Energy Storage and Saving 1 (2022) 44–52

Contents lists available at ScienceDirect

Energy Storage and Saving

journal homepage: http://www.keaipublishing.com/en/journals/energy-storage-and-saving/

Research article

Numerical study of particle fouling deposition on heat transfer surface

Zhiming Xu, Jinhui Li, Zhimin Han∗
School of Energy and Power Engineering, Northeast Electric Power University, Jilin, 132012, China

a r t i c l e i n f o a b s t r a c t

Keywords: An integrated Eulerian model for fouling in turbulent flow was developed to simulate the particle fouling charac-
Particle deposition teristics. The model comprises a deposition process and removal process. The deposition process considers four
Eulerian approach various factors affecting the particle deposition which are Brownian and eddy diffusion, gravity, thermophoretic,
Tube
and turbophoretic forces. This study compares the Lagrangian and Eulerian approaches under the same operat-
Numerical simulation
ing conditions with an emphasis on their performance in predicting particle deposition. The deposition height
calculated by using the Eulerian approach is in good agreement with the experimental data and the deposition
morphology is similar to that observed in experiments. Furthermore, this paper explores the deposition charac-
teristics and predicts the deposition morphology under different particle sizes.

1. Introduction the behavior and mechanism of particle deposition in a ventilation duct

with turbulent flow, analyzing several important parameters affecting
Fouling is a common problem in heat exchanger equipment, and par- this process. The above review of the literatures shows that many re-
ticle fouling occurs because of the solid particle deposition on a heat searchers have used the Eulerian approach to study the deposition be-
transfer surface. This is problematic, as ash deposition increases the ther- havior of particles. But most of them are based on the three-layer model
mal resistance of heating surfaces, and decreases the flow cross-section by Lai and Nazaroff [5], and some empirical relationships are used, such
area, which can lead to the blockage of the gas duct by increasing the as those for the local fluctuation velocity normal to the wall, relative slip
flow resistance of the flue gas. In extreme cases, this can cause heat ex- velocity, temperature gradient perpendicular to the surface and friction
changer failure and even result in the forced shutdown of the furnace, velocity. Till date, published literature does not document the use of the
hence seriously affecting the safety and economy of industrial produc- Eulerian approach to fully describe the process of particles flow across
tion processes [1]. Therefore, it is of great significance to study the de- tube, which was characterized by the different deposition direction from
position characteristics of particle fouling on the heat transfer surface. position to position. The main purpose of this study, therefore, is to de-
Over the years, many numerical studies have been conducted to bet- velop a complete fouling model to simulate the particle fouling on tube
ter understand the physics of particle deposition and fouling formation. surface by Eulerian–Eulerian approach. The model is an improvement
The deposition mechanism and model of particle fouling on the heat on the three-layer model for particle deposition by considering the dif-
transfer surface appeared as early as 1988 [2]. In the Eulerian approach, ferent factors that affect particle deposition. Subsequently, some impor-
particle transport was governed by the species conservation equation. tant parameters are calculated directly from the CFD model by the UDF
Therefore, from a computational point of view, the Eulerian approach code instead of relying on empirical formulas. Finally, the Eulerian ap-
was more effective. A comprehensive review of the physical processes proach was applied to predict particle fouling on the tube surface.
responsible for the transport and deposition of particles and their the-
oretical modeling was given by Guha [3]. In the work of Vargas et al.
[4], two-dimensional particle transport was predicted using the Eule- 2. Mathematical model and simulation method
rian approach. A modified model was proposed by Zhao and Wu, based
on the three-layer model by Lai and Nazaroff [5], that accounts for 2.1. Governing equations
turbophoretic as well as Brownian diffusion, turbulent diffusion, and
gravitational settling [6]. A new type of Eulerian–Eulerian model was The conservation equations solved by the Eulerian-Eulerian model
applied in the study to model the transport and distribution character- for each phase include the continuity and momentum equations, as fol-
istics of particle matter in indoor spaces [7]. Lu and Wang [8] reviewed lows [7]:
𝜕( ) 𝜕 ( )
𝛼 𝜌 + 𝛼 𝜌 𝑢 =0 (1)
𝜕𝑡 b b 𝜕 𝑥𝑖 b b 𝑖
Peer review under responsibility of Xi’an Jiaotong University.
∗
Corresponding author.
E-mail address: [email protected] (Z. Han).

https://doi.org/10.1016/j.enss.2021.11.001
Received 10 October 2021; Received in revised form 25 November 2021; Accepted 26 November 2021
Available online 3 December 2021
2772-6835/© 2021 The Authors. Published by Elsevier B.V. on behalf of Xi’an Jiaotong University. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

Nomenclatures 𝑣̄ ′2+
p non-dimensional particle mean-square velocity
xf deposition height, m
a the distance in the y direction from a position to the
center of the furnace, m Greek symbols
c specific heat capacity, J/(kg⋅K) 𝛼 volume fraction, 𝛼 f =1−𝛼 p
C particle concentration, kg/m3 𝛽 acute angle between the normal outer direction of node
C+ dimension-less particle concentration and x-axis, °
C∞ mainstream particle concentration, kg/m3 𝛿 the thickness of the first layer of grids near the tube wall,
Cc Stokes-Cunningham slim correction factor m
CD the non-linear drag coefficient 𝜀p the eddy viscosity of the fluid, m2 /s
d0 effective diameter of air molecules, m 𝜆f thermal conductivity of fluid, W/(m⋅K)
dp particle diameter, 𝜇m 𝜆p thermal conductivity of particles, W/(m⋅K)
DB Brownian diffusivity of particles, m2 /s 𝜇 dynamic viscosity, kg/(m⋅s)
Ea surface adhesions activation energy, KJ 𝜇 eff effective kinematic viscosity, m2 /s
g gravitational acceleration, m/s2 𝜇t turbulent viscosity, given by 𝜇 t =a∗ 𝜌k/𝜔
h enthalpy, kJ 𝜌f density of flue gas, kg/m3
i correction coefficient 𝜌p density of particle, kg/m3
I turbulence intensity 𝜎k the turbulent Prandtl numbers for k
J the particle mass flux to the wall (deposition mass flux), 𝜎𝜔 the turbulent Prandtl numbers for 𝜔
kg/(m2 ⋅s) 𝜏+ non-dimensional relaxation time
J0 + non-dimensional Brownian and eddy diffusion deposi- 𝜏L the Lagrangian integral time scale, s
tion flux 𝜏p the particle relaxation time, s
k the removal constant 𝜏w the wall shear force, Pa
kd surface adhesion constant, m2 /s2 𝜐 kinematic viscosity, m2 /s
KB Boltzmann constant (1.38 × 10−23 J/K) 𝜓 strength bond factor of the deposit layer
Kn the Knudsen number (Kn=2l0 /dp ) Subscript
l0 the mean free path of gas molecules, m f fluid phase
𝑚̇ d deposition mass rate, kg/(m2 ⋅s) p particle phase
𝑚̇ f net mass rate, kg/(m2 ⋅s) w wall
𝑚̇ r removal mass rate, kg/(m2 ⋅s) dep deposition
P pressure, Pa
R the normal distance from the tube wall, m Superscript
R+ dimension-less normal distance from the wall − Reynolds averaging
R0 the gas constant, J/(mol⋅K) + dimensionless
Re Reynolds number, Re =ud/𝜐
Rep Reynolds number based on velocity of particles relative
to gas ( )
𝜕( ) 𝜕 ( ) 𝜕𝑃 𝜕 𝜕𝑢
S the modulus of the mean rate-of-strain tensor 𝛼 𝜌 𝑢 + 𝛼 𝜌 𝑢𝑢 =− b + 𝜇b 𝑖 − 𝜌 b 𝑢 ′ 𝑖 𝑢 ′ 𝑗 (2)
𝜕𝑡 b b 𝑖 𝜕 𝑥𝑗 b b 𝑖 𝑗 𝜕 𝑥𝑖 𝜕 𝑥𝑗 𝜕 𝑥𝑗
Sp the sticking probability
t time, s where b represents the phase under consideration (with values of f and p
T absolute temperature, K for the fluid and particle phases, respectively). 𝛼 b is the volume fraction
Ts surface temperature, K of phase b.
u the average velocity at the inlet, m/s Neglecting the heat transfer between the particle and fluid phases,
u∗ the friction velocity, m/s the energy equation was solved only for the fluid phase. Consequently,
uf fluid velocity, m/s
( )
un the tangential velocity of the first inner inner grid near 𝜕( ) 𝜕 ( ) 𝜕 𝜆f 𝜕 ℎf
𝜌ℎ + 𝜌 𝑢ℎ = (3)
the wall, m/s 𝜕𝑡 f f 𝜕 𝑥𝑖 f 𝑖 f 𝜕 𝑥𝑖 𝑐f 𝜕 𝑥𝑖
up particle velocity, m/s
uw the velocity at the wall, m/s FLUENT provides many choices of turbulence models [9]. As the k-
Vd + non-dimensional Brownian and eddy diffusion deposi- 𝜔 model is more advantageous for wall-restricted boundary layer and
tion velocity free shear flow, and can more accurately predict flow separation, the
Vdep the particle deposition velocity, m/s turbulence model selected in this paper is the SST k-𝜔 model, which
Vdep + non-dimensional particle deposition velocity may be expressed as follows [10]:
[( ) ]
Vs gravity deposition velocity, m/s 𝜕 (𝜌𝑘) 𝜕 ( ) 𝜕 𝜇 𝜕𝑘
Vs + non-dimensional gravity deposition velocity + 𝜌𝑘𝑢𝑖 = 𝜇+ 𝑡 + 𝜇𝑡 𝑆 2 − 𝜌𝛽 ∗ 𝑓𝛽 ∗ 𝑘𝑤 (4)
𝜕𝑡 𝜕 𝑥𝑖 𝜕 𝑥𝑗 𝜎𝑘 𝜕 𝑥 𝑗
Vt turbophoretic deposition velocity, m/s
[( ) ]
Vt + non-dimensional turbophoretic deposition velocity 𝜕 (𝜌𝜔) 𝜕 ( ) 𝜕 𝜇𝑡 𝜕𝜔 𝜔
+ 𝜌𝜔𝑢𝑖 = 𝜇+ + 𝑎 𝜇𝑡 𝑆 2 − 𝜌𝛽𝑓𝛽 𝑤2 (5)
Vthx thermophoretic deposition velocity in x direction, m/s 𝜕𝑡 𝜕 𝑥𝑖 𝜕 𝑥𝑗 𝜎𝜔 𝜕 𝑥𝑗 𝑘
Vthy thermophoretic deposition velocity in y direction, m/s
Vth + non-dimensional thermophoretic deposition velocity 2.2. Particle deposition model
𝑣̄ ′2
f
fluid mean-square velocity, m2 /s2
𝑣̄ ′2+
f
non-dimensional fluid mean-square velocity The direction of deposition is different for each node on the tube
𝑣̄ ′2
p particle mean-square velocity, m2 /s2 wall, and is along the inner normal direction of the node, denoted as R.
However, the deposition parameters of the particles are normally calcu-
lated under the Cartesian (x-y) coordinate system; hence, it is necessary

45
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

to calculate the deposition parameters of each node in the R-direction where 𝜇 eff is the effective viscosity, uw is the velocity at the wall, which
corresponding to its position. considered to be zero, and un is the tangential velocity of the first layer
If the particle flux in the normal direction is J, the mass transfer of the grid near the tube wall. The velocity of each node in the x and
rate is kept unchanged in R-direction of the tube wall in order to satisfy y directions of the tube wall is first calculated, and then the tangential
the continuity equation and achieve steady-state flow in the direction velocity is calculated according to the size, direction, and location of the
perpendicular to the tube wall [3], i.e.: joint velocity of each node.
𝜕𝐽 The dimensionless forms of the parameters mentioned above are as
=0 (6) follows:
𝜕𝑅
The particle deposition velocity, Vdep , is defined as: 𝐶
𝐶+ = (17)
𝐶∞
|𝐽 (𝑅 = 0)|
𝑉dep = (7)
𝐶∞ 𝑉dep
𝑉dep
+
= (18)
Here, the following four particle deposition mechanisms were con- 𝑢∗
sidered: Brownian and eddy diffusion, gravity, thermophoretic, and tur-
bophoretic forces. 2.2.2. Gravity force
The deposition of small ash particles is mainly governed by diffu-
2.2.1. Brownian and eddy diffusion sion, while that of large particles is mainly affected by gravity. Since
The deposition under diffusion is usually calculated using the modi- the direction of gravity is vertically downward while that of particle
fied Fick diffusion law. Based on the assumption that the flux is constant deposition is in the R-direction (i.e., perpendicular to the surface), the
in the concentration boundary layer, the dimensionless Brownian and gravity deposition velocity, Vs needs to be revised, and can be expressed
eddy diffusion deposition velocity can be written as follows [5]: as:
( ) ( )1∕2
𝐽 + 𝐷B 𝜀p 𝜕 𝐶 +
𝑉d + = 0 = + (8) 4 𝑔 ⋅ 𝑑 p 𝜌p − 𝜌f
𝐶∞ 𝜐 𝜐 𝜕 𝑅+ 𝑉s = 𝑖 ⋅ 𝐶c ⋅ sin 𝛽 (19)
3 𝐶D 𝜌f
The first boundary condition assumes that the particle concentra-
Since the deposition surface in this model is curved, gravity promotes
tion is zero at the position where the particles contact the wall, i.e.,
deposition when the node is located on the upper wall of the tube bun-
C+ =0, and R+ = r+ (r+ = (dp /2) (u∗ /𝜐)). The second boundary condi-
dle, hence the correction coefficient i = 1. On the other hand, gravity
tion assumes that the particle concentration is equal to the value of the
inhibits deposition when the node is located on the lower wall of the
concentration in the main flow region outside the boundary layer, i.e.,
tube bundle, hence i = −1. At the left and right end points, gravity does
C+ = 1, R+ = 200. Therefore, Eq. (8) can be transformed into:
not work, hence i = 0 [8]. CD is the drag coefficient, which is calculated
1 200 ( )
via the modified Stokes equation [13]:
1 d𝐶 + 𝜐
= = d𝑅+ (9)
𝑉d + ∫0 𝑉d + ∫𝑟+ 𝐷B + 𝜀p
⎧24∕Re Rep ≤ 1
DB is the Brownian diffusivity of particles: ⎪ (p )
𝐶D = ⎨24 × 1+0.15Rep 0.678 ∕Rep 1 < Rep ≤ 1000 (20)
𝐾 𝑇𝐶 ⎪0.45 1000 < Rep
𝐷B = B c (10) ⎩
3𝜋𝜇𝑑p
| |
Cc is the Stokes-Cunningham slim correction factor which can be Rep = 𝜌f 𝑑p |𝑢f − 𝑢p |∕𝜇 (21)
| |
expressed as:
( ( )) The dimensionless gravity deposition velocity, Vs + , is defined as:
𝑙 𝑑p
𝐶c =1+ 0 2.514 + 0.8 × exp −0.55 (11) 𝑉s
𝑑p 𝑙0 𝑉 s + =𝐶 + (22)
𝑢∗
l0 is the mean free path of flue gas molecules:
𝐾 𝑇 2.2.3. Thermophoretic force
𝑙0 = √ B (12)
2𝜋𝑑02 𝑃 Due to the temperature gradient between the fluid and the tube wall,
the particles move towards the region of low temperature, in a process
The eddy viscosity of the fluid, 𝜀p , is calculated as follows [11]: known as thermophoretic deposition. The temperature gradients in x
and y directions can be extracted from a user-defined function (UDF)
in the commercial CFD software package ANSYS FLUENT, and the ther-
mophoretic deposition velocities in each direction are then calculated
(13) using the following equations [14–15]:
𝐶c 𝜐𝐻 d𝑇
The dimension-less normal distance from the wall, R+ , can be calcu- 𝑉thx = (23)
𝑇 d𝑥
lated by:
𝛿𝑢∗ 𝐶c 𝜐𝐻 d𝑇
𝑅+ = (14) 𝑉thy = (24)
𝜐 𝑇 d𝑦
where 𝛿 is the distance between the first layer grid and the wall, and The dimensionless thermophoretic deposition velocity along the R-
the friction velocity u∗ is a key parameter, which can be calculated as direction is given by:
follows [12]: ( )
√ 𝐶 𝜐𝐻 𝐶 𝜐𝐻
𝐶 + c𝑇 dd𝑇𝑥 cos 𝛼 + c𝑇 dd𝑇𝑦 sin 𝛼
𝜏w
𝑢∗ = (15) 𝑉th =
+
(25)
𝜌𝑙 𝑢∗
where 𝜏 w is the wall shear force, which is defined as:
( )( 𝜆f ∕𝜆p + 2.18𝐾𝑛
)
𝑣 − 𝑣w 2.34
𝜏w = 𝜇ef f n (16) 𝐻= (26)
𝛿 1 + 3.42𝐾𝑛 1 + 2𝜆f ∕𝜆p + 4.36𝐾𝑛

46
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

2.2.4. Turbophoretic force Table 1

Caporaloni et al. [16] first discovered the turbophoretic phe- Geometric dimensions and calculation parameters.
nomenon of particle transport in 1975. The turbophoretic deposition Parameter Value Parameter Value
process involves the mass transfer from a high to a low turbulence in-
Tube diameter 22 mm Particle concentration 0.02 g/m3
tensity area. The turbophoretic deposition velocity, Vt , is calculated as
Tube wall temperature 773 K Deposition layer density 778.98 kg/m3
follows: Flue gas inlet temperature 793 K Apparent particles density 1692 kg/m3
Porosity of deposition layer 0.54 Specific heat capacity of particles 840 J/kg⋅K
𝜕 𝑣′ p 2
𝑉t = −𝜏p (27)
𝜕𝑅
To obtain Vt , it is necessary to first obtain the particle-wall normal
2.3.2. Particle fouling removal mass rate
fluctuating velocity intensity, 𝑣′ p 2 , which can be estimated using the
Particle deposition is a dynamic (time-variant) process, and the num-
following equation [11]:
ber of particles deposited depends on both the deposition and removal
𝑣′2
p 1 rates, as the particles deposited on the tube wall may leave again due
= ( ) (28) to the scouring effect of the incoming flow and collisions with other
𝑣′2 1 + 0.7 𝜏p ∕𝜏L
f particles. The fouling mass removal rate, 𝑚̇ r , is given by [21–22]:
The particle relaxation time 𝜏 p [17] and the Lagrangian time scale ( )
𝜏
of fluid turbulence 𝜏 L [18] are expressed as: 𝑚̇ r = 𝑘 w 𝑥f (39)
𝜓
𝐶𝑐 𝜌p 𝑑p2
𝜏p = (29) The combination of k and 𝜓 is considered to be a constant by Wang
18𝜇 et al. [23], and its value was assumed to be 0.75 here.
𝜀t
𝜏L = (30)
𝑣′2 2.3.3. Particle fouling net deposition mass rate
R
The net mass rate of the fouling growth process is equal to the dif-
The dimensionless forms of the parameters in Eq. (28) are defined
ference between the deposition and removal mass rates, which can be
as follows:
expressed as [24]:
+ 𝑣′2 p
𝑣′2 p = (31) 𝑚̇ f = 𝑚̇ d − 𝑚̇ r (40)
𝑢∗2
+ 𝑣′2 f 2.4. Simulation method
𝑣′2 f = (32)
𝑢∗2
The r.m.s. fluctuation velocity is given by Guha [11], which is a The FLUENT software, in conjunction with UDF, was used to investi-
function of R+ in this paper: gate the particle deposition in a two-dimensional unsteady state process.
( )2 The calculation grid was generated using the pre-processing geometry
+ 0.005𝑅+2 and mesh generation software GAMBIT. The thickness of the first layer
𝑣′2 f = 0 < 𝑅+ < 200 (33)
1 + 0.002923𝑅+2.218 near the wall was 0.1 mm, and it was densified at the boundary layer
Then, the dimensionless turbophoretic deposition velocity, Vt + , is with a ratio of 1.1, so that the value of R+ would be near 1 to ensure
defined as follows: calculation accuracy. The Eulerian multiphase model was adopted, with
[( ( ))] +
the primary and secondary phases being flue gas and particles, respec-
d 𝜏L ∕ 𝜏p + 𝜏L 𝑣′2 R
𝑉 t =𝜏
+ +
𝐶+ (34) tively. A second-order upwind scheme was used to solve the continuity,
d𝑅+
momentum, and energy equations. When the residual error of each vari-
The non-dimensional relaxation time 𝜏 can be calculated by:
+
able was < 10−6 , it was considered to be convergent.
𝑢∗ 2 In this study, the first grid spacing (R+ ) is 1 and the grid growth
𝜏 + = 𝜏p (35)
𝜐 factor is 1.2.
2.3. Particle fouling model
3. Model description and solution method
2.3.1. Particle fouling deposition mass rate
Considering the four different deposition mechanisms, the total di- 3.1. Geometric model and boundary conditions
mensionless deposition velocity is as follows:
( ) In order to simulate the ash deposition in the furnace and tail flue of
𝐽+ ( ) 𝜕𝐶 + 𝑉 𝑉thy
𝑉dep + = = − 𝐷B + 𝜀P − 𝑖 ⋅ ∗s × 𝐶 + + ∗ × 𝐶 + the boiler, this paper took the settling furnace as a model and simplified
𝐶∞ 𝜕𝑅 𝑢 𝑢
it. The geometric model is shown in Fig. 1. The total length of the chan-
𝑉thx 𝑉t
× sin 𝛼 − ∗ × 𝐶 × cos 𝛼 − ∗ × 𝐶
+ +
(36) nel was 1,400 mm, the width was 50 mm, and a tube with a diameter of
𝑢 𝑢 22 mm was arranged 160mm away from the bottom. Since the flow and
After the deposition velocity is determined using the equation Vdep = convective heat transfer of flue gas mainly occurred on the section per-
Vdep + × u∗ , the deposition mass rate 𝑚̇ d is calculated as follows [19]: pendicular to the tube; hence, the flow field and temperature changes in
𝑚̇ d =𝑆p ⋅ 𝐽 = 𝑆p ⋅ 𝑉d ⋅ 𝐶∞ (37) the axial direction of the tube were ignored, and the model was simpli-
The sticking probability Sp is defined as the ratio of the adhesive fied into two dimensions perpendicular to the tube direction. The geo-
force to the resistance acting on particles [20], which can be calculated metric dimensions and calculation parameters are shown in Table 1. The
by: calculation area consisted of four boundaries: inlet, outlet, side walls,
and tube wall. The simulated flue gas component was air, and it was
𝑒 (− 𝐸 a ∕ 𝑅 0 𝑇 s )
𝑆 p =𝑘 d ⋅ (38) assumed that the air entering the calculation area had uniform velocity
𝑢p 2 uin , temperature Tin , turbulence intensity I = 5%, and that the velocity
The particle velocity at each node on the tube wall is different, which component perpendicular to the flow direction was zero. The simula-
causes a difference in the adhesion probability of each node. Hence, the tion time step was set to 1 × 10−4 s. A non-slip boundary condition with
rate of deposition of the particles is not only related to the deposition constant temperature was assumed at the tube wall. The polynomial ex-
velocity, but also the particle velocity. pression after fitting the temperature in both sides of the furnace T (°C)

47
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

Fig. 1. Geometric model.

Table 2 Table 3
The grid convergence index analysis. Simulation boundary conditions under the different working conditions.

r 𝜏 w (Pa) 𝜀 GCI1 (%) u∗ (m⋅s−1 ) 𝜀 GCI2 (%) Working condition 1 2 3

−3 −3
A1 1.134 1.513 −3.981 × 10 1.770 1.021 −5.911 × 10 2.628 Flue gas velocity (m/s) 1.14 1.57 2.00
A2 1.178 1.507 7.900 × 10−3 2.593 1.015 3.925 × 10−3 1.288 Primary air temperature (°C) 300 300 300
A3 — 1.519 — — 1.019 — — Fly ash concentration (g/m3 ) 0.108 0.076 0.058

Note: The fly ash concentration values under the three working conditions are
converted into ash feed rate of 360 mg/min.

gies at different times obtained using the Eulerian approach are shown
in Fig. 3(a), Fig. 4(a), and Fig. 5(a) for working conditions 1, 2, and
3, respectively. The deposition heights at different times obtained from
the experiments [27] and from Yang’s study [25] using a Lagrangian
simulation approach are compared with the Eulerian approach used in
the present study in Fig. 3(b), Fig. 4(b), and Fig. 5(b). The experimental
images under three working conditions are shown in Fig. 6, Fig. 7, and
Fig. 8, respectively.
As can be seen from Figs. 3(a)–5(a), the ash deposition morpholo-
gies simulated by the Eulerian approach are similar to those obtained
by experiment [27]. Although unlike those in the experiments, the simu-
lated deposition morphologies are symmetrical. It can be inferred that,
due to the use of completely symmetrical boundary conditions in the
simulations, a symmetrical reflux flow was formed around the tube.
However, the asymmetrical distribution of deposition morphologies in
Fig. 2. (a) Computational grid; (b) Local grid enlargement. the experiments is due to the influence of error factors and secondary
air velocity. The average deviations between the simulation results and
and the distance in the y direction from a position to the center of the the experimental data were 15.58%, 11.93%, and 10.22% under work-
furnace a (m) is as follows [25]: ing conditions 1, 2, and 3, respectively. In the experiments, the mini-
𝑇 (𝑦)=1069.19 − 93.62𝑎 − 1890.32𝑎2 + 176.38𝑎3 + 1589.98𝑎4 (41) mum and maximum sizes of the measured particles were 1.445 𝜇m and
275.423 𝜇m, respectively, indicating a large particle size range. In the
3.2. Grid generation and grid independence analysis simulations, however, a single average particle size 85.719 𝜇m was as-
sumed, which may contribute to the discrepancy between the experi-
In order to eliminate the influence of the number of grids on the cal- mental and simulation results. In addition, it can be seen from the three
culation results, this paper uses the grid convergence index (GCI) [26] to figures (Figs. 3–5) that the deposition height tends to increase with the
analyze the independence of the grid. This paper has established three increase of inlet flow velocity, and the difference between working con-
sets of grids, and the calculation results of GCI are shown in Table 2. ditions 2 and 3 is not obvious when the flow time is 30 min. The reason
The calculated GCI values of the average shear force on the tube wall is that although both the deposition rate and the removal rate increase
were 1.770% and 2.593%, respectively, and the GCI were both less than with time, the deposited mass within 30 min is less, and the removal rate
3%; GCI values of the average friction velocity on the tube wall calcu- is very small compared to the deposition rate. Therefore, although the
lated were 2.628% and 1.288%, respectively, and the GCI was less than removal rate increases with time at a larger rate, but the net deposition
3%. That is, the variation range of the calculation results is very small, mass still increases with time. Moreover, as shown in Figs. 3(b)–5(b),
the grid convergence criterion is satisfied, and the third set of grids is the average deviation between the experimental data and the simula-
considered to be independent. Therefore, the number of grids selected tion results in this paper is less than that between experimental data
in the numerical simulation process is 85,979, and the magnified view and the simulation results using Lagrangian approach.
of the calculation grid and the partial grid is shown in Fig. 2(a) and (b). It can be seen from Fig. 6 that at the initial stage of working condition
1, due to the low flow velocity without being affected by the blow-off
4. Results and discussion effect of the flow field, the surface of the deposition layer is uneven,
and then gradually becomes smooth. The whole deposition layer is dis-
4.1. Analysis of particle deposition morphology at different inlet flow tributed like mountain peaks and slightly to the right tilt. As can be seen
velocity from Fig. 7, the growth rate of the deposition layer in working condi-
tion 2 began to increase at the later stage, and the deposition height was
The boundary conditions for the three different working conditions higher than that in working condition 1 at the later stage. It can be seen
used in the simulations are shown in Table 3. The deposition morpholo- from Fig. 8 that the deposition rate of working condition 3 shows a slow-

48
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

Fig. 3. Comparison of experiment and simulation results under working condition 1: (a) deposition morphology simulated using the Eulerian approach; (b) compar-
ison of deposition height between experiments and simulations.

Fig. 4. Comparison of experiment and simulation results under working condition 2: (a) deposition morphology simulated using the Eulerian approach; (b) compar-
ison of deposition height between experiments and simulations.

ing down trend in the later deposition process, and basically maintains before the flow time of about 37 h, the total net deposition mass of 1 𝜇m
the same deposition height in the last five minutes, and the deposition particles is larger than that of 100 nm particles, while the total net depo-
height may have reached the maximum value. sition mass of 100 nm particles is greater after the flow time of 37 h. The
Overall, the data show that the value of deposition height obtained deposition height of particles with particle size of 1 𝜇m in the early flow
using the Eulerian approach is in good agreement with the experimental stage is larger. According to Eq. (39), the removal rate increases with
data, as the deviation is within a reasonable range. Further, the depo- the increase of deposition height. The rate of increase of removal mass
sition morphology is similar to that observed in experiments, which in- gradually increases, resulting in the rate of increase of total net deposi-
dicates that the Eulerian deposition model can accurately characterize tion mass gradually decreasing. Therefore, after 37 h of flow time, the
the deposition process. net deposition mass of 100 nm was larger than that of 1 𝜇m particles.
After the flow time reaches 37 h, the net deposition mass of particles
4.2. Effects of particle diameter on particle deposition characteristics with a particle size of 100 nm is the largest, followed by particles of
1 𝜇m, and the net deposition mass of particles of 10 nm is the smallest.
In order to explore the difference in the deposition characteristics of In addition, it can be seen from Fig. 9 that the deposition mass curve
fly ash particles under different particle sizes, the deposition height is of 1 𝜇m particles has a more obvious asymptotic trend, while the net
calculated when the inlet flow velocity is 5 m/s and the particle concen- deposition mass growth rate of 10 nm and 100 nm particles decreases
tration is 0.02 g/m3 (converted to the ash feed rate is 300 mg/min). The weakly, and it still does not reach an asymptotic value at a flow time of
total net deposition mass at different moments at particle diameters are 50 h. It is speculated that because the wall shearing force of 10 nm and
10 nm, 100 nm, and 1 𝜇m shown in Fig. 9. It can be seen from Fig. 9 that 100 nm particles is small, the removal mass increases slightly with time,

49
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

Fig. 5. Comparison of experiment and simulation results under working condition 3: (a) deposition morphology simulated using the Eulerian approach; (b) compar-
ison of deposition height between experiments and simulations.

Fig. 6. Experimental images of condition 1 at various times.

Fig. 9. Variation of net deposition mass with time of different particle diame-
Fig. 7. Experimental images of condition 2 at various times. ters.

deposition morphologies are drawn, as shown in Fig. 10. It can be found

from Fig. 10 that the deposition characteristics of different particle sizes
are quite different. At the three moments shown in the figures, the parti-
cle deposition height with a diameter of 1 𝜇m is the largest, and a single-
sided wedge-shaped deposition layer is formed on the upper tube wall at
90°. The net deposition mass of particles with a particle size of 100 nm
is the second, followed by 10 nm. But the deposition morphologies of
10 nm and 100 nm particles are different from that of 1 𝜇m particles,
the deposits with these two particle sizes are relatively small, and the
Fig. 8. Experimental images of condition 3 at various times. particles form a double-sided wedge-shaped soot layer on the upper tube
wall at 90° and the lower tube wall at 270°, which was consistent with
the typical deposition morphologies in reference [28].
and the difference between the amount of removal and the deposition The deposition morphology is closely related to the particle velocity
is large, so it takes a long time to reach an asymptotic value. distribution, and the position where the maximum deposition height is
In order to analyze the deposition process of particles with different usually the position where the particle velocity is the minimum. When
particle sizes, when the particle diameters are 1 𝜇m, 100 nm, and 10 nm, the flue gas flows through the tube, the flow velocity is zero at 90° of
taking the flow time of 250 s, 2,500 s, and 5,000 s as examples, the de- the upper tube wall, namely the stagnation point, where the pressure is
position height along each node of the tube wall is calculated and the the largest and the direction of gravity at this position is the direction

50
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

Fig. 10. The deposition morphology at different time of particle diameter 1 𝜇m, 100 nm and 10 nm.

of deposition. At about 270° on the lower wall, the fluid velocity in the
boundary layer decreases to 0, forming a new stagnation point. Due to
the big downstream pressure of the separation point, the fluid is found
to backflow, forming a vortex area. When the particle size is 1 𝜇m, the
particle size in the vortex zone is not enough to overcome its gravity and
is trapped. No obvious deposition occurs in the lower wall stagnation
zone, so the single wedge-shaped ash deposition morphology is formed.
On the contrary, the particles with small particle size are easily ensnared
and thus forming double wedge-shaped ash deposition morphology.
For exploring the reasons for the above-mentioned deposition mor-
phologies, taking a flow time of 5,000 s (1.39 h) as an example, the
dimensionless deposition velocity of different particle sizes under the
action of four deposition mechanisms is calculated, as shown in Fig. 11.
It can be seen from Fig. 11 that the diffusion deposition velocity of
10 nm particles is the largest, and the larger the particle size, the fewer
the number of particles reaching the heat exchange surface under the ac-
tion of Brownian diffusion and eddy diffusion. The dimensionless grav-
itational deposition velocity is obviously the largest when the particle
size is 1 𝜇m, while the other kinds of nano-particles are very weakly af-
fected by gravity due to the small particle size. In addition, the gravity
deposition velocity needs to be corrected according to the node coordi- Fig. 11. Effects of four deposition mechanisms on dimensionless deposition ve-
nates. Gravity force inhibits the deposition on the bottom of the tube, locity under different particle diameters.
and the gravity deposition rate of 1 𝜇m particles on the bottom wall is
greater than the sum of the deposition rates of the other three mech-
anisms, so 1 𝜇m particles hardly deposit on the bottom of tube wall. shearing force. The wall shearing force of 1 𝜇m particles is the largest,
Thus, only a single-sided wedge-shaped soot layer is formed. Further- and the removal mass is also the largest. Therefore, after considering the
more, the gravity deposition velocity of the other particle sizes is lower. corrected gravity deposition velocity and the influence of the removal
Although part of the gravity deposition velocity under the action of the mass. After the flow time of 40 h, the net deposition mass of particles
other three deposition mechanisms is offset, they can still be deposited with a particle size of 100 nm is the largest, followed by 1 𝜇m particles,
on the lower wall to form a double-sided wedge-shaped soot layer, and and the net deposition mass of 10 nm particles is the smallest.
the smaller the particle size is, the greater the deposition height on the
5. Conclusion
lower wall is. Since the inlet temperature and wall temperature under
different particle sizes are the same, the temperature gradient is not
In this study, an Eulerian approach is employed to analyze the parti-
much different. However, the irregular thermal motion of small parti-
cle deposition characteristics on heat transfer surface. In this approach,
cles is more obvious, so the dimensionless thermophoretic deposition
deposition mechanisms such as Brownian and eddy diffusion, gravity,
velocity of particles with a particle size of 10 nm is slightly larger. The
thermophoretic, and turbophoretic forces are considered. Numerical
smaller the particle diameter, the greater the value of the dimension-
simulations based on the Eulerian approach are firstly carried out on
less thermophoretic deposition velocity. In addition, the turbophoretic
three working conditions with different inlet velocities for fly ash depo-
effect of particles with a particle size of 1 𝜇m is the most significant, and
sition on the tube surface. The results show that the deposition mor-
the smaller the particle size, the weaker the effect of turbophoretic on
phologies at different times under all three conditions showed good
particle deposition. In summary, the dimensionless diffusion and ther-
agreement with the experimental data. The average deviations in the
mophoresis deposition velocity of 10 nm particles is the largest, and the
deposition height between the simulation results and the experimen-
dimensionless gravity and turbophoretic deposition velocities of 1 𝜇m
tal values under the three different working conditions are 15.58%,
particles are the largest. In general, the sum of the four dimensionless
11.93%, and 10.22%, respectively. Taking the flow time of 250 s, 2,500 s
deposition rates of 1 𝜇m particles is the largest, followed by 100 nm and
and 5,000 s as examples, the deposition morphologies at three moments
10 nm. In addition, the net deposition mass of particles is not only re-
when the particle size is 10 nm, 100 nm, and 1 𝜇m are drawn respec-
lated to the corrected dimensionless deposition velocity, but also related
tively, and it is found that the particles with a particle size of 1 𝜇m
to the removal mass. The larger the particle size, the greater the wall
showed a single-side wedge-shaped ash deposit layer, while the other

51
Z. Xu, J. Li and Z. Han Energy Storage and Saving 1 (2022) 44–52

two particle sizes all presented a double-side wedge-shaped ash deposit [8] H. Lu, Y.H. Wang, Particle deposition in ventilation ducts: a review, Build. Simul.
layer. In addition, this paper analyses the influence of the four deposi- 12 (2019) 723–734.
[9] FLUENT Incorporation, FLUENT User’s Guide, FLUENT Incorporated, Lebanon, NH,
tion mechanisms, and it is found that for particles with smaller diame- USA, 2011.
ters, the Brownian and eddy deposition and thermophoretic force have [10] D.C. Wilcox, Turbulence modeling for CFD, DCW Industries, 2006.
a greater impact, while for particles with larger diameters, the effect of [11] A. Guha, A unified, A unified Eulerian theory of turbulent deposition to smooth and
rough surfaces, J. Aerosol Sci. 28 (1997) 1517–1537.
gravity and turbophoretic forces are greater. [12] Z.M. Han, Z.M. Xu, X.Y. Yu, CFD modeling for prediction of particulate fouling
of heat transfer surface in turbulent flow, Int. J. Heat Mass Transfer 144 (2019)
Declaration of Competing Interest 118428.1-118428.9.
[13] Z.M. Xu, Z.M. Han, H.W. Qu, Comparison between Lagrangian and Eulerian ap-
proaches for prediction of particle deposition in turbulent flows, Powd. Technol.
The authors declared that there are no conflicts of interest. 360 (2019) 141–150.
[14] M. Abdolzadeh, M.A. Mehrabian, A.A. Solghar, Prediction of particle deposition us-
ing a simplified particle model in fully developed channel flow, Chem. Eng. Com-
Acknowledgments
mun. 202 (2014) 294–302.
[15] L. Talbot, R.K. Cheng, R.W. Schefer, et al., Thermophoresis of particles in a heated
The authors sincerely express gratitude to the National Natural Sci- boundary layer, J. Fluid Mech. 101 (2006) 737–758.
ence Foundation of China (Grant No.: 51976028) for its financial sup- [16] M. Caporaloni, F. Tampieri, F. Trombetti, et al., Transfer of particles in nonisotropic
air turbulence, J. Atmos. Sci. 32 (2010) 565–568.
port. [17] S.T. Johansen, The deposition of particles on vertical walls, Int. J. Multiphase Flow
17 (1991) 355–376.
CRediT author statement [18] L. Tian, G. Ahmadi, Particle deposition in turbulent duct flows – comparisons of
different model predictions, J. Aerosol Sci. 38 (2007) 377–397.
[19] M. Shirdel, D. Paes, P. Ribeiro, et al., Evaluation and comparison of different models
Zhiming Xu: Conceptualization, Supervision, Project administra- for asphaltene particle deposition in flow streams, J. Pet. Sci. Eng. 84–85 (2012)
tion. Jinhui Li: Data curation, Methodology, Investigation, Writing- 57–71 s.
[20] H. Seyyedbagheri, B. Mirzayi, Eulerian model to predict asphaltene deposition pro-
Original draft preparation. Zhimin Han: Software, Validation, Visual- cess in turbulent oil transport pipelines, Energy Fuels 31 (2017) 8061–8071.
ization, Writing- Reviewing and Editing. [21] C. Paz, E. Suárez, A. Eirís, et al., Development of a predictive CFD fouling model for
diesel engine exhaust gas systems, Heat Transfer Eng 34 (2013) 674–682.
References [22] M. Bohnet, Fouling of heat transfer surfaces, Chem. Eng. Technol. 10 (1987)
113–125.
[23] F.L. Wang, Y.L. He, Z.X. Tong, et al., Real-time fouling characteristics of a typical
[1] P. Stehlík, Conventional versus specific types of heat exchangers in the case of pol-
heat exchanger used in the waste heat recovery systems, Int. J. Heat Mass Transfer
luted flue gas as the process fluid – a review, Appl. Therm. Eng. 31 (2010) 1–13.
104 (2017) 774–786.
[2] N. Epstein, Particulate fouling of heat transfer surfaces: mechanisms and models,
[24] D.Q. Kern, R.E. Seaton, A theoretical analysis of thermal surface fouling, Br. Chem.
Springer, Netherlands, 1988.
Eng. 4 (1959) 258–262.
[3] A. Guha, Transport and deposition of particles in turbulent and laminar flow, Annu.
[25] Q. Yang, Research on deposition of fly ash by dynamic mesh techque & immersed
Rev. Fluid Mech. 40 (2008) 311–341.
boundary method, Harbin Institute of Technology, Harbin, 2016 Dissertation.
[4] F.M. Vargas, J.L. Creek, W.G. Chapman, On the development of an asphaltene de-
[26] I. Celik, U. Ghia, P.J. Roache, et al., Procedure of estimation and reporting of uncer-
position simulator, Energy Fuels 24 (2010) 2294–2299.
tainty due to discretization in CFD applications, J. Fluids Eng. 130 (2008) 078001.
[5] A. Lai, W.W. Nazaroff, Modeling indoor particle deposition from turbulent flow onto
[27] L. Wang, Experimental study on characteristics of coal ash deposition, Harbin Insti-
smooth surfaces, J. Aerosol Sci. 31 (2000) 463–476.
tute of Technology, Harbin, 2015 Dissertation.
[6] B. Zhao, J. Wu, Modeling particle deposition from fully developed turbulent flow in
[28] S. Kalisz, M. Pronobis, Investigation on fouling rate in convective bundles of coal–
ventilation duct, Atmos. Environ. 40 (2006) 457–466.
fired boilers in relation to optimization of sootblower operation, Fuel 84 (2005)
[7] S.L Sharma, M. Ishii, T. Hibiki, et al., Beyond bubbly two-phase flow investigation
927–937.
using a CFD three-field two-fluid model, Int. J. Multiphase Flow 113 (2019) 1–15.

52