformation) 104 kg=m3 ; gp ¼ 0:1 ðestimatedÞ; Ep ¼ 0:55 £ 1011
wear factors, fc and jb, published by Nielsen and Gilchrist ðestimatedÞ; and Et ¼ 0:20 £ 1012 Pa; a reduced Young’s
[62] for 210 mm aluminum oxide particles eroding alumi- modulus, Er ¼ 0:14 £ 1012 Pa; and a threshold velocity, Vel ¼
num plate targets, were curve fit in dimensionless form, as 0:056 m=s; are computed and thus V2el can safely be neglected
shown in Fig. 17 (modified from Engel [50]) in the range of with respect to the solids velocity in the N–G experiments.
(100 # V # 200 m/s), as The threshold velocity can also be determined from particle
fc =ðt=rs Þ ¼ 54:21½1 þ expð0:06793ð115:27 2 VÞÞ ð6:10Þ rebound data by using the following relationship [60]
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and V2 ¼ 2V1 Vel 2 Vel2 ð6:13Þ
jb =ðsy =rs Þ ¼ 82:52½1 þ expð0:07942ð110:13 2 VÞÞ ð6:11Þ where V1 and V2 are the particle velocities before and after
where t is the yield shear stress and sy is the yield tensile collision, respectively.
stress (plastic load limit) of the aluminum target ðt ¼ 0:6sy An assessment of the N– G erosion model was carried
and sy ¼ 414 MPaÞ [50]. Since the ductile and brittle wear out using solids velocities typical of those found in fluidized
factors decrease with increasing particle impingement beds. The results given in Lyczkowski et al. [93] indicate
velocity, this may indicate some sort of dynamic softening that the computed erosion rate is strongly dependent on the
of the target. The energy dissipation rates computed from solids velocity. The maximum erosion rate at a specified
the N– G erosion model may be compared with the MED velocity occurs at an angle slightly below a0 ¼ 18:438: The
and Finnie erosion model according to angle at which maximum erosion occurs is obtained by
setting dE_ N – G =da ¼ 0 from Eq. (6.7). Analysis of the
KEN – G ¼ Eq: ð6:7Þ £ jb ð6:12Þ computations shows that the erosion rate is roughly
Since data to determine the wear factors fc and jb for proportional to the solids velocity raised to the power 3.5.
conditions existing in FBCs are lacking, it is recommended In other words, the solids velocity is a dominant parameter
that the values of these two factors evaluated at the lower in determining the erosion rates for the N – G erosion model.
bound of the experimental velocity data (107.8 m/s) be used Comparison of the MED, Finnie, and N– G erosion models
as shown in Fig. 17. If this is done, erosion rates are in rough are given in Lyczkowski et al. [93] for the case of the
agreement with the MED erosion model, as shown by rectangular obstacle described in Section 2.2.
Lyczkowski et al. [93]. N. B: At present, only wear factors
for aluminum target material are in the EROSION/MOD1
computer program [93]. 7. Comparison of erosion predictions for two cold
With a0 chosen to be 18.438, the same as in Finnie’s few-tube FBC geometries
erosion model described in Section 3.1.1, n becomes 4.89
from the relationship a0 ¼ p=ð2nÞ: The wear factors The MED, Finnie, and N – G erosion models are compared
calculated from the empirical formulae given by Eqs. (6.10) in this section for two two-dimensional idealizations of
and (6.11) are in excellent agreement with the experimental the Coal Research Establishment (CRE) 0.3 £ 0.3 m2 cold
data [62] in a velocity range from 100 to 150 m/s. model of the IEA Grimethorpe tube bank ‘C1’ configuration,
Nielson and Gilchrist [62] did not report the threshold Parkinson et al. [96] shown in Fig. 18. In the following
velocity normal to the target, Vel, in their erosion analysis of simulations, the bed consists of glass beads (500 mm
aluminum plates, because they stated that it is usually small diameter) and has a solids viscosity of 0.1 Pa s. Such a
relative to the particle velocity. It can, however, be value is typical of that deduced from bubble shapes in
estimated from the theoretical expression from the Hertz fluidized beds consisting of 550 mm diameter glass beads
Fig. 17. Nielson and Gilchrist wear factors for aluminum oxide particles and aluminum plate target (after Engel [50]).
[97]. The actual cold model consisted of 17 PVC tubes of set 5% lower at 150.76 cm/s. The transient was run to 4.5 s
33 mm outside diameter, three columns wide and eight rows (2.5 s beyond the case of U/Umf ¼ 1.12).
high. The vertical distance between tube row centers was The simulation for the case of U/Umf ¼ 1.7 at 4.5 s was
80 mm and the horizontal distance between tube column used as the initial condition for the case of a still higher jet
centers was 89 mm. More details of the experiment are to be velocity, 2.3 £ Umf (Vjet ¼ 216.21 cm/s). This time, the jets
found in Bouillard et al. [37] and Bouillard and Lyczkowski near the centerline were set to 5% higher at 227.02 cm/s and
[87] who presented some preliminary results. the jets near the sidewalls were set 5% lower at 205.39 cm/s.
The FLUFIX/MOD2 [29] models shown in Fig. 18 The simulation was run to 7.0 s (2.5 s beyond the case of
consist of only two and three rows of tubes. The grid size U/Umf ¼ 1.7). The simulation with three tubes was also run
chosen is Dx ¼ Dy ¼ 1.7 cm with the assumption of to 2.7 £ Umf, while the simulation with five tubes was
symmetry. The number of computational cells in the terminated at U/Umf ¼ 2.3 because of solids being flung to
transverse direction is 9 and 48 in the axial direction, for a the outlet.
total of 432 cells. With this nodalization, the round tubes are
approximated as square obstacles consisting of four nodes 7.1. Hydrodynamic results
each. The perimeter of the square tube is therefore 13.6 cm
as compared with the actual round tube perimeter of Interesting time averaged solids flow patterns are
10.4 cm. The vertical distance between the obstacle centers revealed for U/Umf ¼ 1.2, 1.7 and 2.3 in Figs. 19 and 20.
is 8.5 cm. Solids flow patterns are similar for the two geometries,
The CRE cold model air distributor had 16 standpipes on however, the five tube bed appears to be more quiescent and
a 75 mm square pitch. Each standpipe had 20 round holes, thus will be subject to less erosion as discussed later. The
10 of which were 3.2 mm diameter and 10 of which had a solids are entrained by the gas at the air distributor
diameter of 4 mm for a total open area of 2.06 cm2. The standpipes, converge toward the center of the bed beneath
FLUFIX/MOD2 [29] model standpipes are modeled resem- the upper tube, deflect on the side of this tube, and fall down
bling bubble caps as shown in Fig. 18. These bubble caps are the walls of the fluidized bed toward the distributor. For both
comprised of five obstacle cells in each bed half, each beds, the increase of fluidizing velocity increases the bed
consisting of only one computational cell. The two rows of expansion as well as the strength of solids motion. The time-
obstacles are arranged so that the upper row is situated averaged porosity contours indicate the solids buildup on the
above the two inlets as shown, deflecting the air flow top of the center tube. The reason for the buildup on the
sideways. If the depth of the bed is chosen to be 4.85 cm, the center tube is partly due to the assumption of symmetry
effective flow area of each simulated standpipe is the same which does not allow transverse flow across the centerline to
as in the CRE cold model. The average distance between the sweep solids off the top of the center tube. It was also noted
simulated standpipes is 6.3 cm (3.4 cm on each side and that the bubble formation near the off center tubes alternated
11.9 cm across the centerline). The simulation is performed from one side to the other. This observation leads to the
with glass beads (500 mm in diameter and a density of conjecture that if the symmetry assumption were removed,
2.44 £ 103 kg/m3) and an initial bed height of 44.2 cm so solids would not buildup on top of the center tube to such a
that the obstacles are totally immersed within the bed. large extent as computed.
Hydrodynamic Model B was used [29,37].
For both beds, the superficial gas velocity, U, was 7.2. Bed dynamics
maintained at 23.5 cm/s, just above minimum fluidization
velocity, Umf, equal to 20.9 cm/s (U/Umf ¼ 1.12) for 2 s to First, at each node, the statistical mean, variance, auto
obtain a reasonable initial condition for subsequent runs for correlation, and power spectral density of fluctuation of each
a bed near minimum fluidization (1 , 0.4). The compaction hydrodynamic property, i.e. pressure, porosity, velocities
gas volume fraction, 1 p, was increased to 0.42 to (gas and solids), or axial solids momentum flux, were
accomplish this. computed. Fig. 21 summarizes the statistical means and
Considering that the total open area of the standpipes is variances of the above-mentioned properties around the
2/9 of the total bed area, this means the jet velocity (Vjet) into tubes at 1.12Umf for the three tube simulation. Pressure
each of the standpipes shown in Fig. 18 is 105.81 cm/s. The means are inversely proportional to the node height
jet velocities near the centerline were set 5% higher at reflecting a decreasing bed height. Bubbles appear active
111.10 cm/s and the jets near the sidewalls were set 5% below the tubes where high porosity means and variances
lower at 100.52 cm/s in order to induce sweeping of solids are calculated. Solids are observed to accumulate on top of
off the obstacles near the side walls. tubes where low means and near zero variances are
Several simulations of these fluidized-bed models were computed. Around the tubes, axial gas velocity means are
undertaken. The simulation for U/Umf ¼ 1.12 was used as generally higher than their variances, indicating little chance
the initial condition for the case U/Umf ¼ 1.7 (average of back flow of gas, but most axial solids velocity variances
Vjet ¼ 158.7 cm/s). The jets near the centerline were set 5% are larger than their means, indicating the existence of
higher at 166.64 cm/s and the jets near the sidewalls were strong circulations and back flows of solids. Large axial
Fig. 18. Two CRE cold few-tube FBC geometry FLUFIX models.
R.W. Lyczkowski, J.X. Bouillard / Progress in Energy and Combustion Science 28 (2002) 543–602
Fig. 19. CRE cold few-tube (three tubes) FBC geometry time-averaged porosity and solids velocity distributions.
577
578
R.W. Lyczkowski, J.X. Bouillard / Progress in Energy and Combustion Science 28 (2002) 543–602
Fig. 20. CRE cold few-tube (five tubes) FBC geometry time-averaged porosity and solids velocity distributions.
Fig. 21. Means and variances of fluctuations of pressure (kPa), porosity, axial gas and solids velocity (m/s) and axial solids momentum flux (J/g)
(1.12Umf).
solids momentum fluxes are found on the sides of the lower the trends continue, it appears that the middle tube may
tube. Since some experimental data show that high erosion overtake the lowest tube so that a monotonically increasing
rates occur near these locations, these large momentum erosion rate would result. This is the same trend which
values may be associated with the high erosion rates. occurs when the tube bank is moved closer to the distributor
These results provide much information towards devel- [98]. At higher distances from the distributor, the trend is
oping correlations between the hydrodynamics and erosion reversed. These trends are shown in Fig. 27 modified from
in a cold few-tube bubbling fluidized bed. Porosity means Fig. 3 of Parkinson et al. [98].
and variances appear to be useful in describing bubble The time-averaged transient erosion rates for all three
motions. Analysis of the results reveals major oscillation erosion models were arithmetically averaged and are plotted
modes of 1 Hz as the basic bed frequency and frequencies of in Figs. 28 and 29 as a function of fluidizing velocity. The
up to 4 Hz localized near the underside of the obstacles. average erosion rates for the three-tube geometry decrease
Typical vertical void fraction propagation velocities of from between 1.7 and 2.3 £ Umf and the erosion rates for the five-
0.5 to 1 m/s were predicted. The high cross-correlation tube geometry decrease between 1.12 and 1.7 £ Umf for the
between pressure and the lagging porosity fluctuations (over N– G and Finnie erosion models, respectively, as shown in
0.6) support the use of simple diagnostic measurements of Figs. 27 and 28. The N – G model agrees remarkably well
pressure to correlate bubble dynamics and erosion rates. with the MED erosion model and their trends are similar
except for the decreases already noted. The Finnie erosion
7.3. Erosion predictions model predicts erosion rates almost an order of magnitude
higher than those predicted by the MED and N– G erosion
Local computed time-averaged erosion rates of the models. If the erosion rates are normalized by the maximum
immersed tubes using the closed form MED, Neilson– erosion rates, all of the trends are similar, a basically
Gilchrist (N– G), and the Finnie erosion models given by monotonic increase in erosion rate with a tendency to
Eqs. (6.6), (6.7), and (6.9) are summarized in Figs. 22 – 24 plateau at the highest fluidizing velocities.
for the three- and five-tube models, respectively. The tubes Also shown in Figs. 28 and 29 is the prediction by the
are assumed to be made of aluminum having a hardness of MED erosion model of a monotonic increase in erosion rates
30 kgf/mm2 (294 MPa). As can be seen, the erosion rates are as a function of fluidizing velocity. At higher fluidizing
increasing more rapidly on the sides of the tubes facing each velocity, the erosion rates begin to level off. The shape of the
other as the fluidizing velocity increases. A large erosion curves are remarkably similar to that determined by
rate on the top of the center tube is predicted and may be too Parkinson et al. for 400 and 700 mm particles [96]. They
high partly due to the solids buildup because of the line of found that their experimentally determined erosion rates
symmetry. The erosion rate on the right sides of the tubes flattened out for U/Umf between 5 and 7 for those particle
closest to the walls is almost independent of fluidizing sizes, respectively. In this study, the prediction of this curve
velocity since it corresponds to the downcomer area in shape is considered to be a significant achievement.
which the solids descend at almost constant velocity and The wear patterns on the tubes at the lower fluidizing
porosity to the distributor. On average, the erosion rates velocities are quite uniform, which is the same trend as
predicted for the five-tube geometry are lower than that found in the CRE cold bed erosion experiment using PVC
predicted for the three-tube geometry. tubes at the lowest fluidizing velocities, approximately
As shown in Fig. 22 for the three-tube geometry, 2 £ Umf [96]. Over this range of fluidizing velocities, the
between 2.3 and 2.7 £ Umf, the MED erosion model predicts erosion rates for the MED and N– G erosion models are of
that the erosion rate on the left side of the lower tube is the same order as the experimental data, 0.1 – 0.4 mm/
slightly decreased while the right side of the upper tube is 1000 h as shown in Fig. 29 while those for the Finnie erosion
increasing. The erosion rate on the right side of the lower model are about an order of magnitude higher.
tube is almost independent of fluidizing velocity because For both beds, higher tubes experience larger erosion
this is where solids are decreasing and gas is flowing rates which may be explained by bubbles accelerating as
preferentially between the tubes. The erosion rate on top of they move upwards through the bed entraining with them
the lower tube decreases with fluidizing velocity while the more energetic solids. Table 5 summarizes the comparison
erosion rate on the bottom goes through a maximum. The of the erosion predictions for the three and five-tube CRE
net effect is that the average erosion rate of the lower tube generic FBC geometries. The results are quite close to each
goes through a maximum at 2.3 £ Umf, while the upper tube other which implies that perhaps only a small tube array
erosion rate continues to increase as shown in Fig. 25. may be sufficient to assess FBC erosion.
The average erosion rates for all three rows of tubes for In calculating the erosion rates for the MED erosion
the five-tube geometry are plotted in Fig. 26 as a function of model, we applied the empiricism, like Usimaru et al. [87]
fluidizing velocity. Comparison of these rates with those for that, when metal is removed, fewer than 10% of the grains
two rows of tubes shown in Fig. 25 reveals that they are in contact with the surface actually remove metal and
quite close and the trends are similar. The tube highest in thus the factor C ¼ 0:1: The remaining particles cause
the bed has the highest erosion rate at 2.3 £ Umf and, if only elastic deformation, not resulting in material wear.
Fig. 22. Computed time-averaged 3-tube transient erosion rates for aluminum, for the MED, Neilson–Gilchrist, and Finnie erosion models
(dp ¼ 500 mm), 1.7, 2.3, and 2.7 £ Umf.
Clearly, this factor is influenced by the hardness of the It is not the purpose of this review to investigate materials
erodent relative to the eroding surface and the erosion property models per se, but rather to develop a best-
mechanism, cutting tool (impaction) or sandpaper estimate mechanistic erosion model that incorporates the
(abrasion). This factor of 10% may be justified when state of present knowledge of material properties and the
the hardness of the particle is comparable with that of the results of hydrodynamic modeling.
wall over a fairly wide range, but it is not on a strong The experimental fluidized-bed tube wear data of Wood
theoretical foundation; its value, or a model for it, should and Woodford [65] and Parkinson et al. [98] indicate that
be investigated further. As discussed in Section 6, C is the erosion rate of aluminum is about 0.3 mm/1000 h for
related to the coefficient of restitution given by Eq. (6.4). bed material consisting of silica sand. The erosion rate of
(1.7 £ Umf, dp ¼ 500 mm).
0.3 mm/1000 h for the Wood and Woodford data was 7.4. Simplified closed form MED erosion model
deduced for 500-mm particles by linear interpolating the
data taken over a particle size range of 100– 1900 mm. The Bouillard and Lyczkowski [87] developed a simplified
fluidizing velocity varied from 1.5 to 3.6 m/s. mechanistic MED erosion model using a variational
The erosion rate of approximately 0.3 mm/1000 h for the principle which placed the closed form MED erosion
data of Parkinson et al. [98] was obtained for bare aluminum model derived by Bouillard et al. [37] on a firm foundation.
alloy (RTZ HE 30) tubes. This rate was the average for Run 1, A modification of that model is utilized herein to assist in
which had a bed material consisting of an equal mixture of 1–2 developing erosion design guidelines for the practicing
and 0.5-mm silica sand of mean surface diameter 0.86 mm engineer.
(860 mm) and a fluidizing velocity of 2 m/s. The maximum The erosion rate from the simplified quasi-one-dimen-
erosion rate on any tube was approximately 1 mm/1000 h. sional MED erosion model, E_ EDCF ; may be written in
(2.3 £ Umf, dp ¼ 500 mm).
the form, modified for Hydrodynamic Model B [31,87] as and
ð1 2 1Þð1 2 1gd Þ ð1 2 1gd Þ

E_ EDCF ¼ E_ 0 þK ðU 2 1Vs Þ ð7:1Þ 0:875gxd rg
12 12 K ¼ ð1 2 e2 Þ ð7:3Þ
Esp
where U is the superficial gas velocity, Vs is the solids-phase
velocity, and 1 is the void fraction. The erosion rate group, The units of E_ 0 are in terms of velocity, commonly
E_ 0 ; is given by expressed in mm/1000 h or mm/100 h. K is dimensionless.
e is the particle-surface restitution coefficient which is the
75mg gxd
E_ 0 ¼ ð1 2 e2 Þ ð7:2Þ ratio of particle rebound velocity, Vp2, to particle approach
ðfs dp Esp Þ velocity, Vp1, and xd is a characteristic acceleration distance
Fig. 25. Comparison of average time-averaged 3-tube erosion rates for aluminum tubes for upper and lower tubes for the MED erosion model.
Fig. 26. Comparison of average time-averaged 5-tube erosion rates for aluminum tubes for lowest, middle, and highest tubes for the MED
erosion model.
Fig. 27. Average weight loss of bare aluminum tubes after 48 h of operation (after Parkinson et al. [98]).
Fig. 28. Comparison of time-averaged transient erosion rates for aluminum (three tubes) for the MED, Neilson–Gilchrist, and Finnie erosion
models.
Fig. 29. Comparison of time-averaged transient erosion rates for aluminum (five tubes) for the MED, Neilson–Gilchrist, and Finnie erosion
models.
in the order of the heat exchanger tube spacing in FBCs. The the strikingly simple dimensionless form as
rest of the symbols are defined in the Nomenclature.
E_ EDCF
Experimental evidence [99] strongly supports the ¼ f ð1Þð1 þ 0:01167ReÞ ð7:4Þ
additional approximation that the solids velocity, Vs, in the E_ 0
vicinity of tubes is close to the superficial gas velocity, U. where the fluidization Reynolds number, Re ¼
With the approximation Vs ¼ U, Eq. (7.1) may be written in ðfs dp Þrg U=mg and
ð1 2 1Þð1 2 1gd Þ
Table 5 f ð1Þ ¼ ð7:5Þ
12
Effect of tube number on overall erosion rates
The SCFMED erosion model given above shows clearly the
Geometry Erosion model Erosion rate, mm/1000 h relationship between the operating and dependent variables
Fluidizing velocity, U/Umf and erosion, in particular, the porosity and the solids vel-
1.12 1.7 2.3 2.7 ocity. Note that the erosion rate group, E_ 0 ; has the dimension
of velocity and acts as a natural scaling parameter for metal
Three tubes Energy dissipation 0.18 0.35 0.44 0.47 wastage. The additional scaling parameters are the fluidiza-
Neilson–Gilchrist 0.10 0.3 0.23 0.36 tion Reynolds number, Re, and the porosity dependence
Finnie 1.20 3.17 2.93 3.56 through f ð1Þ given by Eq. (7.5). Eq. (7.4) clearly shows
that erosion measurements performed in laboratory exper-
Five tubes Energy dissipation 0.21 0.26 0.40
iments will scale to field units if the average bed porosity,
Neilson–Gilchrist 0.19 0.13 0.25
Finnie 1.04 1.71 3.01
fluidization Reynolds number, and erosion rate group, E_ 0 ;
given by Eq. (7.2) are the same for both units. These are
the first mechanistic erosion scaling equations known to data. The maximum erosion rate data of Parkinson et al. [96]
these authors. analyzed by Bouillard and Lyczkowski [87] displays a less
By analogy to the energy dissipation used in the prominent maximum around a porosity of 0.7. Parkinson
SCFMED erosion model, Yates [100] obtained a simplified et al. [96] used relationships developed by General Electric
erosion model that is proportional to the kinetic energy of which indicates a maximum normalized erosion rate at
the particles in the wake of the bubbles, Ewp. The final 1 ¼ 0.66, which corresponds to the transition from the
expression obtained depends upon the fourth power (Yates’s bubbling to turbulent fluidization regime. In the bubbling
emphasis) of the equivalent sphere bubble diameter given regime, erosion depends upon dp0:5 ; while in the turbulent
by regime it depends on (1 2 1 ).
For larger particles, both terms in Eq. (7.1) must be taken
Ewp ¼ 0:13ð1 2 1Þfw rs gde4 =ð1 2 fw Þ ð7:6Þ
into account and then the erosion rate is no longer explicitly
where fw is the fraction of bubbles occupied by wakes and de dependent on the void fraction but also upon the fluidizing
is the equivalent sphere diameter. velocity and the calculations become more complicated but
For 500 mm diameter particles, fw , 0.2 and 1 , 0.4, may still be solved on a desk calculator [87]. In that case, a
Eq. (7.6) becomes maximum may not occur with the SCFMED erosion model.
It is recommended that more experiments and calculations
Ewp ¼ 480de4 ð7:7Þ need to be performed. It should be further noted that in
The equivalent sphere bubble diameter, de, may be application of the SCFMED erosion model, the void fraction
estimated from Darton et al. [101] as: in the jet region, 1j, is replaced with the average fluidized
pffiffiffiffi bed void fraction, 1 [87]. Eqs. (7.4), (7.5), and (7.9) are the
0:54ðU 2 Umf Þ0:4 ðH þ 4 A0 Þ0:8 first, to the authors’ knowledge, which express the
de ¼ ð7:8Þ
g0:2 dimensionless erosion rate in terms of fundamental
dimensionless scaling parameters.
In order to link erosion to well-known dimensionless scaling
parameters, the SCFMED erosion model given by Eq. (7.4)
7.5. Erosion guidelines using semi-empirical erosion
may be rewritten in terms of the Archimedes, Ar, and
correlations
Froude, Fr, numbers as
E_ EDCF Zhu et al.’s [103] semi-empirical equation may be used
¼ f ð1Þ½1 þ 0:01167ðArFrÞ1=2 ð7:9Þ
E_ 0 as an erosion guideline. Like the simplified MED [87] and
Yates [100] erosion models, it relies upon use of the
since particle’s kinetic energy. Unlike others, and based upon
Re ¼ ðArFrÞ1=2 ¼ ðfs dp Þrg U=mg ð7:10Þ their own analyses, the work done on the target is taken to be
proportional to the Young’s modulus. The final form is
where implicit in the target material properties and resulted in the
Ar ¼ ðfs dp Þ3 rg ðrs 2 rg Þg=m2g ð7:11Þ following equation
E_ ¼ Cfv rp dpm unv ðb 2 fs Þ ð7:13Þ
and
where the erosion rate, E; _ is given in mm/100 h and the
Fr ¼ rg U 2 =½gðfs dp Þðrs 2 rg Þ ð7:12Þ
particle diameter, dp is given in mm. C is an empirical
Hence for proper dimensionless erosion scaling, the porosity constant depending upon material (erodent) properties and
and Archimedes and Froude numbers must be equal. particle hardness. fv is the void frequency, and uv is the void
The first term of Eq. (7.1), which comes from the Ergun velocity.
equation, predicts a maximum in the erosion rate at a void The values of n, m, and b are given in Zhu et al. [99,104]
fraction, 1 ¼ 0.57 with 1gd ¼ 1mf ¼ 0.4, the minimum (with m ¼ n and b ¼ C4 ), and were refit by Zhu et al. [103]
fluidizing void fraction, a phenomenon missed in almost using multiple linear regression to give b ¼ 1:1: The
all empirical correlations given by Parkinson et al. [96,102], exponents n and m are given in Table 6 and are quite
for example, including all those discussed in Section 7.5 close to those given in Zhu et al. [99,104]. The rec-
below. This maximum occurs at a void fraction of 0.62 for ommended final form of Eq. (7.13) is given by Zhu et al.
Hydrodynamic Model A with 1mf ¼ 0.38 as reported by [103] as:
Bouillard and Lyczkowski [87] and so the value of the
E_ ¼ Cfv rp dp1:2 u2:1
v ð1:04 2 fs Þ ð7:14Þ
maximum is not very sensitive to the hydrodynamic model used.
Analysis of Parkinson et al.’s [96] erosion data indicates It is important to note that the group (1.04 2 fs) plays a role
that a maximum occurs around a porosity of 0.63– 0.65 for similar to the coefficient of restitution, e, in the SCFMED
0.7 mm diameter particles. The erosion data for 1.0 mm erosion model given by Eqs. (7.1)– (7.3). As the sphericity,
diameter particles appears to plateau around the same values fs, decreases, the coefficient of restitution, e, decreases
of porosity. Unfortunately, there is significant scatter in the because more energy is transmitted to the target surface.
Table 6 Table 8
Fitted least-squares experiments for Eq. (7.13) [103] Comparison of Woodford and Wood [106] measured erosion rates
and predictions from Eq. (7.14) with values from Table 7 [103]
Material Experimental, m Experimental, n
Tube material Particle diameter (mm) Erosion rate
Brass 1.4 2.3 (mm/100 h)
Aluminum 2011 1.1 2.1
Copper 1.1 2.1 Measured Predicted
Stainless steel 304 1.2 1.8
Carbon steel 1050 1.1 2.0 Copper 1.9 17 19
Overall 1.2 2.1 SS 316 1.9 3.0 2.8
Copper 0.93 6.7 8.2
SS 316 0.93 1.8 1.2
In addition, as fs decreases, the void fraction, 1, increases

[105], also causing the erosion rate given by the SCFMED studies developed an empirical equation of the form
erosion model to increase up to the maximum value at [107]
around a porosity of 0.6 as discussed in Section 7.4. The cor-
E_ ¼ 1:873 £ 1029 U 2 D0:6 L1:33
s ð1 þ 0:1PÞ ð7:15Þ
relation given by Eq. (7.14), like most of the correlations of
Parkinson et al. [96,102], also fails to predict this maximum. where E_ is the erosion rate in mm/h, D is the tube diameter
The recommended values of C in Eq. (7.14) for silica sand in mm, Ls is the bed depth before fluidization in mm, and P
particles and their corresponding standard deviations are is the tube inclination in deg.
given in Table 7. The BCC concluded that comparison between the
C should be a function solely of the mechanical correlation given by Eq. (7.15) and observed metal loss
properties of both the particulate and tube materials. The for 11 FBC boilers, considering the variety of boiler designs
Young’s modulus of the tube material is implicit in C and from six manufactures, was most encouraging. Remarkably,
varies appreciably. The void frequency may be estimated as Eq. (7.14) predicts erosion rates on the same order as Eq.
discussed by Zhu et al. [103], and has been found to be in the (7.15), approximately 1 mm/100 h, over the same range of
range of 1 – 3 Hz [22]. parameters for steel tube materials [99,104]. The parameters
Zhu et al. [103] successfully analyzed the Woodford and typical for FBCs are dp ¼ 1 mm, U < 1 m/s (U/Umf < 2),
Wood [106] erosion data taken for 0.93 and 1.0 mm silica r < 2 £ 103 kg/m3, fs < 1, and Ls < 500 mm. It should be
sand particles and 10 tube materials by assuming a shape noted that unlike Eq. (7.14), Eq. (7.15) does not contain an
factor, fs, of 0.89, a void rise velocity, uv, of 2 m/s and a explicit particle size dependence. As shown by Zhu [99] and
void frequency, fv, of 2.2 Hz. The results of this analysis are Zhu et al. [104], the dependency of erosion increases
given in Table 8. significantly with particle diameter, with a nearly linear
The British Coal Corporation (BCC). shallow bed dependence for steels and an even stronger dependence for
softer materials in agreement with the data of Wood and
Woodford [65]. Eq. (7.15) does, however, predict erosion
Table 7 rates to increase with static bed height, Ls, and tube
Fitted least-squares values of C for Eq. (7.14) for silica sand [103]a inclination, P. With P ¼ 908 a 10-fold increase in erosion is
predicted for vertical tube banks over horizontal tube banks,
Material C Standard deviation in agreement with the BCC data.
Even more remarkable, given the same range of
Brass 0.115 0.0022 parameters, is that the simplified MED erosion model
Aluminum 201 0.119 0.0046
developed by Bouillard and Lyczkowski [87] also predicts
Copper 0.0683 0.0029
SS 316 0.0097 0.0016 erosion rates that are in the same order as given by Eqs.
CS 1050 0.0199 0.0010 (7.14) and (7.15). For U/Umf < 2, Fig. 3 of Bouillard and
SS 304 0.0101 0.0001 Lyczkowski [87] predicts an erosion rate for polyvinyl
CS 1020 0.0257 0.0009 chloride (PVC) tubes in the order of approximately 1 mm/
Iron 0.0239 0.0008 1000 h. Since the ratio of hardness of steels [2000 –
Pure aluminum 0.237 0.033 4000 MPa [91]] to PVC [40 MPa [87]] is about 100– 1,
Keewatin Steel 1 0.0146 0.0004 the predicted erosion rate for steels is 0.01 mm/1000 h
Keewatin Steel 2 0.0110 0.0005 (1 mm/100 h). As will be shown in Section 7.6, the
Keewatin Steel 3 0.0105 0.0002 SCFMED erosion model also predicts erosion rates on the
Keewatin Steel 4 0.0103 0.0001
same order as the simplified and full MED erosion models
a and brackets available experimental data for aluminum
The units of C and the standard deviation are
(mm s3.1)/(kg h m0.3). tubes.
7.6. Simplified mechanistic erosion guidelines procedure the data, such agreement with the simplified MED erosion
model is remarkable. The simplified MED erosion model,
The SCFMED erosion model given by Eq. (7.4) was Eqs. (7.1)– (7.5) takes a few minutes to compute on a pocket
used to develop a simplified mechanistic erosion guidelines calculator vs. a month on a mainframe computer using the
procedure. The basic concept is to use data as input to the FLUFIX/MOD2 [29] and EROSION/MOD1 [93] computer
SCFMED erosion model which is easily obtained exper- programs.
imentally, such as bed expansion. The time-averaged bed If the physical input parameters input to the SCFMED
expansion was obtained from the detailed hydrodynamic erosion model are missing, they must be estimated by (1)
computations performed by Bouillard and Lyczkowski [87] reviewing available databases for each parameter, (2)
for a few- (three-) tube approximation of the International estimating the range of uncertainties, e.g. tube hardness,
Energy Agency (IEA) Grimethorpe tube bank ‘C1’ and (3) identifying contradictions from different estimation
configuration [96], as shown in Fig. 18, over the range of procedures. Lyczkowski and Bouillard [109] used the
U/Umf from 1.0 to 2.7 for 500 mm diameter glass beads dimensionless groups E_ 0 ; Fr, and Ar in Eqs. (7.2) and
having a density rs ¼ 2.44 £ 103 kg/m3 as summarized (7.9) to develop a mechanistic erosion scaling procedure.
earlier in Section 7. The reader is referred to Bouillard and
Lyczkowski [87] for further details of the calculations.
The fluidized-bed void fraction is obtained as a function 8. Kinetic theory and discrete element method erosion
of the time-averaged fluidized-bed height, H, from the models
following solid overall mass conservation equation
The kinetic theory erosion model developed by Ding and
1 ¼ 1 2 ðHmf 1smf =HÞ ð7:16Þ Lyczkowski [110] is discussed in this section as well as a
brief summary of the discrete element method (DEM) is
where 1smf ¼ (1 2 1mf) is the solid volume fraction at
used to compute erosion rates using the Finnie erosion
minimum fluidization and Hmf ¼ 44.2 cm is the bed height
model.
at minimum fluidization. The rest of the parameters used to
evaluate Eq. (7.4) at 25 8C and 1.01 kPa are given by
8.1. Kinetic theory erosion model
mg ¼ 1.82 £ 1025 Pa s, rg ¼ 1.83 kg/m3, Esp ¼ 294 MPa,
dp ¼ 0.05 cm, xd ¼ 55 mm, e 2 ¼ 0.9, fs ¼ 1.0, and 1gd ¼ 0.4.
It was noted in Section 7 that the Finnie erosion model
Esp corresponds to the hardness of pure aluminum
produced erosion results approximately an order of
(30 kgf/mm2, 294 MPa) and xd is the minimum horizontal
magnitude higher than the N – G and MED erosion models.
spacing between tubes (89– 33.7 mm ¼ 55.3 mm) for Gri-
The kinetic theory of granular flow erosion model developed
methorpe tube bank ‘C1’ Parkinson et al. [96]. The results of
by Ding and Lyczkowski [110] offers an alternative
the calculations are listed in Table 9 and are plotted in
approach to extending single particle erosion models.
Fig. 30, where they are compared with the full MED erosion
A summary of the approach applied to the Finnie erosion
model results from Fig. 28. As can be seen, the agreement is
model is given in this section.
excellent. The results of the full MED and SCFMED erosion
The Finnie erosion model given by Eqs. (3.1a) and (3.1b)
calculations are compared with the data of Wood and
is written as a function of the particle’s instantaneous speed
Woodford [65] and Parkinson et al. [96,102], Zhu et al.
in the vicinity of an eroding surface, cw, angle of attack, a,
[104], and Foster Wheeler (Podolski et al. [108]) taken
and mass M ¼ pdp3 rp =3 as
under similar operating conditions in Fig. 30. As can be
seen, the agreement is excellent. Also shown is the W ¼ BF Mc2w f ðaÞ ð8:1Þ
sensitivity of the results to a three-fold increase in the
where
apparent hardness of aluminum due to surface oxidation
(90 kgf/mm2, 882 MPa). In spite of the diverse sources of 1
BF ¼ ð8:2Þ
8PH
Table 9
Metal wastage predictions from SCFMED erosion modela where PH is the Vickers hardness or flow stress, and f ðaÞ is
given by Eq. (3.3c).
E_ EDCF Finnie’s single particle erosion model is combined with
U/Umf U (cm/s) H (cm) 1 f ð1Þ Re
E_ 0 the kinetic theory of granular flow [111] to obtain an
expression for erosion due to repeated impacts on the eroding
1.0 0.209 44.2 0.40 0 10.5 0 surfaces. The erosion rate, E;_ of a solid surface caused by
1.12 0.234 46 0.42 0.066 11.7 0.075 repeated impacts can be obtained by calculating the number
1.7 0.355 50 0.47 0.168 17.8 0.203
of particles within a range c to c þ dc per unit volume and
2.3 0.481 55 0.52 0.213 24.2 0.273
unit time with the surface times the erosion caused by a single
2.7 0.564 59 0.55 0.223 28.4 0.296
particle impact. Thus, the erosion rate caused by repeated
a_
E0 ¼ 1:8 mm=1000 h: impacts given by Eq. (8.1) is integrated over all impact
Fig. 30. Comparison of closed form MED and SCFMED erosion models with available data for aluminum tubes. Data source key: Parkinson
et al. [96,102], Wood and Woodford [65], Zhu et al. [104], and Foster Wheeler (Podolski et al.) [108].
velocities in the range of (2 1, þ 1) to obtain Details of the integration process are given in Ref. [110].
ð This approach is similar to that used by Soo [64] to
E_ ¼ ðcw ·nÞBF Mc2w f ðaÞfw ðr; c; tÞdcw ð8:3Þ obtain his fluidized-bed erosion models described in
cw ·n.0 Section 3.2.1.
Rogers [112] developed a similar model using Finnie’s
where the single particle velocity distribution function in the
single particle erosion model [54] and the kinetic theory of
vicinity of the eroding surface, fw, is assumed to be
granular flow. That model is more involved and requires a
Maxwellian [34].
certain amount of iteration between the erosion calcu-
Carrying out the integration of Eq. (8.3), the result is
lations and the experimental data to determine adjustable
" rffiffiffiffiffi #
3=2
v2 2T parameters which minimize the difference. The kinetic
_E ¼ 21s rp BF ð2TÞ
pffiffi F1 ðuc Þþ w
3
F ðu Þþ v TF ðuÞ theory model given by Eq. (8.4a) only requires knowledge
p 2 p 1 c 2 w 2
of the granular temperature and some measure of target
ð8:4aÞ hardness.
where The input to the kinetic theory erosion model given by
Eq. (8.4a) may be the computed granular temperature as
p uc 1 1 3 described in Section 8.1.1 or experimental data such as
F1 ðuc Þ ¼ 2 þ sin4 ðuc Þþ sin4 ðuc Þ2 cos4 ðuc Þ
8 4 12 16 4 obtained by Cody et al. [113] using a non-intrusive vibration
¼ 0:1 ð8:4bÞ probe. They obtained the granular temperature data as a
function of fluidizing velocity for glass spheres ranging in
and diameter from 595 to 63 mm encompassing Geldart type B
(and nearly D) and the transition to Geldart type A particles.
2 1 2 Such data can be used as a check on the consistency
F2 ðuc Þ ¼ 2 þ sin5 ðuc Þ2 cosðuc Þsin4 ðuc Þ
5 15 5 between erosion measurements and granular temperature
2 2 measurements.
þ ðcosðuc Þsin ðuc Þþ2 cosðuc ÞÞ
15 The number of particles, Pw, colliding with a wall having
3 2 2 area A in a time interval dt can be estimated as
þ cos ðuc Þsin3 ðuc Þþ sin3 ðuc Þ
5 5
¼ 0:06 ð8:4cÞ Pw ¼ ðcw ·nÞfw ðrw ; cw ; tÞdcw dA dt ð8:5aÞ
Then the number of collisions per unit area per unit time is 187 cm/s. Further details of the computer model are given in
obtained from Ding and Lyczkowski [110].
ð Pw ð
Npw ¼ ¼ ðcw ·nÞfw ðr; c; tÞdcw 8.1.2. Computed erosion results and comparison with data
cw ·n.0 dAdt cw ·n.0
The kinetic theory erosion model was used to compute
rffiffiffiffiffi
61s T erosion rates and particle-wall collision frequencies at
¼ ð8:5bÞ stationary surfaces in the bed.
pdp3 2p
Fig. 32 shows the computed time- and surface-averaged
8.1.1. Three-dimensional computer model erosion rates as a function of tube material hardness
The two-dimensional IFAP (Isothermal Flow Analysis compared with the experimental data of Zhu et al. [104].
Program), which generalizes the FLUFIX/MOD2 code [29] Both two- and three-dimensional predicted erosion rates
used at ANL by adding a kinetic theory granular flow model match the experimental data reasonably well. However, the
and other features [114], was extended to a three-dimen- three-dimensional predictions generally agree better with
sional code for fluid – solids flow. The IFAP code has been the experiments. The clearest indication of the closeness of
demonstrated to be adaptable to a variety of problems the results can be seen in Fig. 32 for the lowest hardness
including a circulating fluidized bed [114,115] and gas – (brass). The differences at higher hardnesses are masked by
liquid slurry flow [116]. The basic numerical scheme for the linear scale since the erosion rates are inversely
solving the equations of gas – solids flow was described in proportional to the hardness which varies by a factor of 30.
detail by Ding [114]. The gas – solids bubbling fluidized bed The computed erosion rates at each brass tube surface are
erosion data of Zhu et al. [104] were simulated in two and shown in Fig. 33. Lower erosion rates were found near the
three dimensions. wall of the bed (y ¼ 8.88 cm) because of the lower solids
Fig. 31 shows a sketch of the IFAP fluidized-bed model motion and lower granular temperature there. The computed
grid used in the simulations. A rectangular tube results lie between the round- and square-tube erosion data
3.2 £ 3.2 £ 20.3 cm3 was used in the computations, instead of Zhu et al. [104]. Zhu et al. [104] and other investigators
of the round tube of diameter 3.2 cm used in the experiment, to [102] found that the highest erosion rates occurred near the
reduce the number of finite-difference cells. The bed material tube bottoms, while very low erosion rates occurred at the
was silica sand having a mean particle diameter of 1 mm; a tube tops. The erosion pattern of the three-dimensional
shape factor, fs, of 0.89; and a density of 2.58 g/cm3. The predictions qualitatively agrees with these experiments.
initial bed height was 32 cm. The minimum fluidization The two-dimensional results, however, do not give
velocity, Umf, was 56 cm/s. The fluidizing velocity was this erosion pattern. Therefore, it is concluded that
Fig. 31. Three-dimensional non-uniform-finite difference grid.

Fig. 32. Comparison of computed time-averaged erosion rates as a function of tube hardness with experimental data of Zhu et al. [104,110].
Fig. 33. Computed time-averaged erosion rates (mm/100 h) at each brass tube surface in the three-dimensional bed [110].
the two-dimensional model is not as good as the three- [Vickers hardness, 98 kgf/mm2 (960 MPa)] tubes as a
dimensional model for computing local erosion rates. function of fluidizing velocity and particle size for silica
The collision frequencies (counts/s) computed from the sand were performed and are shown in Figs. 34 and 35,
kinetic theory model were found by Ding and Lyczkowski respectively. The particle properties used in the simulation
[110] to lie in the range determined by Drennen et al. [117] are listed in Ref. [104, Table 3]. These experiments show
who used a 4-mm diameter impact probe mounted in a 5.72- that either increasing the fluidizing velocity for a given
cm-diameter tube. particle size, or increasing the particle size at a fixed
In order to reduce the computer time needed, the two- fluidizing velocity increases tube wear. The computations
dimensional kinetic theory hydrodynamic and erosion generally display the same trends and agree reasonably well
models were used to assess the capability of the kinetic with the data. Three-dimensional computations may pro-
theory erosion model to predict erosion trends as a function duce better agreement.
of fluidizing velocity and particle size. Comparisons of the The observed phenomena, which were also found in
time- and surface-averaged computed erosion of the brass experiments by Wood and Woodford [65], Parkinson et al.
[Vickers hardness, 155 kgf/mm2 (1519 MPa)], and copper [102], and Zhu et al. [104], for example, can be explained as
Fig. 34. Comparison of computed time- and surface-averaged computed erosion rates erosion rates as a function of fluidizing velocity with
experimental data of Zhu et al. [104,110].
follows. According to the kinetic theory of granular flow concentration. The increasing granular temperature
used in this paper, small particles produce less energy and increases the tube erosion. These trends compete with
dissipate more energy through the inelastic particle – particle each other. When the fluidizing velocity becomes high
collisions and the two-phase drag than do larger particles. enough, the decrease in solids concentration becomes domi-
Hence smaller particles produce a lower granular tempera- nant over the increase in solids velocity, granular tempera-
ture. According to the Finnie kinetic theory erosion model ture, and mean velocity, and erosion tends to decrease.
given by Eq. (8.4a), erosion decreases with decreasing This explains the experimental findings of Parkinson et al.
granular temperature. A higher fluidizing velocity for a [102] who showed that tube wastage increased with fluidiz-
given particle size usually increases the fluctuating intensity ing velocity up to a value of 0.55 times the particle terminal
and mean velocity of solid particles and decreases the solids velocity, and after that point there was a decrease in tube
Fig. 35. Comparison of computed time- and surface-averaged computed erosion rates erosion rates as a function of particle diameter with
experimental data of Zhu et al. [104,110].
wear. This maximum was predicted by Bouillard et al. [87] velocity bracketed data taken at atmospheric conditions
using the original simplified MED erosion model and by pressure. The effect of moderate pressure (1.2 MPa)
Lyczkowski and Bouillard [101] using the SCFMED increased the computed bubble frequency to 4 – 8 Hz as
erosion model summarized in Section 7.4. well as the particle impact velocity. No comparisons were
On the basis of generally good agreement between made with data.
computed erosion and experimental data, it is concluded that The energy calculated by the DEM was used with the
the Finnie kinetic theory erosion model correctly describes Finnie erosion model, however no comparisons were made
significant trends and mechanisms for ductile erosion when with data. Using a hardness of 6 £ 108 kgf/m2 (600 kgf/mm2
used in conjunction with the kinetic theory of granular flow. or approximately 6000 MPa) which corresponds roughly to
Since the erosion predictions are closely related to bed hardened steel (Tables 3 and 4), Rong and Horio computed
dynamics, such good agreement between predicted and erosion rates in the order of 10 mm/1000 h (0.01 mm/100 h),
measured erosion rates gives indirect validation of the which are in rough agreement with rates predicted by
kinetic theory hydrodynamic model. The Finnie kinetic Lyczkowski and Bouillard [109] as discussed in Sections 7.4
theory erosion model explains, from the viewpoint of kinetic and 7.5. Stringer’s [76] caveat applied to Tabakoff’s
theory, the experimental finding that the erosion rates
proposal to use his single-particle erosion methodology for
increase with increasing fluidizing velocity and particle
fluidized beds mentioned in Section 4.1 is applicable to the
diameter. Comparison of predicted erosion patterns of two-
DEM, i.e. it may not be possible to track sufficient particles
and three-dimensional models with experimental results
for a fluidized bed simulation to be meaningful.
again shows the importance of three-dimensional models.
The shortcoming of the DEM is that it uses a system of
Other single-particle-impact models such as the N– G
dashpots and springs which may render it unsuitable for
model can be extended using the kinetic theory of granular
flow to account for additional material properties of eroding truly predictive calculations. An alternative is the direct
targets, such as the Young’s modulus, which has been numerical simulation (DNS) of fluidized particles pioneered
suggested to be a better parameter to correlate erosion [103]. by Joseph and his colleagues [123]. This method requires no
Particle properties such as hardness, shape factor, and adjustable parameters and appears to be truly predictive if
coefficient of restitution may also need to be considered in the number of particles can be made sufficiently large.
further extension of the kinetic theory erosion model. Videos of their simulations may be found at http://www.
aem.umn.edu/Solid-Liquid_Flows/
8.2. Discrete element method erosion model
An alternative to extending single particle erosion

models using the hydrodynamic models contained in the 9. Summary of the Chalmers University of Technology
FLUFIX/MOD2 computer code [29] and kinetic theory of research program
granular flow models contained in the IFAP computer code
[110,111,114] is to use the instantaneous time dependent During the time this review was being written, the
particle velocities computed by the DEM [118,119]. Tsuji
primary author was partially aware of the developments
et al. [120] refer to this method as the discrete particle
ongoing at Chalmers University of Technology in Sweden
simulation. Tsuji [121] summarized Japanese activities on
[35]. At the suggestion of Dr John Stringer of EPRI, one of
discrete particle simulation in 2000. Rong et al. [118]
the reviewers, a synopsis and critique of this important body
reviewed the DEM literature in 1999 and applied the
of work on fluidized-bed hydrodynamics and erosion
technique to describe particle motion and erosion in
experiments integrated with numerical modeling is included
fluidized beds containing an immersed single tube as well
in this section.
as tube arrays. Rong et al. failed to refer to the work of
Walton and collaborators (see Walton and Braun [122] for Experimental studies of fluid dynamics, heat transfer,
example) who use basically the same approach as DEM and and erosion in pressurized bubbling fluidized beds were
refer to the technique as particle dynamics. The DEM initiated around 1990 [124]. More recently, in-bed dynamics
calculations were performed in two dimensions using a stair have been modeled numerically. Some of the group’s earlier
case approximation to the round tubes and treated the modeling work was included in the review by Enwald et al.
particles as circles. Walton and co-workers generally use [35]. Of direct relevance to this review is the program of
spherical particles in three dimensions. Their application is fluid dynamics and erosion experiments [125 – 127] closely
not fluidized beds, but to perform simulations to test the coupled with fluidized-bed CFD and erosion modeling
assumptions made in the kinetic theory of granular flow. [128– 130]. This work builds upon previous numerical CFD
The numerical results from the DEM for particle and development and validation activities [131,132] and is
bubble patterns agreed qualitatively and more quantitative continuing [133]. This program appears to have been
computations such as bubble frequency (2 – 3 Hz), bed heavily influenced by the hydrodynamic and erosion
expansion, bubble velocity (, 0.7 m/s), and particle impact research at ANL and IIT (Section 2).
9.1. Fluidized-bed hydrodynamics experiments atmospheric fluidized beds. Cross-correlations were com-
puted between the erosion rates and various hydrodynamic
Olsson et al. [124] measured bed expansion, mean parameters measured previously by Olsson et al. [124]. This
bubble properties such as pierced length, rise velocity, and concept was put forward by us in 1990 [135]. The
frequency, and power spectra of overall bed pressure drop in correlations showing the strongest influence on tube erosion
a plexiglass rectangular fluidized bed having cross-sectional were found to be the mean bubble velocity, flow rate, and
dimensions of 0.2 m deep by 0.3 m wide. This fluidized bed bubble volume fraction, in that order. Erosion rates were
was placed in an outer pressurized vessel fitted with found to be highest in the center of the tube bank and lower
windows in order to photograph the bed in operation. The near the bed walls, consistent with the findings for the cold
bed material was sand having a wide size range with a mean few-tube erosion predictions summarized in Section 7.3 and
particle diameter of 0.7 mm (700 mm), a density of 2600 kg/ the single square tube modeled by Ding and Lyczkowski
m3 corresponding to group D particles using Geldart’s [110] summarized in Section 8.1.2. The highest erosion
classification (which is only applicable at atmospheric rates were on the bottoms of the tubes with very little wear at
pressure), and a shape factor of 0.8. The bed height at their tops. This finding is at odds with the results for the two
minimum fluidization was 0.86 m at atmospheric pressure off-center tubes for the five-tube erosion computations
and room temperature. Operating pressure ranged from 0.1 shown in Figs. 23 and 24. However, as shown in Ding
to 1.6 MPa at two velocities, U 2 Umf ¼ 0.2 and 0.6 m/s and Lyczkowski [110], more correct trends are found if
where Umf is the minimum fluidization velocity. With three-dimensional computations are performed.
Umf ¼ 0.4 m/s at 0.1 MPa, U/Umf ¼ 1.5, and 2.5. Three The maximum erosion rate for the target tube was found
different tube-bank configurations were used in addition to a to increase with pressure with some of the rates going
bed having no tubes. The tubes were constructed of through maxima between 0.4 and 1.0 MPa. Increasing
aluminum and had a diameter of 20 mm. All the mean pressure while maintaining constant excess fluidizing
bubble parameters except the pierced length were found to
velocity, U 2 Umf, increases U/Umf. For example, at
increase with fluidizing velocity. The bubble frequency and
U 2 Umf ¼ 0.6, pressures of 0.1, 0.8, and 1.6 MPa result
expanded bed height increased with pressure. Interestingly,
in U/Umf ¼ 1.43, 2.5, and 3.3, respectively. In addition, the
the major frequency of the overall bed pressure drop was
fluidizing gas density increases over an order of magnitude,
found to be about 1 Hz at 0.1 MPa, in agreement with the
increasing the mass flow and momentum directly. The
computations for the tube simulations described in Section
tendency for the erosion rate to go through a maximum was
7.2 for the IEA Grimethorpe tube bank ‘C1’ configuration
discussed in Section 7.4. Olsson et al. [124] did not find a
[134]. This low frequency is interpreted by us to mean that
maximum in the bed expansion which would imply a
the fluidized bed is in the slugging regime. As the pressure
maximum in the erosion rate according to the SCFMED
was increased, the power spectra became wider until no
major frequency was predominant, which implies that the erosion model, Eq. (7.4), with pressure but did report earlier
fluidized bed made a transition to the so-called turbulent work that found decreasing bed expansion with increased
regime. pressure for some combinations of high pressure and excess
gas velocity, U 2 Umf. The technique used by Olsson et al.
9.2. Erosion experiments [124] for measuring the bed expansion may account for
these discrepancies and, unfortunately, there is not enough
Wiman et al. [125] measured local tube erosion in the information to convert bed expansion data to average bed
same fluidized-bed facility as the in-bed hydrodynamic porosity using Eq. (7.16).
measurements just described. Several of the aluminum tubes Wiman and Almstedt [127] went on to perform erosion
were replaced with stearin-coated brass-core target tubes to and heat transfer measurements in yet another tube bank
measure erosion. These tubes were placed at strategic having an even denser tube-bank array packing and an
locations slightly over one-half of the tube-bank height. It additional bed material consisting of sand having an average
was stated that the bed height at minimum fluidization was particle diameter of 0.45 mm (450 mm). Erosion rates in
0.86 m but it was not stated for which tube bank the less tightly packed tube bank were found to be much
configuration this was for. The tube-bank height was higher than in the more densely packed tube bank. The
0.75 m, about the same height as the fluidized bed at maximum erosion rate for the larger particle diameter sand
minimum fluidization. Local erosion measurements at 308 was found to be higher by about 25%, in rough agreement
increments around the target tubes were obtained with a with the data of Zhu et al. [103,104] and as predicted by the
micrometer. The use of stearin-coated tubes allowed erosion kinetic theory erosion model in Fig. 34. In none of these
measurements to be made after just 1 h of bed operation, erosion studies was an effective hardness measured for the
orders of magnitude higher than using aluminum target stearin coating on the target tubes. However, if a predictive
tubes. Nonetheless, erosion patterns were found that were in erosion model were to analyze these experiments, the
good agreement with results reported from atmospheric and effective hardness could be backed out. All in all, the
pressurized combustors as well as results obtained in cold erosion measurements serve to confirm the erosion rate
trend predictions from the MED, SCFMED, and kinetic by Enwald et al. [131] who give more details of the CFD
theory erosion models, lending further confidence in them. code, a description of their two-fluid model, and the
Gustavsson and Almstedt [128] conducted erosion numerical solution method.
experiments for the same two-tube T2 configuration used As mentioned above, the fluidized-bed experiments
by Olowson [136] to study fluidized bed hydrodynamics. performed by Olowson [136] contained two tubes, denoted
That experiment and simulations was summarized in T2. The fluidized bed is rectangular in cross section having a
Section 9.3. Gustavsson and Almstedt [128] summarized width of 0.3 m and a depth of 0.2 m. Experiments having no
Wiman’s erosion experiments [126] which were published tubes were also performed. The bed is the same as used in
earlier by Wiman et al. [125] for an excess fluidizing the experiments by Wiman et al. [125], Wiman [126], and
velocity, U 2 Umf ¼ 0.6 ðU=Umf ¼ 2:5Þ to put their erosion Wiman and Almstedt [127]. The 20 mm diameter tubes
measurements in context. They found that their erosion were placed 0.55 m above the distributor plate. The bed
measurements for the two tubes T2 configuration were material was the same silica sand as in the experiments by
higher than for the moderately dense, in-line I4 and dense, Wiman et al., with a mean particle diameter of 0.7 mm
staggered, S4D configurations. They concluded from power (700 m) and a density of 2600 kg/m3. These particle
spectra analysis of the overall pressure drop fluctuations that properties were also used as inputs to the simulations
the tube banks promote the transition from slugging to what described below, together with a sphericity of 0.8. The
they classify as a turbulent behavior, resulting in a lower pressure was varied from 0.1 to 1.6 Pa. The total bed height
erosion rate. This implies that erosion simulations for was 2.16 m including the freeboard. The bed height at
fluidized beds having a few tubes, such as performed for the minimum fluidization was 0.87 m. The excess gas fluidizing
IEA Grimethorpe tube bank ‘C1’ in Sections 7.4 and 7.5, velocity was 0.6 m/s and so U/Umf ¼ 2.5. The five mean
would produce upper bounds for the erosion rates. bubble parameters frequency, pierced length, rise velocity,
Increasing pressure was found to increase the erosion volume fraction and flow rate were obtained this time using
rate as found previously by Wiman et al. [125] but no a capacitance probe mounted between the two tubes.
maximum was claimed to have been found up to 1.6 MPa The experiment was modeled in two dimensions without
for the case of U 2 Umf ¼ 0.6 m/s by Gustavsson and the assumption of symmetry using 7950 computational
Almstedt [129]. However, there is very little increase in the cells. A solids viscosity estimated to be 1.0 Pa s (10 Poise)
maximum erosion between 0.8 and 1.6 MPa, inferring that a was used. Simulation runs were made at 0.1, 0.4, 0.8 and
maximum had been achieved. The bed expansion data of 1.6 MPa. Asymmetries were introduced into the inlet
Olowson [136] for the case of two and four tubes were boundary by using a few jets placed in the distributor
independent of tube configuration. The bed expansion for plate and subsequently turned off after 1.0 s of transient
excess fluiding velocity, U 2 Umf ¼ 0.2 m/s goes through a time. The partial slip boundary condition derived by Ding
broad maximum between roughly 0.5 and 1.0 MPa. The bed and Lyczkowski [110] was used for the solids phase. The
expansion for U 2 Umf ¼ 0.6 m/s increases up to 0.5 MPa, computed volume fractions were processed using the same
and is basically flat up to 1.6 MPa, consistent with very little algorithm as used to process the data from the capacitance
increase in maximum erosion found by Gustavsson and probe for consistency. The capacitance probe data was
Almstedt [129] between 0.8 and 1.6 MPa. No erosion data obtained over a 10 min period, while the computations were
were reported for 0.4 MPa. run for 20 s.
Numerical and experimental results were also presented
9.3. Fluidized-bed hydrodynamics CFD simulations and for the case of a bed having no tubes, referred to as a freely-
validation bubbling bed. The experimental results were obtained by
either Olowson [137] or Olowson and Almstedt [138] (there
Gustavsson and Almstedt [128] used their two-fluid are contradictory references to the data in the text and
fluidized-bed CFD hydrodynamics code GEMINI devel- figures given in the paper [126] and an erroneous reference
oped and validated by Enwald et al. [131] and Enwald to Olsson and Almstedt). The numerical results were
and Almstedt [132] to simulate a cold pressurized obtained by Enwald [139].
fluidized-bed experiment containing only two tubes Agreement with the data for the five bubble parameters
[136]. In 1996 Enwald et al. [35] presented a review of was better at higher pressures. However, trends predicted
worldwide modeling progress, including that at Chalmers from the calculations are in the opposite direction for the
University of Technology. Their CFD code exists in bubble frequency and flow rate. Agreement at low pressures
several versions. The basic version uses a multiblock is poor with severe overprediction of bubble frequency, flow
solver using a single processor computer. A non- rate, pierced bubble length and rise velocity. The authors
orthogonal curvilinear coordinate system is used to more argue that the reasons for disagreement may be due to non-
accurately model round in-bed tubes. The basic solution grid independency and problems modeling the distributor
method is the same as used in the FLUFIX code [29,92], plate. The disagreement might also be due to the fact that
the Implicit Multifield (IMF) numerical technique [38]. the computations were done in two dimensions. Ding and
There also exists a parallel processor version developed Lyczkowski [110] showed that computed erosion results
were better in three dimensions. Nonetheless, the integrated discussed in Section 7, and those of Gustavsson and
computer modeling work and validation with experiments is Almstedt [142], the trends are generally consistent. Con-
quite impressive and promising. sidering the differences in implementation and the use of
non-orthogonal curvilinear coordinates, this is encouraging.
9.4. Erosion simulations Obviously the next step is for them to make predictions of
erosion rates with the MED erosion model (and others) and
Gustavsson and Almstedt [129] adopted the closed form compare them with absolute erosion rate data such as done
MED energy dissipation model to analyze their erosion in Sections 7.4 and 7.5. Gustavssen has initiated such an
experiments. Their reference for this MED energy dissipa- attempt [142] where he has divided the MED energy
tion model was given as Lyczkowski et al. [140] which, dissipation model into deformation and cutting wear and
unfortunately, had a serious typographical error in their Eq. related this model to the Neilson– Gilchrist [62], Bitter [60,
(20). The factor of 1/2 was outside the brackets. The correct 61], and Engel [50] erosion models. A similar approach
form is given by the term in brackets in Eq. (5.32c) above. using the power dissipation model was discussed in Section
Fortunately, we found that the second term representing the 5.1.4 originally published in 1990 by us [135].
energy dissipation due to drag is an order of magnitude
larger than the energy dissipation caused by viscous shear,
and so the error is of little consequence. In fact it was
10. Conclusion
dropped in developing the SCFMED erosion model in
Section 7.6. Their energy dissipation model was expressed
in non-orthogonal curvilinear coordinates and computed A brief review of the status of fluidized bed combustion
using the hydrodynamic output from the fluidized-bed and the erosion issue has been performed. A summary of
hydrodynamic code GEMINI [128]. representative hydrodynamic calculations for a rectangular
Erosion rates were not computed, nor compared with the two-dimensional fluidized bed containing an obstacle
absolute data. Instead the energy dissipation rates are used showed good agreement with the experimental data. The
to interpret trends and to compare relative erosion rates. The bubble frequencies of approximately 4 Hz around and 9–
maximum energy dissipation rates were found to be on the 10 Hz under the obstacle and the size of the first bubble
lower portions of the tubes facing the center of the bed and agree reasonably well with the results of a high-speed
almost none on their tops. The computed angles of motion picture study. The time-averaged porosity distri-
maximum energy dissipation are roughly 110 and 2708 bution also agrees reasonably well with data taken with a
measured counterclockwise from the top of the tube, while gamma-ray densitometer. The computed results are rela-
the erosion data indicates maxima at roughly 170 and 2208. tively insensitive to the hydrodynamic model and solids
The computed energy dissipation rate at 0.8 MPa is below elastic modulus parameter. Other validation experiments
0.1 MPa. The reason given for this is that not a long enough are reviewed, including the integrated bubbling fluidized-
transient time was used for the simulation. The maximum bed experimental and hydrodynamic CFD and erosion
energy dissipation rate was computed to occur at 1.6 MPa, modeling research ongoing at the Chalmers University of
consistent with the erosion data. The shape of the computed Technology.
energy dissipation rate profiles at low pressure are in A critique of six erosion models from the literature was
qualitative agreement with the erosion data profiles, but performed. They were shown to have various flaws. For
their shapes at 0.8 and 1.6 MPa disagree significantly. example, Finnie’s model predicts no erosion at 0 and 908.
By plotting the time dependent energy dissipation rates None of the erosion models include hydrodynamic par-
and particle volume fractions together on the same time ameters. A methodology was developed to use these models
scale, it was observed that the maximum energy dissipation with the hydrodynamic outputs of the FLUFIX computer
rates occurred when the particle volume fractions changed programs [29,92].
from their lowest values rapidly to their highest values, The energy dissipation model was developed, and was
corresponding to the passage of bubbles striking the tube. shown to generalize the so-called power dissipation model
This same phenomenon was reported by Lyczkowski et al. developed for low-angle erosion caused by slurries. The
[141] for the case of erosion calculations for the rectangular basis for the model was shown to be both ball-mill grinding
obstacle discussed in Section 2.2 using the prototype energy and abrasive-wear laws. By generalizing these laws to
dissipation model given by the first six terms in Eq. (5.25). multiple dimensions, it was possible to include impaction
The maximum energy dissipation rate was deduced by effects as well. Thus, it was possible to show for the first
Gustavsson and Almstedt [129] to be caused by the bubble time, to the authors’ knowledge, that impaction and abrasion
wakes. It is gratifying that the Chalmers University of erosion are basically governed by the same mechanism: the
Technology group has adopted the MED erosion model and imparting of the force of the particle stream to the eroding
are improving the resolution of the in-bed tube geometry. In material surface or, alternatively, the transfer of a portion of
spite of detailed differences between our computations the particle energy to the same surface. Lyczkowski et al.
for the cold few-tube IEA Grimethorpe tube bank ‘C1’ [143] used the former concept to extend the Finnie erosion
model to include a constant applied force. The MED erosion Acknowledgements

model was developed and refined.
This review has been able to demonstrate a unification of This work was originally supported by the US
the entire erosion literature developed for over a century Department of Energy, Assistant Secretary for Fossil
beginning with Kick’s law in 1885 and Rittinger’s 1867 ball Energy, Morgantown Energy Technology Center (now
mill model. It is felt that the MED erosion model (and the National Energy Technology Laboratory (NETL)), under
SCFMED erosion models) are capable of quantitatively contract No. W-31-109-ENG-38, and the Cooperative
predicting erosion rates given reasonable estimates of Research and Development Venture ‘Erosion of FBC
material properties because they are mechanistic rather Heat Transfer Tubes’ and the National Science Foundation.
than correlative. Members of the venture were the US Department of
Improvements in the hydrodynamic models contained in Energy, National Energy Technology Laboratory, Electric
the FLUFIX computer program, together with improve- Power Research Institute, State of Illinois Center for
ments in the MED erosion model have enhanced ANLs Research on Sulfur in Coal (now the Illinois Clean Coal
capabilities of analyzing metal wastage in FBCs. Two other Institute), Foster Wheeler Development Corp., ASEA
erosion models, Finnie’s and Neilson– Gilchrist’s (N –G) Babcock PFBC, ABB Combustion Engineering, Inc.,
offer a set of alternative erosion models in the EROSION Tennessee Valley Authority, British Coal Corporation,
[93] computer program. The MED and N– G erosion models CISE, and ANL. Additional support was also by Dr Brian
produce very similar erosion rates and patterns. The Finnie G. Valentine of the US Department of Energy, Office of
erosion model yields erosion rates about an order of Industrial Technologies, Industries of the Future-Chemi-
magnitude higher but all are in reasonable agreement with cals. We thank Thomas J. O’Brien of NETL for his many
available erosion data. suggestions and helpful comments, and, particularly, for his
Parameter sensitivity studies of a generic few-tube moral and financial support as well as for later careful
approximation of the Coal Research Establishment’s cold review of an early draft manuscript. We also thank the late
model erosion experiment have increased the understanding Dr Holmes A. Webb of NETL, for suggesting the MED
of the factors contributing to metal wastage. The erosion erosion model. Our thanks, also, to the late Prof S.L. Soo of
models produce very similar erosion patterns on a normal- the University of Illinois Urbana-Champaign for confirm-
ized basis implying that (1) a few-tube approximation to a ing the calculations in Section 3.2.1. Dr Stephen M. Folga
large-scale FBC may sufficient for understanding the wear is to be thanked for assistance with the computations
mechanisms and (2) the bed dynamics as it affects the performed in Section 7.
energy dissipation appears to control the wear patterns. The authors would like to acknowledge Dr Stringer of
A simplified mechanistic erosion guideline procedure EPRI for his many suggestions and stimulating insights
has been related to semi-empirical correlations and made during his frequent visits to ANL and to the Steering
laboratory and field data in a meaningful manner using the Committee of the Cooperative R&D Venture ‘Erosion of
macroscopic mechanistic SCFMED erosion model. Several FBC Heat Transfer Tubes’. Their many concerns on
examples demonstrate the predictive capability of this understanding small-scale cold model fluidized-bed hydro-
model. The semi-empirical correlations of laboratory and dynamics and their applicability to large-scale reactive
field data and use of the microscopic hydrodynamic and FBCs helped to shape the nature of the simulations, to force
MED erosion models are the alternatives. These semi- us to interpret the significance and implications of the
empirical correlations point the way for extension of the full findings, and to keep the project focused.
closed form MED and macroscopic SCFMED erosion Finally, we thank Professor Almstedt of Chalmers
models to account for effects such as tube inclination. University of Technology for reviewing Section 9.
Use of the kinetic theory erosion model using the Finnie
erosion model produces erosion rates in agreement with
experimental data since it uses the fluctuating rather than References
mean solids velocity. The DEM erosion model hold promise,
but many more particles are required to show convergence. [1] Kunii D, Levenspiel O. Fluidization engineering, 2nd ed.
The integrated bubbling fluidized-bed experimental and Stoneham: Butterworth/Heinemann; 1991.
hydrodynamic CFD and erosion modeling research ongoing [2] Beer JM. Combustion technology developments in power
at the Chalmers University of Technology has been shown generation in response to environmental challenges. Prog
to confirm the trends predicted from the MED and kinetic Energy Combust Sci 2000;301–27.
[3] Erlich S. Fluidized combustion—is it achieving its promise?
theory erosion models. These include the spatial variation of
Conference Preprints, Third International Fluidized Combus-
erosion in a tube bank and around the tubes as well as tion Conference, vol. 2. London: The Institute of Energy;
fluidizing velocity and particle diameter dependency. Thus October 1984. KA/1/1-KA/1/29.
group is also examining refinements to the MED erosion [4] Beer JM. The fluidized combustion of coal. Sixteenth
model using their own two-fluid hydrodynamic CFD Symposium (International) on Combustion, Pittsburgh: The
computer code, GEMINI. Combustion Institute; 1976. p. 439–60.
[5] Fennelly PF. Fluidized bed combustion. Am Scientist 1984; [22] Lyczkowski RW, Gamwo IK, Dobran F, Ai YH, Chao BT,
72:254–61. Chen MM, Gidaspow D. Validation of computed solids
[6] Schwieger R. Fluidized-bed boilers achieve commercial hydrodynamics and pressure oscillations in a model bubbling
status worldwide. Power 1985;129(2):S1 –S16. atmospheric fluidized bed combustor. Powder Technol 1993;
[7] Krishnan RP, Daw CS, Jones Jr.JE. A review of fluidized bed 76(1):65–77.
combustion technology in the United States. In: van Swaaij [23] Rowe PN, Masson H. Intersection of bubbles with probes in
WPM, Afgan NH, editors. Heat and mass transfer in fixed and gas fluidized beds. Trans Inst Chem Engr 1981;59:176–85.
fluidizedbeds.Washington,DC:Hemisphere;1985.p.433– 55. [24] Rowe PN, Everett DJ. Fluidized bed bubble viewed by X-rays.
[8] Stringer J, Wright I. Overview of in-bed erosion in fluidized- Part I. Experimental details and the interaction of bubbles with
bed combustors. Materials and Components in Fossil Energy solid surfaces. Trans Inst Chem Engr 1972;50:42–8.
Applications, US Department of Energy Report DOE/FE- [25] Fakhimi S, Harrison D. The void fraction near a horizontal
0054/54; February 1, 1985. tube immersed in a fluidized bed. Trans Inst Chem Engr
[9] NCB (IEA Grimethorpe) Ltd. Incident Report Tube Bank 1980;58:125–31.
‘C’—Metal Wastage Status Report, November 1982. US [26] Davidson JF. Symposium on fluidization—discussion. Trans
Department of Energy Report DOE/METC-83-54 Inst Chem Engr 1961;39:230–2.
(DE83008239); November 1982. [27] Smoot L. Modeling of coal-combustion processes. Prog
[10] Kobro H. Discussion of the operation and performance of a Energy Combust Sci 1984;10(2):229–72.
2.5 MW fast fluidized bed combustor and a 16 MW bubbling [28] Gidaspow D. Hydrodynamics of fluidization and heat
bed combustor. Conference Preprints, Third International transfer: supercomputer modeling. Twenty-Third Nat Heat
Fluidized Combustion Conference, vol. 1. London: The Transfer Conf, Appl Mech Rev 1986;39(1):1 –23.
Institute of Energy; 1984. DISC/14/110-DISC/14/120. [29] Lyczkowski RW, Bouillard JX, Folga SM. Users manual for
[11] Xu-Yi Z. The progress of fluidized-bed boilers in People’s FLUFIX/MOD2: a computer program for fluid – solids
Republic of China. Proceedings of the Sixth International hydrodynamics. Argonne National Laboratory Sponsor
Conference on Fluidized Bed Combustion. US Department Report, Argonne, IL (April 1992). Reprinted by USDOE
of Energy Report CONF-800428, vol. 1; 1980. p. 36–40.
METC as DOE/MC/24193-3491/NTIS No. DE94000033),
[12] Stringer J, Wright I. Materials issues in fluidized bed
available from NTIS, Springfield, VA; 1994.
combustion. Presented at ASM Conference on Materials for
[30] Gidaspow D, Ettehadieh B. Fluidization in two-dimensional
Future Energy Systems, Washington, DC; May 1984. [Ref.
beds with a jet. 2. Hydrodynamic modeling. I&EC Fundam
[8] is an abstract of this paper.].
1983;22:187–93.
[13] Grace JR. Agricola aground: characterization and interpret-
[31] Bouillard JS, Lyczkowski RW, Gidaspow D. Porosity
ation of fluidization phenomena. In: Weimer AJ, editor.
distributions in a fluidized bed with an immersed obstacle.
Fluidized processes: theory and practice. AIChE Symposium
AIChE J 1989;35(6):908 –22.
Series No. 289, vol. 88. New York: American Institute of
[32] Bouillard JX, Gidaspow D, Lyczkowski RW. Hydrodyn-
Chemical Engineers; 1992. p. 1–16.
amics of fluidization: fast bubble simulation in a two-
[14] Tarmy BL, Coulaloglu CA. Alpha –omega and beyond
dimensional fluidized bed. Powder Technol 1991;66:
industrial view of gas/liquid/solid reactor development.
107 –18.
Chem Engng Sci 1992;47:3231–46.
[15] Dellenback PA, Johansen WR. Surface pressure and wear [33] Lyczkowski RW, Gamwo IK, Gidaspow D. Analysis of flow
characteristics of tubes submerged in bubbling fluid beds. in a non-aerated hopper containing a square obstacle. Powder
Powder Technol 2001;115:298–311. Technol 2000;112:57 –62.
[16] Jansson SA. Erosion and erosion–corrosion in fluidized bed [34] Gidaspow D. Multiphase flow and fluidization, continuum
combustor systems. In: Levy AV, editor. Proceedings on and kinetic theory descriptions. San Diego, CA: Academic
Corrosion –Erosion-Wear of Materials in Emerging Fossil Press; 1994.
Energy Systems. Houston: National Association of Corrosion [35] Enwald H, Peirano E, Almstedt AE. Eulerian two-phase flow
Association Engineers; 1982. p. 548–60. theory applied to fluidization. Int J Multiphase Flow 1996;
[17] Volpicelli G, Massimilla L, Zenz FA. Nonhomogeneities in 22(Suppl. 21):21–66.
solid –liquid fluidization. In: Buckham JA, Levitz NM, [36] Rivard WC, Torrey MD. K-FIX: a computer program for
editors. Fluidized-bed technology. CEP Symposium Series, transient, two-dimensional, two-fluid flow. Los Alamos
vol. 62(67). New York: American Institute of Chemical Scientific Laboratory Report LA-NUREG-6623; April 1977.
Engineers; 1966. p. 42 –50. [37] Bouillard JS, Lyczkowski RW, Folga S, Gidaspow D, Berry
[18] Loew O, Schmutter B, Resnick W. Particle and bubble GF. Hydrodynamics of erosion of heat exchanger tubes in
behavior and velocities in a large-particle fluidized bed with fluidized-bed combustors. Can J Chem Engng 1989;67:
immersed obstacles. Powder Technol 1979;22:45–57. 218 –29.
[19] Merry JMD, Davidson JF. Gulf stream circulation in shallow [38] Harlow FH, Amsden AA. Numerical calculation of multi-
fluidized beds. Trans Inst Chem Engr 1973;51:361 –8. phase fluid flow. J Comput Phys 1975;17:19– 52.
[20] Patrose B, Caram HS. Optical fiber probe transit anemometer [39] Lyczkowski RW, Gidaspow D, Solbrig CW, Hughes EC.
for particle velocity measurements in fluidized beds. AIChE J Characteristics and stability analyses of transient one-
1982;28(4):604–9. dimensional two-phase flow equations and their finite
[21] Lin JS, Chen MM, Chao BT. A novel radioactive particle difference approximations. Nucl Sci Engng 1978;66:378–96.
tracking facility for measurement of solids motion in [40] Rietma K, Mutsers SMP. The effect of interparticle forces on
fluidized beds. AIChE J 1985;31(3):465–73. expansion of a homogeneous gas-fluidization. Proceedings of
the International Symposium on Fluidization, Toulouse, [62] Neilson JH, Gilchrist A. Erosion by a stream of solid
France; 1973. p. 32–3. particles. Wear 1968;11:111–22.
[41] Panelli R, Filho FA. A study of a new phenomenological [63] Sheldon GL, Finnie I. The mechanism of material removal
compacting equation. Powder Technol 2001;114:255 –61. in the erosive cutting of brittle materials. Trans ASME, Ser
[42] Shinora K. Rheological property of particulate solids. In: 13, J Engng Ind 1966;88:393– 400.
Fayed ME, Otten L, editors. Handbook of powder science [64] Soo SL. A note on moving dust particles. Powder Technol
and technology. New York: Van Nostrand Reinhold; 1984. 1977;17:259–63.
p. 129– 69. [65] Wood RT, Woodford DA. Tube erosion in fluidized beds.
[43] Gidaspow D, Syamlal M. Solid–gas critical flow. Presented General Electric Co. Report 11/ET-FUC79. US Energy
at the 1985 Annual Meeting, AIChE, Paper 74e; November Research and Development Administration Report 81-12;
1985. December 1980.
[44] Altiner HK, Davidson JF. Powder flow from an aerated [66] Fiore NF, Anthony NC, Kosel TH. Abrasion in multi phase
hopper. In: Grace JR, Madsen JM, editors. Fluidization. New alloys. In: Levy AV, editor. Proceedings on Corrosion–
York, NY: Plenum Press; 1980. p. 461–8. Erosion-Wear of Materials in Emerging Fossil Energy
[45] Guenther C, Syamlal M. The effect of numerical diffusion on Systems, Berkeley, CA, January 17 –19, 1982. Houston:
simulation of isolated bubbles in a gas–solid fluidized bed. National Association of Corrosion Engineers; 1982.
Powder Technol 2001;116:142–54. p. 266–94.
[46] Ivanov VA, Vasquez SA. A pressure-based method to solve [67] Hutchings I. Some comments on the theoretical treatment of
granular multiphase flows on unstructured meshes. Paper 895 erosive particle impacts. Paper No. 36, Proceedings of the
presented at ICMF-2001, Fourth International Conference on Fifth International Conference on Erosion by Liquid and
Multiphase Flow; May 27–June 1 2001. Solid Impact, Cambridge, England; September 1979.
[47] Gidaspow D, Lin CL, Seo YC. Fluidization in two- [68] Wood RT, Woodford DA. Tube erosion in fluidized beds.
dimensional beds with a jet. 1. Experimental porosity Paper No. 42, Proceedings of the Fifth International
distributions. I&EC Fundam 1983;22:187–93. Conference on Erosion by Liquid and Solid Impact, Cam-
[48] Buyevich YA, Korolyov VN, Syromyatnikov NI. Hydro- bridge, England; September 1979.
[69] Gansley RR, O’Brien TJ. A model for bubble-induced
dynamic conditions for the external heat exchange in
erosion in fluidized-bed combustors and comparison with
granular beds. In: van Swaaij WPM, Afghan NH, editors.
experiment. Wear 1990;137:107– 27.
Heat and mass transfer in fixed and fluidized beds.
[70] Tabakoff W. Effect of environmental particles on a radial
Washington, DC: Hemisphere; 1986. p. 333 –42.
compressor. In: Levy AV, editor. Proceedings on Corrosion–
[49] Humphrey JAC. Fundamentals of fluid motion in erosion by
Erosion-Wear of Materials at Elevated Temperatures,
solid particle impact. Int. J Heat Fluid Flow 1990;11:170– 95.
Berkeley, CA, January 31–February 2, 1990. Houston, TX:
[50] Engel PA. Impact wear of materials. Amsterdam: Elsevier;
National Association of Corrosion Engineers; 1991.
1978.
p. 26.1–26.28.
[51] Sarkar AD. Friction wear. New York: Academic Press; 1980.
[71] Wolak J, Worm P, Patterson I, Bodia J. Parameters affecting
[52] Meng HC, LudemaKC. Wear models and predictive equations:
the velocity of particles in an abrasive jet. Trans ASME, J
their form and content. Wear 1995;181–183:443–57.
Engng Mater Technol 1977;99:147–52.
[53] Padhye AR. Literature review of fundamental erosion/
[72] Tsai W, Humphrey JAC, Cornet I, Levy AV. Experimental
abrasion models as applicable to wear of fluidized bed measurement of accelerated erosion in a slurry pot tester.
internals. Topical Report, EG&G WASC, Inc., Morgantown, Wear 1981;68:289– 303.
W.Va; December 1985. [73] Hussain MF, Tabakoff W. Computation and plotting of solid
[54] Finnie I. The mechanism of erosion of ductile metals. particulate flow in rotating cascades. Comput Fluids 1974;2.
Proceedings of the Third National Congress on Applied [74] Grant G, Tabakoff W. Erosion predictions in turbomachinery
Mechanics, New York: American Society of Mechanical resulting from environmental solid particles. J Aircraft 1975;
Engineers; 1958. p. 527 –32. 12:642–76.
[55] Finnie I. Erosion of surfaces by solid particles. Wear 1960;3: [75] Tabakoff W. Erosion behavior of materials in multiphase flow.
87–103. In: Levy AV, editor. Proceedings on Corrosion–Erosion-
[56] Pourahmadi F, Humphrey JAC. Modeling solid – fluid Wear of Materials in Emerging Fossil Energy Systems,
turbulent flows with application to predicting erosion wear. Berkeley, CA, January 17 – 19 1982. Houston: National
Physico Chem Hydrodyn 1983;4(3):191– 219. Association of Corrosion Engineers; 1982. p. 642–76.
[57] Finnie I. Some observations on the erosion of ductile metals. [76] Stringer J. Electric Power Research Institute, Palo Alto, CA.
Wear 1972;19:81–90. Personal communication; October 25, 1985.
[58] Finnie I, McFadden DH. On the velocity dependence of [77] Sheldon GL, Maji J, Crowe CT. Erosion of a tube by gas
ductile metals by solid particles at low angles of incidence. particle flow. Trans ASME, Ser H, J Engng Mater Technol
Wear 1978;48:181–90. 1977;99:138–42.
[59] Shewmon P, Sundararajan G. The erosion of metals. Annu [78] Crowe CT, Sharma MP, Stock DE. The particle-source-in
Rev Mater Sci 1983;13:303–18. cell (PSI-cell) model for gas-droplet flows. J Fluids Engng
[60] Bitter JGA. A study of erosion phenomena, Part I. Wear 1977;99:325–32.
1963;6:5–21. [79] Rudinger G, Chang A. Analysis of non-steady two-phase
[61] Bitter JGA. A study of erosion phenomena, Part II. Wear flow. Phys Fluids 1964;7:1747–54.
1963;8:161– 90. [80] Gosman AD, Pun WM. Calculation of recirculating flows
(Lecture notes). Imperial College, London, Report HTS/74/ Department of Chemical Engineering, The University of
2; 1974. British Columbia; May 1988.
[81] Mason JS, Smith BV. The erosion of beds by pneumatically [100] Yates JG. On the erosion of metals in fluidized beds. Chem
conveyed suspensions of abrasive particles. Powder Technol Engng Sci 1987;42(2):379–80.
1972;6:323– 35. [101] Darton RC, Lanauze RD, Davidson JF, Harrison D. Bubble
[82] Ushimaru K, Crowe CT, Bernstein S. Design and appli- growth due to coalescence in fluidized beds. Trans Inst Chem
cations of the novel slurry jet pump. Energy International, Engng 1977;55:274 –80.
Inc., Report EI84-108; October 1984. [102] Parkinson MJ, Napier BA, Jury AW, Kempton TJ. Cold
[83] Kick F. Dar Gesetz der proportionalen Widerstande und seine model studies of PFBC tube erosion. Proceedings of the
Anwendung. Germany: Leipzig; 1885. Eighth International Conference on Fluidized-Bed Com-
[84] Walker DR, Shaw MC. A physical explanation of the bustion, vol. II. Springfield, VA: National Technical
empirical laws of comminution. AIME Mining Engng Trans Information Service; 1985. p. 730–8; DOE/METC-85/
1954;6:313– 20. 6021.
[85] Rabinowicz E. Friction and wear of materials. New York: [103] Zhu J, Lim CJ, Grace FR, Lund JA. Tube wear in gas
Wiley; 1965. fluidized beds. II. Low velocity impact erosion and semi-
[86] Shook CA, Ghosh SR, Pilling FE. Wall erosion in slurry empirical model for bubbling and slugging fluidized beds.
couette flow. In: Roco MC, editor. Liquid–solid flows and Chem Engng Sci 1991;46(4):1151–6.
erosion wear in industrial equipment. New York: American [104] Zhu J, Grace R, Lim CJ. Tube wear in gas fluidized beds.
Society of Mechanical Engineers; 1983. p. 63–8. I. Experimental findings. Chem Engng Sci 1990;45(4):
[87] Bouillard JX, Lyczkowski RW. On the erosion of heat 1003–15.
exchanger tube banks in fluidized beds. Powder Technol [105] Foust AS, Wenzel LA, Clump CW, Maus L, Andersen LB.
1991;68:37– 51. Principles of unit operations. New York: Wiley; 1960. Fig.
[88] Mulhearn TO, Samuels LE. The abrasion of metals: a model B-12; p. 537.
of the process. Wear 1962;5:478–98. [106] Woodford AA, Wood RT. Effect of particle size and hardness
on material erosion in fluidized beds. Proc 7th Int Conf
[89] Von Rittinger PR. Lehrbuch der Aufbereitungskunde. Berlin:
Erosion Liquid Solid Impact 1983;56:1–10.
Ernstand Korn; 1867.
[107] British Coal Corporation, Coal Research Establishment.
[90] Bird RB, Stewart WE, Lightfoot EN. Transport phenomena,
Minimizing erosion in coal-fired boilers. Commission of
3rd ed. New York: Wiley; 1960.
the European Communities Report EUR 12360EN, Contract
[91] Lyczkowski RW, Bouillard JX, Gidaspow D, Berry GF.
No. 7220-ED/808. Final Report, Directorate-General, Tele-
Computer modeling of erosion in fluidized beds. Argonne
communications, Information Industries and Innovation,
National Laboratory Report ANL/ESD/TM-1, Argonne, IL,
Luxembourg; 1989.
July 1986; January 1990.
[108] Podolski WF, Lyczkowski RW, Montrone E, Drennen J,
[92] Lyczkowski RW, Bouillard JX. Interim User’s manual for
Ai YH, Chao BT. A study of parameters influencing metal
FLUFIX/MOD1: a computer program for fluid – solids
wastage in fluidized bed combustors. In: Anthony EJ,
hydrodynamics. Argonne National Laboratory, ANL/EES-
editor. Proceedings of the 11th International Conference
TM-361, Argonne, IL, October 1986 (published February
on Fluidized Bed Combustion, vol. 2. New York:
1989). American Society of Mechanical Engineers; 1991.
[93] Lyczkowski RW, Bouillard JX, Chang SL, Folga SM. User’s p. 609–18.
manual for EROSION/MOD1: a computer program for fluid [109] Lyczkowski RW, Bouillard JX. Scaling and guidelines for
solids erosion. Argonne National Laboratory Sponsor Report, erosion in fluidized beds. Paper 935 presented at ICMF-2001,
Argonne, IL (April 1992). Reprinted by USDOE METC as Fourth International Conference on Multiphase Flow; May
DOE/MC/24193-3500/NTIS No. 94000032. Available from 27 –June 1, 2001.
NTIS, Springfield, VA; 1994. [110] Ding J, Lyczkowski RW. Three-dimensional kinetic theory
[94] Savage SB. Granular flow at high shear rates. New York: modeling of hydrodynamics and erosion in fluidized beds.
Academic Press; 1982. p. 339 –57. Powder Technol 1992;73:127 –38.
[95] Baumeister T, editor. Mark’s standard handbook got mech- [111] Ding J, Gidaspow D. A bubbling model using kinetic theory
anical engineers, 7th ed. New York: McGraw Hill; 1967. of granular flow. AICHE J 1990;36(4):523–38.
[96] Parkinson MJ, Jury AW, Napier BA, Kempton TJ, Holder JC. [112] Rogers WA. The prediction of wear in fluidized beds. Trans
Cold model erosion studies in support of pressurized ASME J Pressure Vessel Technol 1995;117:142–9.
fluidized bed combustion. Electric Power Research Institute [113] Cody GD, Goldfarb DJ, Storch Jr. GV, Norris AN. Particle
Final Report for Project, 1337-2; April 1986. granular temperature in gas fluidized beds. Powder Technol
[97] Grace JR. In: Hetsoroni G, editor. Fluidized-bed hydrodyn- 1996;87:211–32.
amics in handbook of multiphase systems. New York, NY: [114] Ding J. A fluidization model using kinetic theory of granular
McGraw-Hill; 1982. p. 8.5–8.64. chapter 8.1. flow. PhD Thesis. Illinois Institute of Technology, Chicago;
[98] Parkinson MJ, Grainger JFG, Jury AW, Kempton TJ. Tube 1990.
erosion at IEA Grimethorpe: cold model studies at CRE. [115] Neri A, Gidspow D. Riser hydrodynamics: simulation using
Reports Commissioned by the Project from Outside kinetic theory. AIChE J 2000;46:52–67.
Consultants and Others, vol. 2, NCB (AEA Grimethorpe) [116] Wu Y, Gidaspow D. Hydrodynamic simulation of methanol
Ltd., Barnsley, S. Yorkshire, UK; September 1984. synthesis in gas– liquid slurry bubble column reactors. Chem
[99] Zhu J. Tube erosion in fluidized beds. PhD Thesis. Engng Sci 2000;55:573 –87.
[117] Drennen JF, Hocking WR, Howard DA. Particle motion at [132] Enwald H, Almstedt AE. Fluid dynamics of a pressurized
fluidized bed tube surfaces. Final Report prepared for fluidized bed: comparison between numerical solutions from
Morgantown Energy Technology Center, Contract DE- two-fluid models with experimental results. Chem Engng Sci
AC21-86MC23034, Combustion Engineering, Inc., Windsor, 1999;54:329–42.
CT; 1989. [133] Johansson K, Magnusson A, Rundqvist R, Almstedt AE.
[118] Rong D, Mikami T, Horio M. Particle and bubble movements Study of two gas–particle flows using Eulerian/Eulerian two-
around tubes immersed in fluidized beds—a numerical study. fluid models. Paper 329 presented at ICMF-2001, Fourth
Chem Engng Sci 1999;54:5737–54. International Conference on Multiphase Flow; May 27–June
[119] Rong D, Horio M. Behavior of particles and bubbles around 1, 2001.
immersed tubes in a fluidized bed at high temperature and [134] Chang SL, Lyczkowski RW, Berry GF. Spectral dynamics
pressure: a DEM simulation. Int J Multiphase Flow 2001;27: of computer-simulated, two-dimensional, few-tube flui-
89–105. dized-bed combustors. Int J Heat Mass Transfer 1991;34:
[120] Tsuji Y, Kawaguchi T, Tanaka T. Discrete particle 1773–81.
simulation of two-dimensional fluidized bed. Powder [135] Lyczkowski RW, Bouillard JX. State-of-the-art review and
Technol 1993;77:79– 87. status of erosion modeling in fluid/solids systems. In: Levy
[121] Tsuji Y. Activities in discrete particle simulation in Japan. AV, editor. Proceedings on Corrosion –Erosion-Wear of
Powder Technols 2000;113:278–86. Materials at Elevated Temperatures, Berkeley, CA, January
[122] Walton O, Braun RL. Simulation of rotary-drum and repose 31 –February 2, 1990. Houston, TX: National Association of
tests for frictional spheres and rigid sphere clusters. DOE/ Corrosion Engineers; 1991. p. 27.1–27.66.
NSF Workshop on Flow of Particulates on Fluids, [136] Olowson PA. Influence of pressure and fluidization velocity
September 29 –October 1, 1993, Proceedings. Available on the hydrodynamics of a fluidized bed containing
from NTIS, US Department Commerce, Springfield, VA; horizontal tubes. Chem Engng Sci 1994;49(15):2437–46.
1994. p. 131–48. [137] Olowson PA. Hydrodynamics of a fluidized bed—influence
[123] Joseph DD. Lift correlations from direct numerical simu- of pressure and fluidization velocity. PhD Thesis. Chalmers
lation of solid–liquid flow. Paper 385 presented at ICMF- University of Technology, Goteborg, Sweden; 1991.
2001, Fourth International Conference on Multiphase Flow; [138] Olowson PA, Almstedt AE. Influence of pressure and
May 27–June 1, 2001. fluidization velocity on the bubble behaviour and gas flow
[124] Olsson SE, Wiman J, Almstedt AE. Hydrodynamics of a distribution in a fluidized bed. Chem Engng Sci 1990;45(7):
pressurized fluidized bed with horizontal tubes: influence of 1733–41.
pressure, fluidization velocity and tube-bank geometry. [139] Enwald H. A numerical study of the fluid dynamics of
Chem Engng Sci 1995;50(4):581 –92. bubbling fluidized beds. PhD Thesis. Chalmers University of
[125] Wiman J, Mahpour B, Almstedt AE. Erosion of horizontal Technology, Goteborg, Sweden; 1997.
tubes in a pressurized fluidized bed-influence of pressure, [140] Lyczkowski RW, Folga S, Chang SL, Bouillard JX, Wang
fluidization velocity and tube-bank geometry. Chem Engng CS, Berry GF, Gidaspow D. State-of-the-art computation of
Sci 1995;50(21):3345–56. dynamics and erosion in fluidized bed tube banks. In:
[126] Wiman J. Hydrodynamics, erosion and heat transfer in Manakar AM, editor. Proceedings of the 10th 1989
pressurized fluidized beds. PhD Thesis. Chalmers University International Conference on Fluidized Bed Combustion,
of Technology, Goteborg, Sweden; 1997. vol. 1. New York: American Society of Mechanical
[127] Wiman J, Almstedt AE. Hydrodynamics, erosion and heat Engineers; 1989. p. 465–78.
transfer in a pressurized fluidized bed: influence of pressure, [141] Lyczkowski RW, Bouillard JX, Berry GF, Gidaspow D.
fluidization velocity and tube-bank geometry. Chem Engng Erosion calculations in a two-dimensional fluidized bed. In:
Sci 1997;52:2677– 95. Mustonen JP, editor. Proceedings of the Ninth 1987
[128] Gustavsson M, Almstedt AE. Numerical simulation of fluid International Conference on Fluidized-Bed Combustion,
dynamics in fluidized beds with horizontal heat exchanger vol. 2. New York: American Society of Mechanical
tubes. Chem Engng Sci 2000;55:857 –66. Engineers; 1987. p. 697–706.
[129] Gustavsson M, Almstedt AE. Two-fluid modelling of [142] Gustavsson M. Fluid dynamic mechanisms of particle flow
cooling-tube erosion in a fluidized bed. Chem Engng Sci causing ductile and brittle erosion. Wear 2002;252:845–58.
2000;55:867–79. [143] Lyczkowski RW, Folga S, Chang SL, Bouillard JX.
[130] Gustavsson M. Fluidized bed operating parameters and Computational of dynamics and erosion for small tube
Eulerian erosion models. PhD Thesis. Chalmers University arrays in fluidized beds. Fifth Miami International Sym-
of Technology, Goteborg, Sweden; 2000. posium on Multiphase Transport and Particulate Phenomena,
[131] Enwald H, Peirano E, Almstedt AE, Leckner B. Simulation Miami Beach, FL; December 12–14, 1988. In: Veziroglu
of the fluid dynamics of a bubbling fluidized bed. TN, editor. Proceedings of Condensed Papers; 1988. p. 147–
Experimental validations of the two-fluid model and 8. Veziroglu TN, editor. Full paper in Multiphase Transport
evaluation of a parallel multiblock solver. Chem Engng Sci and Particulate Phenomena, vol. 3, New York: Hemisphere;
1999;54:311 –28. 1990. p. 325 –53.

State of The Art Review of Erosion Modeling in Fluid Solid Systems PDF

Caricato da

Informazioni sul documento

Titolo 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

State of The Art Review of Erosion Modeling in Fluid Solid Systems PDF

Caricato da

Copyright:

Formati disponibili

Progress in Energy and Combustion Science 28 (2002) 543–602

State-of-the-art review of erosion modeling in fluid/solids systems

Received 11 June 2001; accepted 5 September 2002

* Corresponding author. Tel.: þ 1-630-252-5923; fax: þ1-630-252-9281.

2. Modeling of fluidized-bed solids motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548

1. Background terms which came into use in the petroleum industry in

Nomenclature K1 constant given by Eq. (3.10b), (m/s)1/2

x lateral coordinate, m sf tensile strength at failure, Pa

The treatment of the pressure gradient term in the gas Table 1

velocity, 26.0 cm/s. There are no solids entering, hence, the

accounted for in the model. Hence, the formation of a vortex

the obstacle. Also revealed is a larger general solids

3.1. Single-particle erosion models E_ t ¼ 2E_ ¼ 2W=A

M{V~ 2 cos2 ðaÞ 2 K1 ½V~ sinðaÞ 2 Vel 3=2 }

where fc is the material-dependent cutting wear factor. The

C0 ¼ 0:288ðrp =sy Þ1=4 =sy ð3:10aÞ

K1 ¼ 8:036s2y Er22 ðsy =rp Þ1=4 ð3:10bÞ

The angle a0 may be estimated from Finnie’s model (Eqs.

Finnie’s erosion model can be written as

Fig. 13. Computed solids velocities time averaged over 2.5 s.

conclusion that it is highly two-dimensional. Hence, the

For erosion on a horizontal surface (shown in Figs. 12 and

E_ PD ¼ E_ SPD þ E_ IPD ð5:21bÞ ›~vs

2 1s rs ~vs g~ 2 7·½ð1s t SV Þ·~vs  ð5:26Þ UEDKE ¼ ð2dKEs =dtÞdp ¼ EEDKE dp : ð5:29Þ

UEDvREL ¼ ½ð1s t SV Þ : 7~vs þ 1s b B ðvf 2 ~vs Þ2 dp

It has been our experience that the evaluation of Eq. (5.32a)

6. Implementation of erosion models in

The implementation of single particle and differential

where UEDa and EEDa are given by Eqs. (5.32a) – (5.32d)

Fig. 16. Coupling of hydrodynamics and erosion models.

Material Crystal structure Diamond pyramid hardness Composition

Aluminum fcc 18.5 99.99% Al

the form, modified for Hydrodynamic Model B [31,87] as and

ð1 2 1Þð1 2 1gd Þ ð1 2 1gd Þ

In addition, as fs decreases, the void fraction, 1, increases

Fig. 31. Three-dimensional non-uniform-finite difference grid.

An alternative to extending single particle erosion

model to include a constant applied force. The MED erosion Acknowledgements

Potrebbero piacerti anche

M{V~ 2 cos2 ðaÞ 2 K1 ½V~ sinðaÞ 2 Vel 3=2 }

2 1s rs ~vs g~ 2 7·½ð1s t SV Þ·~vs ð5:26Þ UEDKE ¼ ð2dKEs =dtÞdp ¼ EEDKE dp : ð5:29Þ

UEDvREL ¼ ½ð1s t SV Þ : 7~vs þ 1s b B ðvf 2 ~vs Þ2 dp