Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Wear 460–461 (2020) 203450

Contents lists available at ScienceDirect

Wear
journal homepage: http://www.elsevier.com/locate/wear

CFD modelling of the influence of particle loading on erosion using dense


discrete particle model
Hassan Pouraria a, Farzin Darihaki b, Ki Heum Park c, Siamack A. Shirazi b, Yutaek Seo c, *
a
Department of Chemical and Material Engineering, New Mexico State University, Las Cruces, NM, USA
b
The Erosion/Corrosion Research Center, Department of Mechanical Engineering, The University of Tulsa, Tulsa, OK, 74104, USA
c
Department of Naval Architecture and Ocean Engineering, Seoul National University, 1 Gwanak-ro Gwanak-gu, Seoul, 08826, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: The present study aims at modelling erosion using Dense Discrete Particle Model (DDPM). The numerical model
Erosion modelling is based on Eulerian-Eulerian-Lagrangian approach which has been developed for modelling both dilute and
Dense discrete particle model dense particulate flows. In this approach, the continuous phase is modelled on a fixed Eulerian grid and particles
Particle concentration
are tracked by using Lagrangian model. However, the particulate phase is also represented on the Eulerian grid
Erosion reduction
and the inter-particle interactions are modelled by employing the kinetic theory of granular flows. Numerical
simulations are first performed for a low Stokes number flow of slurry jet test and a high Stokes number flow of
dry jet test. The numerical results for slurry jet test indicate a reasonable accuracy of the model in predicting the
erosion distribution. Furthermore, the obtained numerical results suggest a reduction of erosion ratio at high
particle loadings which agree well with the experimental data in this study. However, the predicted erosion
reduction for the high Stokes number flow of dry jet test is significantly higher than that in low Stokes number
slurry flow. The present results indicate that there exist a critical particle loading beyond which the erosion
reduction becomes significant (more than 20%). However, the magnitude of the critical particle loading for the
high Stokes number flow is significantly lower than that in low Stokes number flow. Moreover, the influence of
particle loading on the erosion ratio and the flow-field in an elbow conveying two-phase gas-sand flow is
investigated. The numerical results clearly revealed a reduction of erosion ratio at high particle loading condi­
tions. Furthermore, a comparison of the obtained results by using DDPM and one-way coupled DPM model shows
the importance of considering the influence of inter-particle collisions as wells as coupling between the fluid and
particles at high particle loading conditions.

1. Introduction the erosion rate [3,4]; 79; [5,6]. The proposed models have been used
along with the CFD models to predict the erosion patterns under
Sand production is one of the major challenges for the safe operation different working conditions. A comprehensive review of the published
of flow-lines. The existence of sand particles along with produced fluids works can be found in Ref. [2].
can potentially result in several issues [1,2]. Erosion, due to the sand CFD is known to be the most comprehensive approach for predicting
particles, is one of major issues that pose a great menace to the oil and the vulnerable spots in piping systems and estimating the associated
gas production systems. Underestimating the damage from erosion may damage from erosion. CFD models can also be used to minimize such
result in catastrophic health, safety and environmental problems. Hence, damages by optimizing the geometry and the flow conditions of the
any attempt towards predicting the erosion damage in a quantitative system [2,7–9]. Thus far, several researchers used CFD to investigate the
manner is of great importance for oil and gas production as well as erosion pattern in elbow, plugged tee, valves, etc [10–14]. The accuracy
several other industries. of different erosion models has been evaluated by many researchers
Due to the importance of erosion, several experimental studies were [15–20]. Furthermore, the influence of several factors such as geomet­
conducted to identify the influential factors affecting the severity of rical and flow conditions have been investigated [11,15,17,18,21–24].
erosion. Furthermore, several models have been proposed to quantify The ultimate strength of the damaged structures has also been

* Corresponding author.
E-mail address: yutaek.seo@snu.ac.kr (Y. Seo).

https://doi.org/10.1016/j.wear.2020.203450
Received 5 March 2020; Received in revised form 17 August 2020; Accepted 18 August 2020
Available online 29 August 2020
0043-1648/© 2020 Elsevier B.V. All rights reserved.
H. Pouraria et al. Wear 460–461 (2020) 203450

investigated in several works [25–28]. also differs from DEM since in DDPM particles are not tracked individ­
Eulerian-Lagrangian approaches have been used by many re­ ually but groups of parcels, each containing many particles, are tracked
searchers to predict the erosion pattern. In this approach, the fluid phase in a Lagrangian fashion and the inter-particle collisions are modelled
is modelled by using the Eulerian model and the particles are tracked by indirectly using the computationally less demanding KTGF model.
Lagrangian framework. Traditionally, a one-way coupled DPM is used Hence, this approach is potentially suitable for comparatively large
for erosion prediction in which the effect of inter-particle collision is systems. Furthermore, using this approach will eliminate the inherent
neglected [10,11,13,15,18–20]. This model is very efficient for flows weakness of the traditional Eulerian-Eulerian method in modelling wall
with very low particle concentrations. However, as the particle con­ interactions and delta shock as well as fine grid requirement [45–47].
centration increases the accuracy of such models is under question. For Hence, unphysical merging as a result of averaging of the opposite
such conditions the Discrete Element Method (DEM) can be used as an momentum fluxes in Eulerian-Eulerian approach is avoided by applying
alternative Eulerian-Lagrangian model that is capable of modelling a direct simulation of particle parcels via Lagrangian framework. Apart
inter-particle interactions [29–31]; 78). Nevertheless, due to the high from this, due to the Lagrangian tracking of particles, particle size dis­
computational cost, application of such models is limited to the tribution can be easily included without adding additional momentum
small-scale simulations and the fundamental studies. As a result, such equations which is required in Eulerian-Eulerian approach, thereby
models are not yet suitable for simulating large scale flow-fields reducing the computational cost [1,46].
observed in oil and gas as well as many other industries. It is worth Although the overall volume fraction of the sand in the oil and gas
mentioning that the traditional DPM model can also be improved to take production is not generally high, the particle concentration in specific
into account the inter-particle collisions [32,33]. reported a successful regions of flow may become significantly high. Recent studies show that
implementation of inter-particle collision model via stochastic even for comparatively dilute particulate flows the volume fraction of
hard-sphere approach. particles near the hot spots of the elbows is significantly high so that the
Another alternative for modelling erosion in particulate flows is inter-particle collisions cannot be neglected [32]. Previous studies also
using Eulerian-Eulerian models [2]. In this approach, both phases are revealed that for plugged-tees the particle concentration is significantly
treated as interpenetrating continua resulting in a system of averaged higher than that in elbows [11,48]. The experimental study of [11] and
Navier-Stokes equation. In such model various empirical models ob­ the numerical study reported by Ref. [48] indicate a significant erosion
tained from rheological measurements may be used to evaluate the reduction in plugged tees at high particle loading conditions when
apparent shear viscosity of the particulate flow [34]. However, for more compared to that in elbows. Furthermore, at high particle concentra­
complex systems, appropriate closure laws based on the kinetic theory of tions, the influence of solid particles on the hydrodynamics of the
granular flows (KTGF) are adopted to close the system of equations. continuous phase becomes significant. Due to the importance of
Such model referred to as Eulerian-Granular models is widely used for handling highly concentrated slurries in oil sands industry, drilling and
modelling various particulate systems. This approach has been validated fracturing applications developing an alternative model for modelling
to give adequate representations of the complex flows occurring in the particulate flow-field with confidence is essential. The capability of
risers, bubbling fluidized beds as well as slurry flows [35–38]. Thus far, DDPM in modelling both dilute and dense particulate flows with
few researchers tried to use such approach for predicting the erosion rate reasonable computational cost makes this model a suitable candidate for
[39–42]. Although such models are reliable tools for predicting the investigating different industrial scale flow and geometry
overall flow pattern, they cannot give reliable information about the configurations.
particle impact velocity and the angle of impact at the wall. However, The objective of the present study is to investigate the influence of
determining these quantities is the ultimate goal of using CFD in erosion particle loading on the erosion ratio which is defined as the thickness
modelling. Hence, the predicted erosion rate using this approach is not loss per unit sand mass flow rate. To this end, the influence of particle
trustworthy. loading on erosion in jet impingement test is first studied for two
Recently [43], proposed a hybrid CFD model in which the particulate different cases including a low Stokes number flow (St = 26.5) in a
flow in regions without significant erosion was modelled by using the submerged slurry jet test and a high Stokes number flow (St = 1480) of
Eulerian-Eulerian approach while, in hot spots where the erosion occurs, dry jet test (sand in gas) by using both one-way DPM and DDPM.
the Lagrangian model was employed to accurately track the particles Furthermore, numerical modelling is performed for an elbow conveying
and avoid the uncertainties involved in Eulerian-Eulerian model. This gas-solid flow. The capability of DDPM in predicting the erosion pattern
methodology was further developed in Refs. [34]. as well as the influence of particle volume fraction on the erosion ratio is
Another alternative for modelling particulate flows is Eulerian- investigated. Moreover, the influence of particle loading on the distri­
Eulerian-Lagrangian approach known as DDPM. This model takes a bution of particle volume fraction and the velocity of the carrier fluid in
middle ground between DEM and Eulerian-Eulerian method. In this the elbow is investigated by using DDPM.
approach, the fluid phase is modelled on a fixed Eulerian grid and the
particles are tracked by using a Lagrangian model which is similar to the 2. Numerical model
traditional Eulerian-Lagrangian model known as DPM. However, on
contrary to the Eulerian-Eulerian model, the transport equations for the 2.1. Dense discrete particle method
particulate phase are not solved and the velocity and volume fraction of
the particulate phase obtained from the Lagrangian framework are The conservation equation for the continuous phase is written based
interpolated onto the Eulerian grid. Furthermore, inter-particle colli­ on the standard Eulerian multi-fluid model. For an incompressible flow
sions and translations are modelled using the well-established KTGF without mass transfer between the phases, the continuity and mo­
model. This methodology eliminates the inherent shortcomings of the mentum equations are written as follows:
traditional DPM framework that makes it valid only for very dilute
∂( ) ( )
particulate flows where the inter-particle interactions can be ignored. It α ρ + ∇ ⋅ α f ρf →
v f =0 (1)
∂t f f

∂( ) ( ) [ ( T )] →
α f ρf →
v f + ∇ ⋅ αf ρf → v f = − αf ∇p + ∇ ⋅ αf μf ∇→
v f→ v f + ∇→
vf + αf ρf →
g + F exchange (2)
∂t

2
H. Pouraria et al. Wear 460–461 (2020) 203450

where, α, ρ, v, and p indicate the volume fraction, density, velocity and represent shear and bulk viscosities arising from particle momentum
pressure, respectively. The subscript f denotes the fluid phase. On the exchange due to translation and collision. The.solids shear viscosity is
contrary to the standard multi-fluid approach, the conservation equa­ generally computed by summation of the collisional, kinetic, and fric­
tions for the particulate phase are not solved. As a matter of fact, the tional parts as follows [83].
volume fraction and the velocity field are directly obtained from the μs = μs,col + μs,kin + μs,fr (10)
Lagrangian approach in which the trajectory of each parcel is calculated
by integrating the force balance on the parcels. Furthermore, the ex­ 4 (θ )
(11)
s 1

change forces between two phases are obtained based on the employed
μs,col = αs ρs ds g0 (1 + ess ) ⋅ 2 ⋅αs
5 π
forces in DPM approach. The governing equation of the motion of par­ √̅̅̅̅̅̅̅ [ ]
cels is written as follows: αs ds ρs θs π 2
μs,kin = 1 + (1 + ess )(3ess − 1)αs g0 (12)
⃒ ⃒ →( ) 6(3 − ess ) 5
d→
us 3Cd ⃒→us− → v f⃒( → → ) g ρs − ρf ρf
ms = ms v f − u s + ms + ms → v f ∇→
vf Since, in the present study, the volume fraction of particles is less
dt 4 ds ρs ρs
→ than the packing limit, the effect of friction viscosity was neglected.
+ F KTGF Furthermore, for non-spherical particles an equivalent spherical particle
(3) diameter is considered as the solid particle size ds .
The first three terms on the right-hand side of the equation represents In the present study the bulk viscosity λs, that accounts for the
the drag force, buoyancy force and pressure gradient force, respectively. resistance of the granular particles to compression and expansions is
These terms are also observed in the traditional DPM approach. How­ calculated according to [53].
ever, the last term is added in DDPM framework to account for the ef­ 4 (θ )12
(13)
s
fects of inter-particle collisions and translations using the kinetic theory λs = α2s ρs ds g0 (1 + ess )
3 π
of granular flows (KTGF).
Several models can be used to quantify the drag coefficient, Cd . For
spherical and semi spherical particles as well as non-spherical particles 2.2. Solids pressure is calculated as follows
with low particle Reynolds numbers the model proposed by Gidapsow
[49] can be used which is a combination of [50,51] models. For fluid
volume fractions higher than 0.8, it resembles the [50] model and can be ps = αs ρs θs + 2ρs (1 + ess )α2s g0 θs (14)
written as follows:
where ess is the coefficient of restitution for particle collisions, g0 is the
radial distribution function and θs is the granular temperature. In the
2.65
24αs ρf α−f [ ( )0.687 ]
Cd = 1 + 0.15 αf Res (4)
Res present study the restitution coefficient was set to 0.9.
Radial distribution function is a correction factor that modifies the
where Res is the particle Reynolds number, which is defined as follows: probability of collisions between particles when the solid phase becomes

ρf d s ⃒→us− →

v f⃒ dense. In the present study the following description is used for this
Res = (5) function [54].
μ
⎡ ⎤− 1
For non-spherical particles the model proposed by Ref. [52] can be (
αs
)13
used to quantify the drag coefficient which reads as follows: g0 = ⎣1 − ⎦ (15)
αs,max
24 [ ] b3 Res
Cd = 1 + b1 Res b2 + (6)
Res b4 + Res where, αs,max denotes the packing limit and is set to 0.63.
As seen in the above equations, the solid pressure, shear viscosity and
Where the constants are defined as follows: bulk viscosity are calculated by quantifying the granular temperature.
( )
b1 = exp 2.3288 − 6.4581ϕ + 2.4486ϕ2 Analogous to the thermodynamic temperature for the gases, the concept
of granular temperature is based on the kinetic energy due to the
b2 = 0.0964 + 0.5565ϕ (7) random translational motion of the particles.

( ) 1
b3 = exp 4.905 − 13.8944ϕ + 18.4222ϕ2 − 10.2599ϕ3 θs = us,i us,i (16)
3
(
b4 = exp 1.4681 12.2584ϕ − 20.7322ϕ2 + 15.8855ϕ3 The default setting in ANSYS-Fluent uses algebraic expression for
modelling the granular temperature [80]. In the algebraic expression of
The shape factor, ϕ, is the ratio of the surface area of the equivalent the granular temperature, the convection and diffusion terms are
spherical particle, s, to the surface area of the non-spherical particle, S. neglected and it is assumed that the granular energy is dissipated locally
The KTGF model, developed for predicting the generated stresses due which can be valid for modelling dense fluidized beds where the volume
to inter-particle collisions and translation of spherical particles, is used fraction of the particles is high and the velocity is low [55–57]. How­
to calculate the exerted forces on the particles. The KTGF forces are ever, such approach is not reliable for modelling erosion in high flow
modelled from solid phase stress tensor. rates. In the present study the granular temperature was also determined
by using Lagrangian approach. Hence, the following equation in
→ 1 =
F KTGF = − ms ∇ τ s (8) Lagrangian framework is solved at every DPM iteration.
ρs
3 dθ =
The solid phase stress tensor includes the normal and shear viscous αs ρ s = τ s : ∇→
u s + ∇ ⋅ (kθs ∇θs ) − γ θs + ϕfs (17)
2 s dt
stresses as well as normal pressure:
( ) The first term at the right-hand side indicates the generation of
( T) 2 fluctuating energy by the solid stress tensor. The second term is energy
= =
τ s = − ps I + αs μs ∇→ u s + αs λs − μs ∇⋅→
u s + ∇→ (9)
=
v sI
3 diffusion and the third term represents the energy dissipation due to the
= inelastic particle-particle collisions and the fourth term is energy ex­
where ps indicates solid pressure, I is unit stress tensor. μs and λs change between fluid and particle. The magnitude of the granular

3
H. Pouraria et al. Wear 460–461 (2020) 203450

temperature on the Eulerian grid can be calculated as the mass weighted


HV + 0.1023
average of granular temperature considering all the parcels inside a cell BH = (20)
0.0108
[46].
The dispersed formulation of the standard k− ε model which uses the where BH, is the Brinell hardness of pipe material, C is an empirical
Tchen theory of the dispersion of discrete particles in a homogeneous constant and Hv is Vicker’s hardness of the pipe material. The model
turbulence is adopted for modelling the turbulence. The computational proposed by Ref. [19] suggests that for stainless steel (Hv = 1.83 GPa)
domain was initialized with the steady state solution of single-phase the appropriate values of C and K for sand-air flow are 4.62 × 10− 7 and
flow. Then, the unsteady state solution was carried out for two-phase 2.16 × 10− 8 , respectively. According to the [58] the appropriate value of
particulate flows. The transient simulations were conducted by adopt­
K for sand-water flow is 1.15 × 10− 8 .
ing equal time steps for the fluid flow and the Lagrangian particle Fs is the particle shape coefficient (sharpness factor), which is
tracking. The adopted time step for the slurry jet, gas jet and the elbows
considered equal to 0.2 for fully rounded particles, 0.53 for semi
were 1E-4S, 2E-5S and 1E-4S, respectively. The transient simulation of
rounded particles and 1 for sharp particles, respectively. Furthermore, f
DDPM model was carried out by injecting 1408 and 1216 parcels at each
(θ) is the impact angle function which accounts for the effect of impact
time step for the elbows and the impingement tests, respectively. A
angle on the erosion rate. A modified version of the angle function
trapezoidal discretization scheme was employed to solve equation (3).
suggested by Ref. [4] was recently proposed by Vieira el at., [19,58],
The pressure-velocity coupling in two-fluid model was achieved via
which reads as follows:
phase-coupled SIMPLE (PC-SIMPLE) algorithm which is the extension of (
Semi-Implicit Method for Pressure-Linked Equations (SIMPLE) algo­ 1
F(θ) = (sinθ)n1 1 + HV n3 (1 − sinθ))n2 (21)
rithm for multiphase flows. Bounded second order implicit scheme was f
employed for transient simulations. The under-relaxation factors were
The appropriate constants for sand-gas and sand-liquid flows are
reduced to half of their default values in the solver to ensure the stability
given tin Table 1.
of the numerical simulations. The ANSYS-Fluent 16.1 package was used
As the particles hit the fitting wall, their reflected velocity becomes
to solve the transport equations for the elbows and the slurry jet test. The
less than that of the incoming velocity. Previous studies revealed that
same transport equations were solved in ANSYS-Fluent 19.2 for the gas
the angle of impact has a significant effect on the coefficients of resti­
jet impingement test.
tution [61,62]. The respective perpendicular and parallel coefficients of
restitution for steel material are incorporated into the model according
2.3. One-way coupled discrete particle method to Refs. [63]:

Numerical simulation was also carried out using a one-way coupled eper = 0.988 − 0.78θ + 0.19θ2 − 0.02θ3 + 0.027θ4 (22)
Discrete Particle Method (DPM). The description of this model can be
found in several published papers [10,15,20]; Peng and Gao, 2016; [23]. epar = 1 − 0.78θ + 0.85θ2 − 0.21θ3 + 0.028θ4 − 0.022θ5 (23)
Hence, the detailed information is not presented for the sake of brevity.
In this approach Eq. (1) and Eq. (2) are solved along with the standard where θ is the impact angle of the particles. The impact angle function f
k− ε turbulence model to resolve the fluid motion. Then, Eq. (3) is solved (θ) and the restitution coefficient e were introduced to the model as
to find the trajectory of the particles. However, the last term in Eq. (3) piecewise linear functions.
which accounts for the inter-particle collisions and translations is not
included. The steady state simulation of the continuous phase is first 4. Geometry and boundary conditions
carried out and then the Lagrangian particle tracking is performed using
the resolved flow-field information. In the present study, two different experimental data were used to
In the present study 100,000 and 400,000 parcels were released from assess the accuracy of the CFD model. The experiment of [41] was used
the inlet to quantify the erosion rate in the elbow and the jet impinge­ to validate the accuracy of the model for slurry flows. The erosion test
ment tests, respectively. The adopted parcel numbers were based on was conducted using a submerged slurry jet impacting a target spec­
achieving the independency of the predicted erosion patterns with imen. In the experiment a mixture of water-sand flow with sand volume
respect to the parcel numbers. fractions ranging from 0.38% to 5.66% were delivered by a nozzle. The
slurry jet produced by the nozzle was normal to the sample surface.
Fig. 1 shows a schematic of the experiment as well as the computational
3. Erosion model
domain, boundary conditions and the dimensions for the submerged
slurry jet test. The nozzle diameter and the distance between the nozzle
By using the described DDPM and DPM CFD models, the impinge­
and the specimen were 7 mm and 12.7 mm, respectively. Water was the
ment data such as the speed and the angle of impact of the sand particles
continuous phase and sharp sand particles with the average diameter of
are determined as the particles hit the walls. Using such information, the
300μm were used in the experiment.
erosion rate can be calculated according to different proposed erosion
The CFD model was also used to model the erosion in an elbow
models.
conveying gas-sand flow. The experiment reported by Ref. [19] was used
In the present study the models proposed by Ref. [19,58] are used to
to validate the CFD model. The experiments were performed with a
quantify the erosion for sand-air and sand-water flows, respectively. The
standard elbow (R/D = 1.5) in vertical-horizontal orientation and with
used models are the modified version of E/CRC erosion model and reads
as follows [6,12,59,60]:
ER = KFs Vpn f (θ) (18) Table 1
Constants used in the angle functions.
where ER is the erosion ratio, which is defined as the ratio of the mass Empirical constants Sand-gas model Sand-liquid model
loss rate at the wall to the particle mass flow rate. Vp is the velocity of the
n1 0.15 1.52
particles, n is a constant equal to 2.41, and K is a constant that depends
n2 0.85 8.9
on the material of the pipe which is defined as follows:
n3 0.65 0.01
K = C(BH)− 0.59
(19) HV (GPa) 1.83 1.83
f 1.53 18.74

4
H. Pouraria et al. Wear 460–461 (2020) 203450

first layer thickness of the grids for the experiment one and two were set
to 300 μm and 184 μm, respectively. A uniform velocity along with
turbulent intensity and length scale were specified at the inlets for both
the experiments. Pressure outlet boundary condition was applied at the
outlets. A no-slip boundary condition was imposed at the wall for the
continuous phase. Appropriate angle function and restitution functions
for stainless steel 316 were imposed at the walls for both the experi­
ments according to the described erosion model in section 3.

5. Results and discussions

5.1. Erosion in jet impingement test

The experimental work published by Ref. [41] was used to evaluate


the accuracy of the described numerical model in predicting the erosion
Fig. 1. A schematic of the geometry and boundary conditions (a) and the used
grid in CFD model(b) for the jet impingement tests.
of dense slurries. The liquid-sand erosion model proposed by Ref. [58]
was used to predict the erosion in slurry flow. The average flow velocity
in the experiment was 14 m/s and sharp sands with the average diam­
eter of 300 μm were used in the experiment. Hence, a sharpness factor of
1 was used in erosion modelling and a shape factor of 0.66 was adopted
in the non-spherical drag model. In the experiment water was used as
the carrier fluid and the sand volume fraction at the inlet was varied
from 0.38% to 5.66%. The duration of each test was 30 min.
Numerical modelling was first carried out for the sand volume
fraction of 0.38% for which the erosion profile was measured in the
experiment. Fig. 3 shows the predicted contours of erosion ratio in terms
of thickness loss per unit sand mass flow rate as predicted by one-way
coupled DPM and four-way coupled DDPM. As seen, the predicted
erosion profiles by both models are very similar. Fig. 4 illustrates a
comparison of the predicted radial distribution of erosion as obtained by
two CFD models and measured in the experiment. As seen, the predicted
maximum erosion rates by DDPM and one-way DPM agree well with the
measured value in the experiment. It is also observed that the location of
Fig. 2. A schematic of the geometry and boundary conditions (a) and the used the predicted maximum erosion rates slightly deviated from the
grid in CFD model(b) for the gas-sand flow in elbow. measured profile in the experiment.
At a constant flow velocity, the erosion rate changes proportionally
the internal diameter of 76.2 mm. Furthermore, air was used as a carrier with the particle volume fraction at the inlet, that is, an increase in the
fluid. Experiments were conducted for different particle shapes and sizes inlet particle volume fraction proportionally increases the mass flow
as well as different flow velocities. Fig. 2 shows the schematic of the used rate of the particles hitting the wall. However, experimental observa­
geometry as well as the grid. tions indicate that, at high particle loadings, by increasing the inlet
The geometries and the grids shown in Figs. 1 and 2 were generated particle volume fraction the erosion rate does not necessarily increase at
by using GAMBIT 2.4.6 [64]. A grid sensitivity test was carried out and the same rate as the inlet particle volume fraction. As a matter of fact, by
the grids consisting 303,040 and 746,240 hexahedral cells were found to increasing the particle volume fraction beyond a critical range the
be suitable for the experiment 1 and 2, respectively. Following (81) the erosion ratio which is defined as the erosion rate per unit sand mass flow
rate decreases [41,42,44,65,66,68,69].

Fig. 3. The erosion ratio distribution for the particle volume fraction of 0.38% in submerged slurry jet test as predicted by one-way coupled DPM and DDPM.

5
H. Pouraria et al. Wear 460–461 (2020) 203450

this study the reference Stokes number is 26.5.


In order to compare the performance of one-way DPM and DDPM at
different particle loading conditions, numerical modelling was also
performed for 0.038%, 1.88%, 3.78% and 5.66% sand volume fractions
while the flow velocity was kept constant. Fig. 5 shows the predicted
maximum erosion ratios at different sand volume fractions as obtained
by using one-way DPM and DDPM. As expected in this figure, the pre­
dicted erosion ratios by using one-way DPM is constant for all the par­
ticle volume fractions. Moreover, the predicted erosion profiles using
one-way DPM were identical to the erosion profile at 0.38% sand vol­
ume fraction shown in Fig. 3. However, the obtained results by using
DDPM in Fig. 5 indicate a decrease in erosion ratio when increasing the
sand volume fraction. The present results suggest that for volume frac­
tions below 1% the influence of particle volume fraction on the pre­
dicted erosion ratio is negligible, while further increase in particle
volume fraction results in a considerable erosion reduction. According to
Fig. 4. Radial distribution of erosion ratio for the particle volume fraction of this figure, by further increasing the volume fraction beyond a critical
0.38% in the submerged slurry jet test as predicted by one-way coupled DPM, value of around 2% the erosion reduction becomes significant (more
DDPM and measured in the experiment.
than 20%).
Fig. 6 shows the contours of erosion ratio at different sand volume
On contrary to the one-way DPM, DDPM takes into account the in­ fractions as obtained by DDPM.
fluence of inter-particle collisions as well as the influence of particles on In the experiment conducted by Ref. [41]; the integral erosion ratio
the continuous phase. Hence, it is expected to better predict the flow- defined as the total mass loss of the eroding sample per unit sand mass
field and the erosion rate at high particle loadings. Considering the flow rate was used to quantify the erosion ratio at different particle
fact that both CFD models employed the same erosion model, any dif­ volume fractions.
ferences between the predicted erosion profiles by using DDPM and one- The integral erosion ratio was measured for a specimen with the
way DPM could be attributed to the influence of particle concentration dimensions of 25mm × 37.5 mm in the experiment as well as the CFD
taken into account by DDPM. Hence, the similarity between the pre­ simulations. The measured integral erosion ratio for the particle volume
dicted erosion profiles and the maximum erosion ratios by using one- fraction of around 0.38% in the experiment was 6.04E-6 Kg/Kg. The
way DPM and DDPM as shown in Figs. 3 and 4 indicates a negligible predicted integral erosion ratio by using one-way DPM and DDPM are
influence of particle concentration on the erosion ratio at the low par­ 2.06E-6 Kg/Kg and 1.88E-6 Kg/Kg, respectively. Of note, Messa and
ticle volume fraction of 0.38%. It is worth mentioning that due to the Malvasi [34] used an axis-symmetric CFD model based on a mixed
high density and viscosity of the carrier fluid in slurry flows the influ­ Eulerian-Eulerian/Eulerian-Lagrangian approach and reported an inte­
ence of particle loading on the flow-field generally becomes important at gral erosion ratio of 2.63E-6 Kg/Kg for the same experiment.
comparatively higher particle volume fractions when compared to that Despite the reasonable agreement among the predicted and
in particulate gas flows. The tendency of particles to follow the flow measured erosion patterns in Fig. 3, the predicted integral erosion ratios
streamlines is generally described by the dimensionless Stokes number by CFD models in this study are considerably lower than that measured
(St) which is defined as the ratio of the particle relaxation time to the in the experiment. A study conducted by Ref. [70] indicates the occur­
fluid-flow relaxation time. For a given flow-field the reference Stokes rence of abrasion erosion in regions far from the center of the sample.
number reads as follows: However, the employed erosion model in the present CFD model does
ρp dp2 Vf not take into account the influence of abrasion erosion. Furthermore, in
St = (24) profilometer, the volume loss is measured from a reference surface that
18μf D
is being eroded by abrasion. Hence, the influence of abrasion erosion in
regions far from the center of the sample is overlooked in both the CFD
Where, ρp and dp denote the particle density and diameter. Vf , μf and D
simulations and the measurement using profilometer. As a result, the
indicate the fluid-flow velocity, dynamic viscosity and the flow length
predicted and measured integral erosion ratios by using the CFD simu­
scale (pipe diameter), respectively. For the submerged slurry jet test in
lation and the profilometer are less than that measured by using the
direct mass balance test.
Fig. 7 shows the measured and the predicted relative integral erosion
ratios obtained by normalizing the integral erosion ratios with respect to
their value at 0.38% sand volume fraction. This figure also illustrates the
obtained results by using Eulerian-Granular model as reported by
Ref. [41]. It should be noted that all the CFD results conducted using
one-way DPM, DDPM and the Eulerian-Granular approaches were ob­
tained by using the erosion model proposed by Ref. [58]. Hence, the
observed differences among the predicted results could be attributed to
the performance of different CFD methods in modelling the particulate
flow-field. As seen in this figure, the experimental data indicate a sig­
nificant reduction of integral erosion ratio when the sand volume frac­
tion increases, while the one-way DPM results suggest a constant value.
Furthermore, the obtained results by DDPM indicate a considerable
reduction of integral erosion ratio that reasonably agree with the
experimental data. From this figure, it is also observed that the
Fig. 5. Maximum erosion ratio at different particle volume fractions for the Eulerian-Granular model predicted a noticeable reduction of integral
submerged slurry jet test as predicted by one-way coupled DPM and erosion ratio which is significantly low when compared to the experi­
DDPM models. mental data. Furthermore, the CFD results using the mixed

6
H. Pouraria et al. Wear 460–461 (2020) 203450

Fig. 6. Erosion ratio distribution at different particle volume fractions, vf, for the submerged slurry jet test as predicted by DDPM.

Fig. 8 shows the radial distribution of the impact velocity of the


particles for different particle loading conditions as predicted by DDPM.
According to this figure, as the particle volume fraction at the inlet of the
nozzle increases the impact velocity of the particles hitting the sample
decreases. The present results suggest that by increasing the particle
loading and, thereby the influence of inter-particle collisions the energy
of attacking particles decreases. These results agree with the numerical
study reported by Ref. [33]. Fig. 9 illustrates the radial distribution of
parcel number density (1/m3) at the sample for different particle
loading conditions. The distribution of parcel number density at regions
close to the center indicates that by increasing the particle loading the
parcel number hitting the sample increases, while at the location of
maximum impact velocity (7–10 mm), the parcel number density de­
creases with increasing particle concentration at the inlet.
In order to investigate the influence of reference Stokes number on
the erosion ratio at different particle volume fractions, numerical
modelling was carried out for a particulate gas jet impingement test in
Fig. 7. Relative integral erosion ratio at different particle volume fractions in which the geometry, flow velocity as well as the particle properties were
submerged slurry jet test as measured in the experiment and predicted by one- identical to those in the submerged slurry jet test used in this study. In
way coupled DPM, Eulerian-Granular, DDPM and mixed EE-EL models. fact, due to the significantly lower viscosity of the gas compared to that
of the water in the slurry flow the reference Stokes number of the flow in
Eulerian-Eulerian/Eulerian-Lagrangian (EE-EL) approach reported by the particulate gas jet impingement becomes as high as 1480.
Messa and Malvasi [34] is depicted in Fig. 7. This CFD model considers Fig. 10 shows the erosion ratios at different particle volume fractions
the interplay among the testing time and the particle volume fractions in ranging from 3.8E-7 to 0.038. According to this figure, for particle
which the influence of the evolution of the surface shape due to the volume fractions below 1E-5, both DPM and DDPM predict similar
erosion is taken into account. As observed in this figure, the mixed EE-EL erosion ratios, while at higher volume fractions DDPM suggest a
approach also clearly shows the erosion reduction at high particle decrease in erosion ratio when compared to the DPM results. The ob­
loading conditions, thereby, suggesting the importance of surface shape tained results by using DDPM indicate that for the particle volume
evolution on the erosion reduction. fractions beyond a critical value of around 1E-4 the erosion reduction
The observed difference between the predicted results using DDPM becomes significant (more than 20%). Furthermore, for the higher
and the measured relative integral erosion ratio could be attributed to particle volume fractions beyond 5E-3, the erosion reduction of around
several factors such as the assumption of spherical particles in KTGF two order of magnitude is observed. In fact, due to the significantly
model or neglecting the surface roughness and the evolution of surface higher reference Stokes number of the particulate jet impingement test
shape due to the erosion during the testing time. compared to that of the submerged slurry jet test, the erosion reduction

Fig. 8. Radial distribution of particle impact velocity at different input particle Fig. 9. Radial distribution of particle number density at different input particle
volume fractions (vf) as predicted by DDPM. volume fractions (vf) as predicted by DDPM.

7
H. Pouraria et al. Wear 460–461 (2020) 203450

that in the peripheral region. While in the slurry jet, the stagnation zone
near the center reduces the erosion in that region, in high concentration
particulate gas jet, inter-particle collisions prevent severe impacts near
the center of the jet. Furthermore, at the higher particle volume fractions
the intense inter-particle collisions results in a higher degree of particle
dispersion, thereby increasing the eroded area.
The published works regarding the influence of particle loading on
the erosion ratio shows conflicting opinions. Generally, for steels and
ceramics an increase in particle loading beyond a certain range results in
erosion reduction, while for rubbers an increase in erosion ratio is
observed [71]. Furthermore, previous experiments reveal the impor­
tance of testing time on the erosion ratio. Recently [34], could model the
influence of testing time by simultaneously updating the geometry of
eroding specimen during the simulation. Their results revealed a sig­
nificant erosion reduction as the testing time increased. Furthermore,
several studies indicate that the influence of particle loading on erosion
ratio depends on the impact velocity, the particle type, the angle of
Fig. 10. Maximum erosion ratio at different particle volume fractions for the impact as well as the particle size [71,72]. The influence of particle
particulate gas jet impingement test as predicted by one-way coupled DPM
loading for different particle type was related to the crushing of the
and DDPM.
particles that depends on the material properties. The influence of ve­
locity was attributed to the higher probability of the impact of attacking
is significantly higher. particles with those rebounding from the surface. This, in turn, results in
Fig. 11 shows the erosion ratio distribution at different particle decreasing the energy of attacking particles and changing their direction
volume fractions as obtained by one-way DPM and DDPM. The erosion [71]. Such interpretation agrees with the observed erosion distribution
ratio distribution obtained by one-way DPM are identical for all the in Fig. 8. However, a quantitative assessment of the obtained results by
particle loading conditions. Hence, only one of them is shown in this DDPM needs further investigation and validation against experimental
figure. From these figures it is seen that the predicted erosion profiles by data.
DDPM at low particle loading conditions are very similar to the pre­
dicted results by one-way DPM. On contrary to the low Stokes number 5.2. Erosion in elbow
flow in slurry jet test, particles in high Stokes number flow of gas jet
impingement do not follow the flow-streamlines. In fact, due to the high The experimental work published by Ref. [19] was used to evaluate
relaxation time of the particles compared to that for the fluid, the par­ the performance of the described numerical model in predicting the
ticles do not have enough time to adjust their path with the sudden erosion ratio in an elbow conveying two-phase gas-sand flow. Two tests
changes in flow path-lines. Hence, instead of following the fluid-flow with different fluid velocities, particle sizes, particle shapes and particle
and turning their way around the specimen surface, particles directly loadings were used to assess the accuracy of the numerical model.
hit the specimen so that the maximum erosion appears at the center of Table 2 shows the important parameters regarding the operating con­
the specimen. Moreover, unlike the slurry flow in which the rebounded ditions of the two tests as well as the corresponding erosion rates as
particles are swept away from the impact region, in the particulate gas measured in the experiments and predicted by the CFD models.
jet particles rebound back toward the incoming particles. Hence, by Fig. 12 shows the predicted erosion patterns for the test 1 using the
increasing the particle volume fraction and subsequently the role of DDPM and one-way coupled DPM models. The flow velocity and particle
inter-particle collisions the erosion profile gradually changes. As size are 11 m/s and 150 μm, respectively. Semi-round particles were
observed in Fig. 11 the erosion distribution at high particle volume used in test 1. Hence, the sharpness factor was set to 0.53 in the CFD
fractions is similar to that for the submerged slurry jet test so that the model. As seen in this figure, a so-called V-shape erosion pattern is
erosion ratio at the center of the specimen is significantly lower than predicted by both the DDPM and DPM approaches. The same erosion

Fig. 11. Erosion distribution at different particle volume fractions, vf, for the particulate gas jet impingement test as predicted by one-way coupled DPM and DDPM.

8
H. Pouraria et al. Wear 460–461 (2020) 203450

Table 2
Operating condition of the experiment [19]].
Experimental condition Test 1 Test 2

Pipe diameter 0.0762 m 0.0762 m


Fluid density 1.2 Kg/m3 1.2 Kg/m3
Fluid viscosity 0.000017 Pa s 0.000017 Pa s
Particle diameter 150 μm 300 μm
Particle density 2650 Kg/m3 2650 Kg/m3
Flow velocity 11 m/s 15 m/s
Particle volume fraction 2.2E-5 8.83E-6
Particle shape Semi-round Sharp
Measured erosion rate 6.5 mm/year 22.1 mm/year
Predicted erosion rate by 1-way DPM 6.22 mm/year 27 mm/year
Predicted erosion rate by 4-way DDPM 6.18 mm/year 25.89 mm/year

Fig. 13. The predicted erosion pattern as obtained by CFD models for test 2 (a)
DDPM, (b) one-way coupled DPM.

and the measured erosion rates. As seen in Fig. 13, both models pre­
dicted a V-shape erosion pattern. It is worth mentioning that in the
present CFD simulations the influence of surface roughness has been
neglected. Recent studies on the influence of surface roughness indicate
that by increasing the roughness the particle dispersion increases which
in turn prevents the formation of V-shape erosion profiles observed in
Figs. 12 and 13 [13,33]. A detailed discussion regarding the influence of
surface roughness on the particle impact angle and velocity can be found
in Ref. [33].
As mentioned before, the main advantage of DDPM compared to the
one-way coupled DPM approach relies on its capability in modelling
Fig. 12. The predicted erosion patterns as obtained by CFD models for test 1 (a) dense particulate flows where the effect of inter-particle collisions and
DDPM, (b) one-way coupled DPM. the influence of particulate phase on the continuous phase become
important. In order to investigate the influence of particle concentration
pattern was also reported by Ref. [19] according to the experiment as on the erosion pattern, numerical modelling was conducted for test 1
well as using one-way coupled DPM simulation. According to Table 2, and test 2 while the input volume fractions varied from 1E-6 to 1E-3. In
the measured erosion rate in the experiment is 6.5 mm/year. The pre­ order to make a reliable comparison of the erosion patterns at different
dicted erosion rates using DDPM and one-way coupled DPM are 6.18 particle concentrations, all the parameters except the particle mass flow
mm/year and 6.22 mm/year, respectively. A comparison among the rate were kept constant. The particle mass flow rate at the inlet was
predicted erosion rates by the CFD and the experimental measurement adjusted at a constant inlet fluid velocity of 11 m/s and 15 m/s to
verifies the accuracy of the present DDPM and one-way coupled DPM achieve the desired particle volume fractions at the inlet.
models. The particle volume fraction for test 1 is around 2.2E-5. A Fig. 14 illustrates the predicted erosion ratios for test 1 at different
comparison of the predicted erosion rates by using one-way DPM and particle volume fractions as obtained by the DDPM and one-way coupled
DDPM indicates that the influence of inter-particle collisions on the DPM approaches. The flow velocity and the particle size are 11 m/s and
erosion can be neglected at this particle loading. 150 μm, respectively. As seen in this figure, the DDPM approach pre­
Fig. 13 shows the predicted erosion patterns for test 2. The flow dicted a reduction of erosion ratio. However, as expected, the one-way
velocity and sand particle sizes are 15 m/s and 300 μm, respectively. The coupled DPM model shows a constant erosion ratio for all the volume
California 60 sand was used as abrasive particle. SEM image of the sand fractions. According to the obtained results by using DDPM, as the
particles showed very sharp edges [19]. Hence, a sharpness factor (Fs ) of particle volume fraction at the inlet increases the penetration per unit
1 was used in the CFD model. The sand mass flow rate at the inlet in­ inlet sand mass flow rate decreases. From this figure, it is seen that for
dicates a very dilute flow with an average volume fraction of 8.83E-6. volume fractions less than 1E-4, the influence of particle loading is not
The measured erosion rates in the experiment as well as the pre­ significant. However, for particle concentrations beyond 1E-4, a
dicted values using DDPM and one-way coupled DPM are shown in comparatively significant reduction in erosion ratio is observed.
Table 2. Again, a reasonable agreement is observed among the predicted Fig. 15 shows the influence of particle volume fraction for the test 2

9
H. Pouraria et al. Wear 460–461 (2020) 203450

Fig. 14. The influence of particle volume fraction on the erosion ratio for test 1
(velocity = 11 m/s, particle size = 150 μm, semi-rounded particle).

Fig. 15. The influence of particle volume fraction on the erosion ratio for test 2
(velocity = 15 m/s, particle size = 300 μm, sharp particle).

in which the flow velocity and the particle size are 15 m/s and 300 μm,
respectively. As seen in this figure, the predicted erosion ratio using
DDPM approach clearly shows the erosion reduction phenomenon at the
high particle loading conditions. According to Figs. 14 and 15, as the
particle loading increases the erosion reduction becomes more signifi­
cant. However, the rate of reductions in these figures are different. The
observed differences might be attributed to the different operating
conditions such as the flow velocity and the particle size. Previous
experimental observations also indicate that the erosion reduction de­
pends on several factors such as particle size, particle density, velocity
and the property of the carrier fluid [44]. As mentioned earlier, for a
certain geometrical configuration, the dimensionless reference Stokes
number is often used to describe the particle-fluid interactions. The
reference Stokes number for the test 1 and test 2 conducted in the elbow
are 26.7 and 145.7, respectively. Hence, the higher erosion reduction
observed in the test 2 could be attributed to the higher reference Stokes
number in this test when compared to that in the test 1.
It is also interesting to compare the obtained results of the test 1 in
the elbow with those obtained in the submerged slurry jet test. Although
the reference Stokes number in these flows are almost identical the
observed erosion reduction trends are different. A comparison of Figs. 5
and 14 indicates that the erosion reduction phenomenon in the elbow Fig. 16. The influence of particle volume fraction on the erosion pattern as
emerges at the lower particle loadings when compared to that in the obtained by DDPM for test 1.
submerged slurry jet test. Such difference reveals the importance of
other geometrical and flow parameters.
Fig. 16 a-d show the predicted erosion patterns using DDPM for
different particle concentrations as obtained for test 1. According to

10
H. Pouraria et al. Wear 460–461 (2020) 203450

these figures, the erosion patterns for all the volume fractions are very mentioning that the particle volume fraction in a particulate system
similar. A V-shape erosion pattern is observed in all of them, while the cannot exceed the packing limit which is 0.63 for a particulate system
maximum erosion ratio is observed at the centreline of the extrados of with uniform particle size. However, the obtained results by the one-way
the elbow. From these figures, it is seen that an increase in the inlet DPM in Fig. 17 suggest that a further increase of the inlet particle vol­
volume fraction from 1E-6 to 1E-5 does not influence the observed ume fraction to 5E-3 would result in a maximum particle volume frac­
erosion pattern and the erosion ratio. Moreover, further increase of the tion of 0.855 in the elbow section which is unrealistic. By employing a
particle volume fraction from 1E-5 To 1E-4 results in a decrease in four-way coupling approach in DDPM, however, the particle volume
erosion ratio while the erosion pattern is not affected. However, by fraction cannot exceed the packing limit. At higher particle volume
further increase of particle volume fraction from 1E-4 to 1E-3 a fractions the inter-particle collisions become important, which in turn,
comparatively more significant reduction in erosion ratio is observed. In result in a higher degree of dispersion. The different predicted erosion
addition, comparing Fig. 16-c and Fig. 16-d, it is observed that the pattern at the input volume fraction of 1E-3 might be attributed to the
erosion pattern at the higher volume fractions slightly changes and the higher particle dispersion and thus a less increase of particle volume
erosion gradient at the maximum erosion point decreases. fraction at the wall when compared to the predictions by using one-way
As mentioned, the reduction of erosion ratio at higher particle vol­ coupled DPM. Hence, a comparatively more uniform erosion distribu­
ume fractions was already observed in the experiments [41,44,58,69]. tion as well as a reduction in erosion ratio is observed at the higher
This phenomenon has been related to the so-called shielding effect. It is volume fraction of 1E-3 (see Fig. 16-d).
believed that at higher particle concentrations a shield of concentrated For flow-fields with highly concentrated particle regions the influ­
particle region is formed near the wall, which in turn, partially protects ence of particles on the hydrodynamics of the continuous phase cannot
the wall from erosion [33,44]. The numerical study conducted by be neglected. Fig. 19 and Fig. 20 show the velocity contours of the
Ref. [33] could capture this phenomenon by employing a four-way continuous phase for the test 1 at different input particle volume frac­
coupled DPM model. Their numerical results suggested that by tions as obtained by one-way coupled DPM and DDPM, respectively. As
increasing the particle loading the intensity of inter-particle collisions as expected, the obtained results by using one-way coupled DPM indicated
well as the particle-wall collision frequency increase while the energy of that the velocity field for all the runs remained identical regardless of the
the incoming particles decreases. The present CFD results using DDPM changes in the input particle volume fraction. In fact, the velocity con­
could also predict the erosion reduction at high volume fractions. tour depicted in Fig. 19 is observed for all the operating conditions with
However, due to the lack of experimental data for the standard elbow, a the input volume fractions ranging from 1E-6 to 1E-3. A comparison of
quantitative assessment of the erosion reduction cannot be made. Figs. 19 and 20 reveals that the velocity field predicted by one-way
Fig. 17 and Fig. 18 show the contours of particle volume fraction at coupled DPM is similar to those obtained by DDPM at low particle
the extrados of the elbow for different input particle volume fractions as loading conditions.
obtained by one-way coupled DPM and DDPM, respectively. A com­ When the gas phase enters the elbow section, the dominant centrif­
parison among these figures and Fig. 16 indicates that the location of the ugal force increases the velocity near the intrados of the elbow while the
maximum particle volume fraction and maximum erosion are identical. velocity in the extrados region of the elbow decreases. However, at the
As seen in Fig. 17, the predicted distribution of volume fraction using exit of the elbow (the entrance of the horizontal pipe), a significant
one-way coupled DPM indicates that the maximum particle volume reduction of velocity is observed near the inner edge. On contrary to the
fraction in the elbow section is two order of magnitude higher than that elbow section, the maximum velocity at the outlet of the elbow is
at the inlet. Moreover, the overall distribution is similar for all the runs observed near the outer edge. The obtained velocity profiles agree well
regardless of the different input volume fractions. In fact, the obtained with the previous studies on the singe phase flows in the bends [74–77].
results by using one-way coupled DPM suggest that by increasing the However, as the particle loading increases, the predicted velocity con­
input particle volume fraction the maximum particle volume fraction tours by DDPM starts to deviate from those predicted by DDPM at low
increases with the same rate as the input volume fraction. It is also particle loadings. The velocity distributions as obtained by DDPM sug­
observed that the particle volume fraction in elbow section, for input gest that an increase in particle loading results in a reduction of the
volume fraction of 1e-4 and 1e-3, is greater than 0.01. It is worth velocity in the extrados of the elbow. As a result, for very high particle
mentioning that for particle volume fractions higher than 0.01, loading conditions, the velocity profile at the exit of the elbow signifi­
neglecting the inter-particle collisions is not a reasonable assumption cantly differs from that at the low particle loading conditions. The
[73]. Hence, the obtained results using one-way coupled DPM may not DDPM results indicate that, for high particle loading conditions, the
be reliable. velocity of fluid at the outer edge of the elbow in the exit section is
As seen in Fig. 18, the predicted distribution of particle volume significantly lower than that at the inner edge.
fraction using four-way coupled DDPM is different. According to these The significant changes observed in the velocity profile of the fluid
figures, as the particle volume fraction at the inlet increases the by using DDPM is attributed to the volume fraction of the particles in the
maximum particle volume fraction in the elbow section also increases. elbow section. As seen in Figs. 17 and 18, at high particle loadings the
However, on contrary to the obtained results using one-way coupled volume fraction of the particles in the outer region of the elbow signif­
DPM, the maximum particle volume fraction in elbow does not increase icantly increases, which in turn, increases the drag force between two
with the same factor as the inlet particle volume fraction. It is worth phases. As a matter of fact, the high particle concentration region

Fig. 17. Particle volume fraction distribution at different input volume fractions as predicted by one-way coupled DPM for test 1.

11
H. Pouraria et al. Wear 460–461 (2020) 203450

Fig. 18. Particle volume fraction distribution at different input volume fractions as predicted by DDPM for test 1.

by further improving the closure laws the obtained results may improve.
It is worth mentioning that employing different homogenous turbulence
models such as the one used in this study is a common practice in most of
the published works using DEM and two-fluid model. However, the
particulate flows are inherently inhomogeneous.

6. Conclusions

Numerical modelling of erosion was carried out by using dense


discrete particle method (DDPM) and one-way coupled discrete particle
method (DPM). Different experiments consist of a submerged slurry jet
impingement test and an elbow conveying gas-sand flow was used to
evaluate the performance of the CFD models.
CFD simulations of the submerged slurry jet test indicated that both
one-way DPM and DDPM predicted similar erosion profiles at low par­
ticle loading condition which reasonably agree with the experimental
measurement.
Numerical modelling of erosion ratio at high particle loading con­
ditions in the slurry flow revealed a significant difference among the
predicted results by using one-way DPM and DDPM. The DDPM results
suggest that there exist a critical particle loading beyond which the
erosion reduction becomes important. Apart from that, it was observed
that the erosion reduction rate predicted by DDPM reasonably agreed
with the experimental data. Furthermore, a comparison of the predicted
Fig. 19. Fluid velocity contours for different input volume fractions as pre­
results by DDPM with those obtained by Eulerian-Granular model
dicted by one-way coupled DPM.
revealed the higher accuracy of DDPM in predicting the influence of
particle loading on the erosion ratio.
partially blocks the path of the fluid flow, thereby increasing the fluid
According to the DDPM predictions and the experiment, for the
velocity near the inner edge of the elbow. By considering a four-way
slurry flow with the low Stokes number (St = 26.5) considered in this
coupling in DDPM approach a significant influence of the particulate
study the erosion reduction becomes significant after a critical particle
phase on the carrier fluid is revealed while in one-way coupled DPM
loading of around 2% by volume. Furthermore, the obtained results by
such features are overlooked. The four-way coupling approach
DDPM suggest that by increasing the particle loading the impact velocity
employed in DDPM simultaneously depicts the influence of inter-
of the particle decreases.
particle collisions and the two-way coupling between the fluid and
The obtained DDPM results for the particulate gas jet impingement
particles. Hence, the contribution of these two cannot be distinguished
test with a comparatively high Stokes number (St = 1480), however,
well. A study conducted by Ref. [33] reports on the influence of
indicates that the onset of erosion reduction appears at particle loadings
considering both two-way and four-way couplings as the particle
as low as 0.01%. Furthermore, because of more inter-particle collisions
loading increases.
in the high Stokes flow of particulate gas jet impingement, the predicted
The numerical results obtained by DDPM approach depends on the
erosion reduction was significantly higher than that in the low Stokes
adopted closure laws such as those in KTGF model, the turbulence model
flow of submerged slurry jet test.
and the interfacial forces. Similar to the other approaches such as DEM
Numerical modelling of erosion for particulate gas flow in an elbow
(hard-sphere or soft-sphere models) or the Eulerian-Eulerian approach,

Fig. 20. Fluid velocity contours for different input volume fractions as predicted by DDPM.

12
H. Pouraria et al. Wear 460–461 (2020) 203450

was conducted for two different Stokes numbers of 26.7 and 145.7. [14] C.Y. Wong, C. Solnordal, A. Swallow, S. Wang, L. Graham, J. Wu, Predicting the
material loss around a hole due to sand erosion, Wear 276–277 (2012) 1–15.
At low particle loading conditions both one-way DPM and DDPM
[15] F. Darihaki, E. Hajidavalloo, A. Ghasemzadeh, G.A. Safian, Erosion prediction for
give comparatively similar results that reasonably match the experi­ slurry flow in choke geometry, Wear 372 (2017) 42–53.
mental data. However, by increasing the particle volume fraction the [16] G. Hu, P. Zhang, G. Wang, H. Zhu, Q. Li, S. Zhao, T. Wang, Performance study of
DDPM shows a reduction in erosion ratio which is more significant for erosion resistance on throttle valve of managed pressure drilling, J. Petrol. Sci.
Eng. 156 (2017) 29–40.
the flow with the higher Stokes number. [17] G.V. Messa, S. Malavasi, The effect of sub-models and parameterizations in the
Both one-way DPM and DDPM models indicate that for input volume simulation of abrasive jet impingement tests, Wear 370 (2017) 59–72.
fractions higher than1E-4, the maximum particle volume fraction in the [18] W. Peng, X. Cao, Numerical prediction of erosion distributions and solid particle
trajectories in elbows for gas–solid flow, J. Nat. Gas Sci. Eng. 30 (2016) 455–470.
elbow section exceeds the critical value of 1E-2 beyond which the inter- [19] R.E. Vieira, A. Mansouri, B.S. McLaury, S.A. Shirazi, Experimental and
particle collisions and the influence of particulate phase on the fluid computational study of erosion in elbows due to sand particles in air flow, Powder
becomes important. Technol. 288 (2016) 339–353.
[20] M. Zamani, S. Seddighi, H.R. Nazif, Erosion of natural gas elbows due to rotating
The obtained results using DDPM indicate a lower increase of the particles in turbulent gas-solid flow, J. Nat. Gas Sci. Eng. 40 (2017) 91–113.
particle volume fraction at the hot spot of the elbow when compared to [21] S. Karimi, S.A. Shirazi, B.S. McLaury, Predicting fine particle erosion utilizing
those obtained by one-way DPM. computational fluid dynamics, Wear 376 (2017) 1130–1137.
[22] M. Parsi, M. Agrawal, V. Srinivasan, R.E. Vieira, C.F. Torres, B.S. McLaury, S.
A comparison of the predicted velocity contours by using DDPM and A. Shirazi, CFD simulation of sand particle erosion in gas-dominant multiphase
one-way DPM also shows the importance of considering a two-way flow, J. Nat. Gas Sci. Eng. 27 (2015) 706–718.
coupling between the continuous and the dispersed phase for high [23] H. Pouraria, J.K. Seo, J.K. Paik, Numerical study of erosion in critical components
of subsea pipeline: tees vs bends, Ships Offshore Struct. 12 (2) (2017) 233–243.
particle loading conditions. The numerical results obtained by DDPM
[24] J. Wang, S.A. Shirazi, A CFD based correlation for erosion factor for long-radius
revealed remarkable velocity changes at high particle volume fractions elbows and bends, ASME Journal of Energy Resource Technology 125 (2003)
while the one-way DPM solution suggests a flow-field similar to a single- 26–34.
phase gas flow. [25] E.S. Firoozabad, B.G. Jeon, H.S. Choi, N.S. Kim, Failure criterion for steel pipe
elbows under cyclic loading, Eng. Fail. Anal. 66 (2016) 515–525.
[26] G.H. Lee, H. Pouraria, J.K. Seo, J.K. Paik, Burst strength behaviour of an aging
subsea gas pipeline elbow in different external and internal corrosion-damaged
Declaration of competing interest positions, International Journal of Naval Architecture and Ocean Engineering 7 (3)
(2015) 435–451.
The authors declare that they have no known competing financial [27] M. Mokhtari, R.E. Melchers, A new approach to assess the remaining strength of
corroded steel pipes, Eng. Fail. Anal. 93 (2018) 144–156.
interests or personal relationships that could have appeared to influence
[28] Q. Wang, W. Zhou, Burst pressure models for thin-walled pipe elbows, Int. J. Mech.
the work reported in this paper. Sci. 159 (2019) 20–29.
[29] J. Chen, Y. Wang, X. Li, R. He, S. Han, Y. Chen, Erosion prediction of liquid-particle
two-phase flow in pipeline elbows via CFD–DEM coupling method, Powder
Acknowledgements Technol. 275 (2015) 182–187.
[30] N. Climent, M. Arroyo, C. O’Sullivan, A. Gens, Sand production simulation
This work was supported by the National Research Foundation of coupling DEM with CFD, European Journal of Environmental and Civil Engineering
18 (9) (2014) 983–1008.
Korea (NRF) grantfunded by the Korea government (Ministry of Science
[31] L. Xu, Q. Zhang, J. Zheng, Y. Zhao, Numerical prediction of erosion in elbow based
and ICT) (No. 2019R1A2C1006786). on CFD-DEM simulation, Powder Technol. 302 (2016) 236–246.
[32] C.A.R. Duarte, F.J. de Souza, V.F. dos Santos, Numerical investigation of mass
loading effects on elbow erosion, Powder Technol. 283 (2015) 593–606.
Appendix A. Supplementary data [33] C.A.R. Duarte, F.J. de Souza, R. de Vasconcelos Salvo, V.F. dos Santos, The role of
inter-particle collisions on elbow erosion, Int. J. Multiphas. Flow 89 (2017) 1–22.
Supplementary data to this article can be found online at https://doi. [34] G.V. Messa, S. Malavasi, A CFD-based method for slurry erosion prediction, Wear
398 (2018) 127–145.
org/10.1016/j.wear.2020.203450.
[35] N. Ellis, M. Xu, C.J. Lim, S. Cloete, S. Amini, Effect of change in fluidizing gas on
riser hydrodynamics and evaluation of scaling laws, Ind. Eng. Chem. Res. 50 (8)
References (2011) 4697–4706.
[36] B. Pang, S. Wang, Q. Wang, K. Yang, H. Lu, M. Hassan, X. Jiang, Numerical
prediction of cuttings transport behavior in well drilling using kinetic theory of
[1] R. Dabirian, R.S. Mohan, O. Shoham, G. Kouba, Sand transport in stratified flow in
granular flow, J. Petrol. Sci. Eng. 161 (2018) 190–203.
a horizontal pipeline, in: SPE Annual Technical Conference and Exhibition, Society
[37] F. Taghipour, N. Ellis, C. Wong, Experimental and computational study of gas–solid
of Petroleum Engineers, 2015.
fluidized bed hydrodynamics, Chem. Eng. Sci. 60 (24) (2005) 6857–6867.
[2] M. Parsi, K. Najmi, F. Najafifard, S. Hassani, B.S. McLaury, S.A. Shirazi,
[38] R. Tebowei, M. Hossain, S.Z. Islam, M.G. Droubi, G. Oluyemi, Investigation of sand
A comprehensive review of solid particle erosion modeling for oil and gas wells and
transport in an undulated pipe using computational fluid dynamics, J. Petrol. Sci.
pipeline applications, J. Nat. Gas Sci. Eng. 21 (2014) 850–873.
Eng. 161 (2018) 747–762.
[3] I. Finnie, Erosion of surfaces by solid particles, Wear 3 (2) (1960) 87–103.
[39] G.J. Brown, Erosion prediction in slurry pipeline tee-junctions, Appl. Math. Model.
[4] Y.I. Oka, K. Okamura, T. Yoshida, Practical estimation of erosion damage caused by
26 (2) (2002) 155–170.
solid particle impact, Wear 259 (1–6) (2005) 95–101.
[40] Y. Lu, M. Agrawal, A computational-fluid-dynamics-based Eulerian-granular
[5] H. Arabnejad, A. Mansouri, S.A. Shirazi, B.S. McLaury, Development of
approach for characterization of sand erosion in multiphase-flow systems, SPE J.
mechanistic erosion equation for solid particles, Wear 332 (2015) 1044–1050.
19 (4) (2014) 586–597.
[6] Y. Zhang, Application and Improvement of Computational Fluid Dynamics (CFD) in
[41] M. Mahdavi, S. Karimi, S.A. Shirazi, B.S. McLaury, Parametric study of erosion
Solid Particle Erosion Modelling, (Ph.D. Dissertation) Department of Mechanical
under high concentrated slurry: experimental and numerical analyses, in: ASME
Engineering, The University of Tulsa, Tulsa, Oklahoma, USA, 2006, 2006.
2016 Fluids Engineering Division Summer Meeting Collocated with the ASME 2016
[7] N.A. Barton, Erosion in Elbows in Hydrocarbon Production System: A Review
Heat Transfer Summer, 2016.
Document, Research report Scottish Enterprise Technology Park, East Kilbrede,
[42] A. Mansouri, M. Mahdavi, S.A. Shirazi, B.S. McLaury, Investigating the effect of
2003. Research Report 115.
sand concentration on erosion rate in slurry flows, Paper 6130 (2015) 15–19.
[8] C.A.R. Duarte, F.J. de Souza, Innovative pipe wall design to mitigate elbow erosion:
[43] G.V. Messa, G. Ferrarese, S. Malavasi, A mixed Euler–Euler/Euler–Lagrange
a CFD analysis, Wear 380 (2017) 176–190.
approach to erosion prediction, Wear 342 (2015) 138–153.
[9] C.A.R. Duarte, F.J. de Souza, V.F. dos Santos, Mitigating elbow erosion with a
[44] R. Macchini, M.S.A. Bradley, T. Deng, Influence of particle size, density, particle
vortex chamber, Powder Technol. 288 (2016) 6–25.
concentration on bend erosive wear in pneumatic conveyors, Wear 303 (1–2)
[10] X. Chen, B.S. McLaury, S.A. Shirazi, Application and experimental validation of a
(2013) 21–29.
computational fluid dynamics (CFD)-based erosion prediction model in elbows and
[45] X. Chen, J. Wang, A comparison of two-fluid model, dense discrete particle model
plugged tees, Journal of Computers and Fluids 33 (2004) 1251–1272.
and CFD-DEM method for modeling impinging gas–solid flows, Powder Technol.
[11] X. Chen, B.S. McLaury, S.A. Shirazi, Numerical and experimental investigation of
254 (2014) 94–102.
the relative erosion severity between plugged tees and elbows in dilute gas/solid
[46] S. Cloete, S.T. Johansen, S. Amini, Performance evaluation of a complete
two-phase flow, Wear 261 (2006) 715–729.
Lagrangian KTGF approach for dilute granular flow modelling, Powder Technol.
[12] O.D. Njobuenwu, M. Fairweather, Modelling of pipe bend erosion by dilute particle
226 (2012) 43–52.
suspensions, Comput. Chem. Eng. 42 (2012) 235–247.
[47] A. Klimanek, W. Adamczyk, A. Katelbach-Woźniak, G. Węcel, A. Szlęk, Towards a
[13] C.B. Solnordal, Y.W. Chong, J. Boulanger, An experimental and numerical analysis
hybrid Eulerian–Lagrangian CFD modeling of coal gasification in a circulating
of erosion caused by sand pneumatically conveyed through a standard pipe elbow,
fluidized bed reactor, Fuel 152 (2015) 131–137.
Wear 336 (2015) 43–57.

13
H. Pouraria et al. Wear 460–461 (2020) 203450

[48] C.A.R. Duarte, F.J. de Souza, D.N. Venturi, M. Sommerfeld, A numerical assessment [65] T. Deng, A.R. Chaudhry, M. Patel, I. Hutchings, M.S.A. Bradley, Effect of particle
of two geometries for reducing elbow erosion, Particuology 49 (2020) 117–133. concentration on erosion rate of mild steel bends in a pneumatic conveyor, Wear
[49] D. Gidaspow, Multiphase Flow and Fluidization: Continuum and Kinetic Theory 258 (2005) 480–487.
Descriptions, Academic Press, 1994. [66] T. Frosell, M. Fripp, E. Gutmark, Investigation of slurry concentration effects on
[50] C.Y. Wen, Y.H. Yu, Mechanics of fluidization, Chem. Eng. Prog. Symp. (1966) solid particle erosion rate for an impinging jet, Wear 342 (2015) 33–43.
100–113. [68] D. Mills, J.S. Mason, Particle concentration effects in bend erosion, Powder
[51] S. Ergun, Fluid flow through packed columns, Chem. Eng. Prog. 48 (1952) 89–94. Technol. 17 (1) (1977) 37–53.
[52] A. Haider, O. Levenspiel, Drag coefficient and terminal velocity of spherical and [69] S. Turenne, M. Fiset, J. Masounave, The effect of sand concentration on the erosion
nonspherical particles, Powder Technol. 58 (1) (1989) 63–70. of materials by a slurry jet, Wear 133 (1) (1989) 95–106.
[53] C.K.K. Lun, S.B. Savage, D.J. Jeffrey, N. Chepurniy, Kinetic theories for granular [70] H. Arabnejad, A. Mansouri, S.A. Shirazi, B.S. McLaury, Abrasion erosion modeling
flow: inelastic particles in Couette flow and slightly inelastic particles in a general in particulate flow, Wear 376 (2017) 1194–1199.
flow field, J. Fluid Mech. 140 (1984) 223–256. [71] I. Kleis, P. Kulu, Solid Particle Erosion: Occurrence, Prediction and Control,
[54] S. Ogawa, A. Umemura, N. Oshima, On the equation of fully fluidized granular Springer Science & Business Media, 2007.
materials, J. Appl. Math. Phys. 31 (1980) 483–493. [72] H. Uuemǒis, I. Kleis, A critical analysis of erosion problems which have been little
[55] M. Hamzehei, CFD modeling and simulation of hydrodynamics in a fluidized bed studied, Wear 31 (2) (1975) 359–371.
dryer with experimental validation 2011, ISRN Mechanical Engineering, 2011, [73] S. Subramaniam, Lagrangian–Eulerian methods for multiphase flows, Prog. Energy
p. 2011. Combust. Sci. 39 (2) (2013) 215–245.
[56] B.G.M. Van Wachem, J.C. Schouten, R. Krishna, C.M. Van Den Bleek, Eulerian [74] P. Dutta, S.K. Saha, N. Nandi, N. Pal, Numerical study on flow separation in 90◦
simulations of bubbling behaviour in gas-solid fluidised beds, Comput. Chem. Eng. pipe bend under high Reynolds number by k-ε modelling, Engineering Science and
22 (supplement 1) (1998) S299–S306. Technology, an International Journal 19 (2) (2016) 904–910.
[57] B.G.M. Van Wachem, J.C. Schouten, C.M. Van den Bleek, R. Krishna, J.L. Sinclair, [75] A. Jain, Experimental Investigation of Turbulent Flow in a Pipe Bend Using Particle
CFD modeling of gas-fluidized beds with a bimodal particle mixture, AIChE J. 47 Image Velocimetry, Master Thesis, McMaster University, Alberta, Canada, 2017.
(6) (2001) 1292–1302. [76] A. Pantokratoras, Steady laminar flow in a 90◦ bend, Adv. Mech. Eng. 8 (9) (2016)
[58] A. Mansouri, H. Arabnejad, S.A. Shirazi, B.S. McLaury, A combined CFD/ 1–9.
experimental methodology for erosion prediction, Wear 332 (2015) 1090–1097. [77] K. Sudo, M. Sumida, H. Hibara, Experimental investigation on turbulent flow in a
[59] K. Ahlert, Effects of Particle Impingement Angle and Surface Wetting on Solid circular-sectioned 90-degree bend, Exp. Fluid 25 (1) (1998) 42–49.
Particle Erosion of AISI 1018 Steel, M.S. Thesis, University of Tulsa, OK, 1994. [78] S. Akhshik, M. Behzad, M. Rajabi, CFD–DEM approach to investigate the effect of
[60] B.S. McLaury, Predicting Solid Particle Erosion Resulting from Turbulent drill pipe rotation on cuttings transport behavior, J. Petrol. Sci. Eng. 127 (2015)
Fluctuation in Oilfield Geometries, PhD Dissertation, University of Tulsa, OK, 229–244.
1996. [79] G. DNV, Recommended Practice–DNVGL-RP-O501–Managing Sand Production and
[61] G. Grant, W. Tabakoff, Erosion prediction in turbomachinery resulting from Erosion, DNV GL, 2015, pp. 3–60.
environmental solid particles, J. Aircraft 15 (5) (1975) 471–478. [80] Fluent User’s Guide, Release 14, Ansys Inc., USA, 2011.
[62] M. Sommerfeld, N. Huber, Experimental analysis and modelling of particle-wall [81] J. Zhang, F. Darihaki, S.A. Shirazi, A comprehensive CFD-based erosion prediction
collisions, Int. J. Multiphas. Flow 25 (6–7) (1999) 1457–1489. for sharp bend geometry with examination of grid effect, Wear 430 (2019)
[63] A. Forder, M. Thew, D. Harrison, Numerical investigation of solid particle erosion 191–201.
experienced within oilfield control valves, Wear 216 (2) (1998) 184–193. [83] M. Syamlal, W. Rogers, T.J. O’Brien, MFIX Documentation: Volume1, Theory
[64] Gambit User’s Guide, Release 2. 3. 16, Ansys Inc., USA, 2006. Guide>, National Technical Information Service, Springfield, VA, 1993. DOE/
METC-9411004, NTIS/DE9400087, 1993.

14

You might also like