Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Numerical simulations of viscoelastic fluid flows are always a hot issue in the field of computational fluid dy
SPH namics and the study of their complex rheological properties has important academic and engineering appli
Phan-Thien-Tanner model cation value. In the present work, we develop a smoothed particle hydrodynamics (SPH) method for simulating
Viscoelastic fluid flow
transient viscoelastic fluid flows governed by the Phan-Thien-Tanner (PTT) constitutive equation. To improve
Poiseuille flow
Droplet dynamics
the computational accuracy of the SPH method, the mixed symmetric correction algorithm of kernel gradient is
implemented. To remove the particle clustering and unphysical fracture in fluid stretching which is named as the
tensile instability, the artificial stress model is added into the momentum equation. We firstly apply the proposed
SPH method to solve the plane Poiseuille flow of a PTT viscoelastic fluid, in which the effectiveness and
advantage of the method are verified by comparing the SPH solution with those obtained by the finite volume
method (FVM) and analyzing the l2 norm error of different SPH solutions to the FVM solution. Then, the method
is employed to simulate the impact behavior of a PTT viscoelastic droplet with a rigid plate. In particular, we not
only investigate the spreading behavior of PTT viscoelastic droplet after impacting the rigid plate, but also for the
first time capture and analyze the bouncing behavior of droplet by decreasing the Reynolds number. The in
fluences of the Reynolds number, the Weissenberg number, the solvent viscosity ratio, and the PTT elongational
parameter on the droplet dynamics behavior are further deeply studied. Numerical results demonstrate that the
SPH method proposed in this paper is a powerful computation tool for simulating PTT viscoelastic fluid flows and
is capable of effectively describing their complex rheological properties and free surface variation characteristics.
1. Introduction marker and cell (MAC) [2], or level set [3] methods need to be employed
additionally, which makes the implementation procedure very complex.
Viscoelastic fluid flows are widely present in many industrial pro In the last few years, grid-based methods have been employed to
duction processes, such as extrusion and injection of polymers, ink jet simulate viscoelastic fluid flows governed by different constitutive
devices, container filling in the food and cosmetic industries, and so on. models such as upper Convected-Maxwell (UCM) model [4,5],
Usually, these flows involve complex free surfaces. To improve pro Oldroyd-B model [6,7], FENE-CR model [8–10], Giesekus model [11,
ductive efficiency and product performance, the study of complex 12], Phan-Thien-Tanner (PTT) model [13,14], among others.
rheological properties of viscoelastic fluids has attracted the attention of Among these models, the PTT model has been the subject of work by
many scientists. Viscoelastic free surface flows do not have analytic some scientists since it provides a better fit to the rheology of polymer
solutions and so for this reason, development of numerical methods for melts and concentrated solutions than some other simple models such as
predicting this class of flow problems has been a hot issue in the field of UCM or Oldroyd-B models [13]. At present, several works have been
computational fluid dynamics. Typical numerical techniques for dealing made to investigate viscoelastic free surface flows of PTT fluids. de Paulo
with viscoelastic fluid flows are grid-based, including finite volume et al. [15] developed a MAC technique to simulate two-dimensional (2D)
(FVM), finite element (FEM), and finite difference (FDM) methods. PTT viscoelastic free surface flows. In their work, the extrudate swell
Nevertheless, to capture the moving free surface, the volume of fluid [1], and the jet flowing onto a rigid plate were investigated. Tomé et al. [16]
* Corresponding author.
E-mail address: xiaoyang.xu@xust.edu.cn (X. Xu).
https://doi.org/10.1016/j.enganabound.2023.11.020
Received 25 September 2023; Received in revised form 18 November 2023; Accepted 18 November 2023
0955-7997/© 2023 Elsevier Ltd. All rights reserved.
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
extended the work of de Paulo et al. [15] to three-dimensional (3D) 2. Governing equations
space and simulated 3D extrudate swell and a 3D jet flowing onto a rigid
plate. Evans et al. [17] simulated transient planar stick-slip flow of a PTT In a Lagrangian frame, the mass and momentum conservation
fluid and discussed the polymer stress behavior around the stress sin equations for the flow of an isothermal, viscoelastic, weakly compress
gularity at the stick-slip point. França et al. [18] adopted FDM to ible fluid are respectively written as [44]
investigate 2D PTT viscoelastic binary droplet collision, where the free
dρ ∂vβ
surface dynamic was accomplished by the front-tracking representation = − ρ β, (1)
with marker particles, combined with the MAC method.
dt ∂x
In the past few decades, the smoothed particle hydrodynamics (SPH) dvα 1 ∂σ αβ
method [19,20] has shown a substantial potential in dealing with free = + Fα , (2)
dt ρ ∂xβ
surface flows and gradually become an alternative to grid-based nu
merical methods. The SPH method is actually a Lagrangian particle where ρ is the fluid density, t is the time, xβ is the spatial coordinate, vβ is
method, in which the problem to be solved is discretized by a set of the velocity, σαβ is the total stress tensor, Fα is the external force acting
particles rather than fixed meshes. These particles could be moving on the fluid, and dtd is the material derivative, i.e., dtd = ∂∂t + vβ ∂x∂β .
according to the solutions of governing equations. Thus, the governing
The total stress tensor σαβ in (2) is usually decomposed into the
equations are finally converted into the equations of motion for these
pressure p, the viscous stress τsαβ , and the elastic stress ταp β , i.e.,
particles. In comparison with grid-based methods, the SPH method has
the following advantages for flow simulations: (a) the moving free sur σ αβ = − pδαβ + ταs β + ταp β , (3)
faces can be modeled without the need of any explicit surface tracking
techniques; (b) the advection and transport of the system could be where δαβ is the Kronecker function. The viscous stress ταs β is calculated
calculated very well due to its Lagrangian and adaptive nature; (c) nu based on the Newtonian fluid model
merical implementation for complex physical problems is a compara
tively easy task. Owing to these merits, the SPH method has been τsαβ = 2ηs dαβ , (4)
adopted to simulate a wide variety of scientific and engineering prob
lems including dam-break flow [21–24], explosion [25–27], biome where ηs is the Newtonian solvent viscosity and dαβ is the deformation
chanics [28,29], multi-phase flows [30–32], heat transfer [33–35], rate tensor
fluid-structure interaction [36–38], viscoelastic fluid flows [39–42], etc. (
1 ∂vα ∂vβ
)
Regarding the application of SPH to viscoelastic free surface flows, d αβ = + . (5)
2 ∂xβ ∂xα
Fang et al. [43] added an artificial stress term into the SPH method to
alleviate the tensile instability problem and simulated the spreading
deformation of a 2D Oldroyd-B droplet impacting a rigid plate. Their 2.1. PTT constitutive equation
method was validated by comparing the SPH results with those of Tomé
et al. [6]. Xu et al. [44] explored the feasibility of SPH in handing 3D-un To close the governing equations, a constitutive equation related to
steady viscoelastic free surface flows, where the impact of a 3D the elastic stress ταp β is required. In this paper, we consider the PTT
Oldroyd-B droplet with a rigid wall and 3D jet buckling problem were constitutive model [13,15]
investigated. Zainali et al. [45] developed an improved multiphase SPH
method with an improved interface treatment to simulate the defor ∇
f (tr(τ))ταpβ + λ1 τpαβ = 2ηp dαβ . (6)
mation of a Newtonian droplet in a viscoelastic Oldroyd-B fluid. Vahabi
et al. [46] studied the interaction of a pair of in-line bubbles ascending in In the PTT constitutive model, f(tr(τ)) is defined as
an Oldroyd-B liquid using a weakly compressible SPH method. Moinfar
ελ1
et al. [47] analyzed the deformation of a Giesekus droplet under simple f (tr(τ)) = 1 + tr(τ), (7)
ηp
shear flow. Despite these successful applications of SPH to viscoelastic
fluid flows, most of these works are made based on the UCM, Oldroyd-B,
where ε is the PTT elongational parameter, λ1 is the relaxation time of
or Giesekus models, and SPH simulations of viscoelastic fluid flows ∇
based on the PTT model are very rare. the fluid, ηp is the polymer viscosity, and ταp β is the upper-convected
In this paper, we develop the SPH method for simulations of 2D derivative of ταp β
transient viscoelastic fluid flows governed by the PTT constitutive
equation. To improve the computational accuracy of the SPH method, ∇ dταp β ∂vα γβ ∂vβ αγ
ταpβ = − τ − τ . (8)
the mixed symmetric correction algorithm of kernel gradient is imple dt ∂xγ p ∂xγ p
mented. To remove the particle clustering and unphysical fracture in
From Eqs. (6)–(8), we obtain the PTT constitutive equation as
fluid stretching which is named as the tensile instability, the artificial
stress model is added into the momentum equation. We firstly apply the αβ
d τp ∂vα ∂vβ αγ f (tr(τ)) αβ 2ηp αβ
= γ τγβ
p + τ − τp + d . (9)
proposed SPH method to solve the plane Poiseuille flow of a PTT dt ∂x ∂xγ p λ1 λ1
viscoelastic fluid, in which the effectiveness and advantage of the
It is obvious that the Oldroyd-B constitutive equation can be recov
method are verified by comparing the SPH solution with those obtained
ered by simply setting the PTT elongational parameter ε=0 in the code.
by the FVM method and analyzing the l2 norm error of different SPH
solutions to the FVM solution. Then, the method is employed to simulate
the impact of a PTT viscoelastic droplet with a rigid plate. In particular, 2.2. Equation of state
we not only investigate the spreading behavior of PTT viscoelastic
droplet after impacting the rigid plate, but also for the first time capture There are two commonly-used methods for solving the governing
and analyze the bouncing behavior of droplet by decreasing the Rey equations in the framework of SPH. One is the incompressible SPH
nolds number. The influences of the Reynolds number, the Weissenberg method [48], in which the fluid pressure is implicitly evaluated from a
number, the solvent viscosity ratio, and the PTT elongational parameter pressure Poisson equation; the other is the weakly compressible SPH
on the droplet dynamics behavior are further deeply studied. method [49], in which the pressure is explicitly calculated from a stiff
equation of state. In this paper, we follow the latter by using the
following equation of state [49]
474
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
( )
∂vα ∑mj ( ) ∂W where Sαijβ is the artificial stress. Sαijβ is defined as
(15)
ij
kiαβ = = vαj − vαi ,
∂xβ i ρ j ∂xβi ( )
(20)
j
Sijαβ = fijn Rαi β + Rjαβ ,
i and j are the index of particle, m is the particle mass, Wij = W(|xi − xj|,
h)is the kernel function, h is the smooth length of the kernel function. where n = W(0, h)/W(Δx, h), fij = Wij/W(Δx,h), and Rαβ is the artificial
With respect to the kernel function, the quintic spline kernel is stress tensor. To compute Rαβ, a new coordinate frame (x′, y′) is specially
employed because of its better accuracy and stability resulting from the constructed by rotating the original (x, y). The angle of coordinate
continuous second derivative [50]: rotation θ is given by
⎧ 1 2σ xy
⎪
⎪ (3 − q)5 − 6(2 − q)5 + 15(1 − q)5 , 0 ≤ q < 1 θ = tan− 1 xx , (21)
⎨ 2 σ − σyy
(3 − q)5 − 6(2 − q)5 , 1≤q<2
W(r, h) = W0 × , (16)
⎪
⎪ (3 − q)5 , 2≤q<3 where σxx, σyy, and σxy are the total stress tensor in the original (x, y).
⎩
0, q≥3
Note that the purpose of introducing the rotation angle is to maintain the
where q = r/h and r is the distance between two particles. The conservation properties of the system [43,59]. Then, σ′xx and σ′yy in the
normalization factor W0 takes the value of 7/64πh2 in 2D space. new (x′, y′) is calculated according to:
In comparison with grid-based methods, the computational accuracy xx
σ′yy = σxx cos2 θ + 2σxy sinθcosθ + σyy sin2 θ, . (22)
of SPH is comparatively low, which limits its application to some com σ′ = σxx sin2 θ − 2σ xy sinθcosθ + σ yy cos2 θ,
plex flow problems. The low computational accuracy originates from the
particle approximations adopted in SPH and discontinuities caused by Once we obtain σ′xx and σ′yy, the diagonal components R′xx and R′yy of
irregular distribution of particles. To improve the computational accu the artificial stress tensor in the new (x′, y′) is determined by
racy of SPH, some researchers have developed various modified SPH ⎧ ′xx
algorithms, including reproducing kernel particle methods [51], finite ⎨ − ε σ , σ′xx > 0;
particle method [52,53], modified SPH method [54,55], symmetric SPH R′xx
= ρ2 , (23)
⎩
method [56,57], improved SPH method [58], and so on. In this paper, 0, otherwise
the mixed symmetric correction algorithm of kernel gradient [49,56,57]
where ε is the artificial stress parameter with the value ranging from 0 to
is implemented, that is, the kernel gradient ∂Wij /∂xβi in the SPH dis
1. Finally, the artificial stress tensor Rαβ is calculated from the relations
cretization (11)-(15) is modified as
xx yy
⎛ ̃ ⎞ ⎛ ⎞ Rxx = R′ cos2 θ + R′ sin2 θ,
∂W ij ∂Wij xx yy
(24)
⎜ ∂xi ⎟ ⎜ ∂xi ⎟ Ryy = R′ sin2 θ + R′ cos2 θ, .
′xx yy
⎜
⎜
⎟ ⎜
⎟ = M− 1 ⎜
⎟
⎟, (17) R = cosθsinθ(R − R′ )
xy
⎝ ∂W̃ ⎠ ⎝ ∂Wij ⎠
ij
∂yi ∂yi 3.3. Artificial viscosity
475
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
the artificial viscosity is also added into the momentum equation. particles are also kept unchanged. Each dummy particle is associated
Accordingly, we have the SPH discretization of the momentum equation with a boundary particle in the normal direction of boundary. To satisfy
( α) ( αβ
) the Newman boundary conditions of pressure and elastic stress, the
dv ∑ σ αβ σ j ∂W̃ ij
pressure and elastic stress of dummy particles are set to be the same as
= mj i2 + 2 + Sijαβ − Πij δαβ + Fiα , (25)
dt i j
ρi ρj ∂xβi those of the associated boundary particle. The velocities of dummy
particles are set to be zero to satisfy no-slip boundary condition.
where Πij is the artificial viscosity. Πij has the form [60]:
⎧
3.5. Time integration scheme
2
⎨ − αΠ cij φij + βΠ φij , v ⋅r < 0
⎪
(26)
ij ij
Πij = ρij .
⎪
⎩
0, vij ⋅rij ≥ 0 To solve the governing equations, the leapfrog time integration
scheme is adopted due to its high computational efficiency. Let Xi de
where αΠ and βΠ are the artificial viscosity parameters, notes the vector of variables (ρi , vαi , ταp,iβ ) and Bi denotes the vector of the
ci + cj right-hand sides of governing equations. Then, the leapfrog time inte
cij = , (27) gration scheme is summarized as follows [44,50]: at the end of the first
2
time step t0, Xi is advanced at half a time step, while the particle position
φij = ⃒ ⃒2
hvij ⋅rij
, (28) ri is advanced by a full time step
⃒rij ⃒ + 0.01h2 ⎧
⎪
⎪ t = t0 + Δt
⎪
⎨
Δt
ρij =
ρi + ρj
, (29) Xi (t0 + Δt/2) = Xi (t0 ) + Bi (t0 ) . (31)
2 ⎪
⎪ 2
⎪
⎩
ri (t0 + Δt) = ri (t0 ) + ui (t0 + Δt/2)Δt
vij = vi − vj, and rij = ri − rj. The term related to αΠ produces the shear
and the bulk viscosities, while that of βΠ is employed to prevent unre To ensure consistency with the particle position at each subsequent
alistic particle penetration, which is similar to the von-Neumann- time step, Xi is advanced by half a time step
Richtmeyer viscosity adopted in FDM. The term of 0.01h2 is added in Δt
order to avoid numerical instability when two particles are too close to Xi (t) = Xi (t − Δt / 2) + Bi (t − Δt) . (32)
2
each other.
At the end of the subsequent time step, the density, velocity, elastic
stress, and displacement of the particles are advanced according to the
3.4. Boundary conditions standard leapfrog scheme
⎧
⎨ t = t + Δt
In this paper, there are two kinds of boundary conditions involved: X (t + Δt/2) = Xi (t − Δt/2) + Bi (t)Δt . (33)
one is the free surface boundary condition; the other is the wall ⎩ i
ri (t + Δt) = ri (t) + ui (t + Δt/2)Δt
boundary condition.
For the free surface boundary condition, the free surface particles Finally, in order to maintain the numerical stability, the time step Δt
need to be identified. The stress-free condition σ ⋅ n = 0 should be should be constrained by the following CFL and mass and viscous-
imposed on these particles. However, this condition is implicitly satis diffusion conditions [44,50]:
fied with SPH as the stress-free condition is represented in a surface ( √̅̅̅̅̅ )
integral which is neglected in SPH approximation for gradients and h h h2
Δt ≤ min , min , 0.125 , (34)
happens to be zero [61]. As such, the free surface can be modeled c i Fi υ0
naturally without the need of any explicit surface tracking techniques.
For the wall boundary condition, we handle it by using two types of where Fi is the force per unit mass, equivalent to the magnitude of the
virtual particles, i.e., boundary particles and dummy particles [21]. particle acceleration, and υ0 = η/ρ is the kinematic viscosity.
First, only one layer of boundary particles is placed on the wall
boundary. The initial particle spacing for the boundary particles is Δx. 4. Numerical examples
However, the boundary particles do not exert an artificial repulsive
force. Instead, they participate in calculation of velocity, pressure, and 4.1. Plane Poiseuille flow of a PTT fluid
elastic stress in the governing equations in order to effectively prevent
the fluid particles from penetrating the wall boundary. The density and At present, there has no works on SPH simulation of plane Poiseuille
position of the boundary particles remain unchanged during the calcu flow based on the PTT model. Therefore, we select it as the first example
lation and their velocities are set to be zero in order to impose a no-slip of numerical simulation. Meanwhile, in order to better describe the flow,
boundary condition. The pressure and elastic stress are calculated ac we introduce the dimensionless Reynolds number
cording to the following Shepard interpolation of the adjacent fluid
ρVL
particles [21] Re = , (35)
η
∑mj
ρj Aj Wij
j
Ai = ∑mj , (30) the dimensionless Weissenberg numbers
ρ Wij j
λ1 V
(36)
j
Wi = ,
L
where A denotes the pressure or elastic stress, i is the boundary particle, j
is the fluid particle adjacent to the particle i. Although the boundary and the solvent viscosity ratio
particles are placed along the wall boundary, the supporting domain of ηs
fluid particles is still truncated by the wall boundary. Therefore, in order β= , (37)
η
to further resolve this problem, three layers of dummy particles are ar
ranged outside the boundary with an initial particle spacing Δx. With where V and L denote the characteristic velocity and characteristic
the same as the boundary particles, the density and position of dummy length of the flow, respectively.
476
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
and 8. The time changes of the u-velocity and the shear stress τxy at
points A, B, and C are further shown in Fig. 2b and Fig. 2c, respectively.
Obviously, these three monitoring points all exhibits the phenomenon of
velocity and stress overshoot. The flow reaches the steady state after t =
8. The closer the monitoring point to the rigid plate is, the smaller the u-
velocity value is. Also, points A and C are symmetrical about y = 0, and
hence their velocity values are identical. As for the shear stress τxy, the
closer the monitoring point to the rigid plate is, the larger the shear
stress value is. It is always 0 at point B, while those of points A and C are
equal in magnitude and opposite in direction.
So far, there is no transient analytical solution for the plane Poiseulle
Fig. 2. SPH simulation of the plane Poiseuille flow of a PTT fluid (Re = 2, Wi = 1, β = 0.1, ε = 0.25): (a) the velocity profile; (b) the time change of velocity u at
points A to C; (c) the time change of elastic shear stress τxy at points A to C.
477
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
Fig. 6. SPH simulation of the impact of a PTT droplet with a rigid plate (Re = 5, Wi = 1, β = 0.1, ε = 0.25).
478
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
479
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
Table 1 parameters on the time evolution of the droplet’s spread width. Except
Calculation parameters for different models. the target parameter, the data adopted for different models are the same
Models Calculation parameters as those of Fig. 6 and are summarized in Table 1, which reads:
Influence of Re Wi = 1, β = 0.1, ε = 0.25 Re: 2, 5, 10
Influence of Wi Re = 5, β = 0.1, ε = 0.25 Wi: 0.5, 1, 2, 4, 8 • Influence of Re
Influence of β Re = 5, Wi = 1, ε = 0.25 β: 0.1, 0.3, 0.5, 0.7, 0.9
Influence of ε Re = 5, Wi = 1, β = 0.1 ε: 0.01, 0.1, 0.25, 0.5 To investigate the influence of the Reynolds number Re on the
droplet’s spreading behavior, we employ three different Re values, i.e.,
Re = 2, 5, 10. The droplet’s spread widths obtained by three different Re
initial particle spacings (i.e., Δx = 0.0004, 0.0002, 0.000125, and
values are presented in Fig. 9(a). It is observed that the maximum
0.0001) to simulate it. During the simulation process, all the other
spreading width of the droplet greatly increases with an increasing Re.
physical parameters remain unchanged. Fig. 8 shows the time evolutions
This is because: the larger Re is, the stronger the inertial force of the
of the droplet’s spread width obtained by four different Δx. As can be
droplet is, resulting in a fast spread of the droplet after impacting the
seen, there are no large differences between these four different SPH
rigid plate. Due to the strong inertial force, the retraction behavior of the
solutions, indicating that the SPH method proposed in this paper is
droplet becomes weak. The retraction behavior of Re = 2 is obviously
convergent for the simulations of PTT viscoelastic free surface flows.
stronger than that of Re = 10. The retraction behavior of the droplet is in
Meanwhile, in order to better quantify the convergence of the method,
nature caused by the elasticity of the droplet. However, the elastic force
we similarly employ Eq. (38) to calculate the l2-norm error by
of the droplet becomes relatively small with a large Re. Therefore, there
comparing the SPH results obtained on Δx = 0.0004, 0.002, and
is a weak retraction behavior of the droplet after impacting the rigid
0.000125 to that obtained on Δx = 0.0001, where the SPH solution of
plate.
Δx = 0.0001 is taken as the reference solution. The calculated l2-norm
errors for Δx = 0.0004, 0.002, and 0.000125 compared to Δx = 0.0001
• Influence of Wi
are 1.16%, 0.67% and 0.23%, respectively. Obviously, the l2-norm error
decreases with refining particle, indicating the good convergence of the
As for the influence of the Weissenberg number Wi on the flow
proposed SPH method.
process, we observe from Fig. 9(b) that increasing Wi from 0.5 to 8 also
Below, we investigate the influence of the PTT rheological
Fig. 9. Influence of different rheological parameters on the time change of the PTT droplet’s spread width d(t*)/d0: (a) Re; (b) Wi; (c) β; (d) ε.
480
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
481
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
• Influence of β
• Influence of ε
482
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
Fig. 14. Variation of droplet bounce height with time under different β values Fig. 16. Influence of Re on the maximum bounce height of the droplet.
(Re = 0.2, Wi = 1,ε = 0.25).
bouncing behavior is and the higher the bounce height of the droplet is.
483
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
Data availability
Acknowledgments
Fig. 17. Influence of ε on the maximum bounce height of the droplet. The authors are grateful to Dr. Zhen Li of Xi’an University of Science
and Technology for assisting in the revision and participating in dis
bounce height of the droplet is plotted in Fig. 17, where eight different ε cussions and suggestions to improve the quality of the paper. This work
values, i.e., ε = 0.01, 0.02, 0.05, 0.1, 0.2, 0.4, 0.8, are selected. Mean is financially supported by the National Natural Science Foundation of
while, similar to that of Wi and Re, three different β with the value of β = China (Grant No. 12071367, 12172163), the Department of Science and
0.01, 0.05, and 0.1 are employed for the analysis of ε. It is observed that Technology of Guangdong Province (Grant No. 2023B1212060001),
the maximal bounce height of the droplet slightly decreases with an and Shenzhen Science and Technology Innovation Commission (Grant
increase of ε. Indeed, the larger ε is, the larger the elongational viscosity No. JSGG20220831101400002).
of the droplet at the same shear rate is and the more pronounced the
elongational effect of the droplet is. In this situation, the percentage of References
the elastic force of droplet relative to the total force decreases, leading to
a slightly weak bouncing behavior of the droplet. [1] Hirt CW, Nicholls BD. Volume of fluid (VOF) method for dynamics of free
boundaries. J Comput Phys 1981;39:201–21.
[2] McKee S, Tomé MF, Ferreira VG, Cuminato JA, Castelo A, Sousa FS,
Conclusions Mangiavacchi N. The MAC method. Comput Fluids 2008;37:907–30.
[3] Osher S, Fedkiw RP. Level set methods: an overview and some recent results.
J Comput Phys 2001;169(2):463–502.
In this paper, we have developed a SPH method for simulations of [4] Poole RJ, Pinho FT, Alves MA, Oliveira PJ. The effect of expansion ratio for
transient viscoelastic fluid flows governed by the PTT constitutive creeping expansion flows of UCM fluids. J Non-Newton Fluid Mech 2009;163:
equation. To improve the computational accuracy of SPH, the mixed 35–44.
[5] Cruz DOA, Pinho FT. Analytical solution of steady 2D wall-free extensional flows of
symmetric correction algorithm of kernel gradient is implemented. To
UCM fluids. J Non-Newton Fluid Mech 2015;223:157–64.
remove the tensile instability, the artificial stress model is added into the [6] Tomé MF, Mangiavacchi N, Cuminato JA, Castelo A, McKee S. A finite difference
momentum equation. The SPH method is employed to simulate the technique for simulating unsteady viscoelastic free surface flows. J Non-Newtonian
plane Poiseuille flow of a PTT fluid and the impact of a PTT droplet Fluid Mech 2002;106:61–106.
[7] Viezel C, Tomé MF, Pinho FT, McKee S. An Oldroyd-B solver for vanishingly small
impacting a rigid plate. The following conclusions are drawn from the values of the viscosity ratio: application to unsteady free surface flows. J Non-
results: Newton Fluid Mech 2020;285:104338.
[8] Malheiro JM, Oliveira PJ, Pinho FT. Parametric study on the three-dimensional
distribution of velocity of a FENE-CR fluid flow through a curved channel. J Non-
(1) For the plane Poiseuille flow of a PTT fluid, good agreement Newton Fluid Mech 2013;200:88–102.
between SPH and FVM solutions is achieved and the error of the [9] Paulo GS, Oishi CM, Tomé MF, Alves MA, Pinho FT. Numerical solution of the
present SPH in comparison with FVM is smaller than that of the FENE-CR model in complex flows. J Non-Newton Fluid Mech 2014;204:50–61.
[10] Ohta M, Furukawa T, Yoshida Y, Sussman M. A three-dimensional numerical study
traditional SPH, indicating that the present SPH method is on the dynamics and deformation of a bubble rising in a hybrid Carreau and FENE-
effective and has higher computational accuracy than traditional CR modeled polymeric liquid. J Non-Newton Fluid Mech 2019;265:66–78.
SPH for solving PTT viscoelastic flows. [11] Peng S, Li J, Xiong Y, Xu X, Yu P. Numerical simulation of two-dimensional
unsteady Giesekus flow over a circular cylinder. J Non-Newton Fluid Mech 2021;
(2) The PTT parameters have an important influence on the 294:104571.
spreading behavior of the droplet after impacting the rigid plate. [12] Tomé MF, Araujo MT, Evans JD, McKee S. Numerical solution of the Giesekus
That is, the larger Re or Wi are, the faster the droplet spreads. The model for incompressible free surface flows without solvent viscosity. J. Non-
Newton Fluid Mech 2019;263:104–19.
larger β or ε are, the weaker the retraction behavior of the droplet
[13] Alves MA, Pinho FT, Oliveira PJ. Study of steady pipe and channel flows of a
becomes. single-mode Phan-Thien–Tanner fluid. J Non-Newton Fluid Mech 2001;101:55–76.
(3) The PTT parameters have an important influence on the bouncing [14] Ferrás LL, Nóbrega JM, Pinho FT. Analytical solutions for channel flows of Phan-
behavior of the droplet. That is, the smaller Re, β, or ε are, the Thien–Tanner and Giesekus fluids under slip. J Non-Newton Fluid Mech 2012;171:
97–105.
higher the maximum bounce height of the droplet is. The [15] De Paulo GS, Tomé MF, McKee S. A marker-and-cell approach to viscoelastic free
maximum bounce height first increases and then decreases with surface flows using the PTT model. J Non-Newton Fluid Mech 2007;147(3):
an increasing Wi. 149–74.
[16] Tomé MF, Paulo GS, Pinho FT, Alves MA. Numerical solution of the PTT
(4) The SPH method proposed in this paper is a powerful computa constitutive equation for unsteady three-dimensional free surface flows. J Non-
tion tool for simulating PTT viscoelastic fluid flows and is capable Newton Fluid Mech 2010;165:247–62.
of effectively describing their complex rheological properties and [17] Evans JD, Cuminato JA, Palhares Junior IL, Oishi CM. Numerical study of the stress
singularity in stick-slip flow of the Phan-Thien Tanner and Giesekus fluids. Phys
free surface variation characteristics. Fluids 2019;31(9):093101.
[18] França HL, Oishi CM, Thompson RL. Numerical investigation of shear-thinning and
CRediT authorship contribution statement viscoelastic binary droplet collision. J Non-Newton Fluid Mech 2022;302:104750.
[19] Gingold RA, Monaghan JJ. Smoothed particle hydrodynamics theory and
application to non-spherical stars. Mon Not R Astron Soc 1977;181:375–89.
Xiaoyang Xu: Methodology, Supervision, Conceptualization. Jie [20] Lucy LB. A numerical approach to the testing of the fission hypothesis. Astron J
1977;82:1013–24.
484
X. Xu et al. Engineering Analysis with Boundary Elements 158 (2024) 473–485
[21] Xu X, Jiang YL, Yu P. SPH simulations of 3D dam-break flow against various forms [43] Fang J, Owens RG, Tacher L, Parriaux A. A numerical study of the SPH method for
of the obstacle: toward an optimal design. Ocean Eng 2021;229:108978. simulating transient viscoelastic free surface flows. J Non-Newton Fluid Mech
[22] Chang CC, Wu YT. SPH modeling of dam-break bores on smooth and macro- 2006;139:68–84.
roughness slopes. Ocean Eng 2023;279:114484. [44] Xu X, Ouyang J, Jiang T, Li Q. Numerical simulation of 3D-unsteady viscoelastic
[23] Sun PN, Pilloton C, Antuono M, et al. Inclusion of an acoustic damper term in free surface flows by improved smoothed particle hydrodynamics method. J Non-
weakly-compressible SPH models. J Comput Phys 2023;483:112056. Newtonian Fluid Mech 2012;177:109–20.
[24] Lyu HG, Sun PN, Liu PZ, et al. Derivation of an improved smoothed particle [45] Zainali A, Tofighi N, Shadloo MS, et al. Numerical investigation of Newtonian and
hydrodynamics model for establishing a three-dimensional numerical wave tank non-Newtonian multiphase flows using ISPH method. Comput Methods Appl Mech
overcoming excessive numerical dissipation. Phys Fluids 2023;35:067102. Eng 2013;254:99–113.
[25] Zhang ZL, Liu MB. Numerical studies on explosive welding with ANFO by using a [46] Vahabi M, Hadavandmirzaei H, Kamkari B, et al. Interaction of a pair of in-line
density adaptive SPH method. J Manuf Process 2019;41:208–20. bubbles ascending in an Oldroyd-B liquid: a numerical study. Eur J Mech B Fluids
[26] Peng YX, Zhang AM, Ming FR. Numerical simulation of structural damage 2021;85:413–29.
subjected to the near-field underwater explosion based on SPH and RKPM. Ocean [47] Moinfar Z, Vahabi S, Vahabi M. Numerical simulation of drop deformation under
Eng 2021;222:108576. simple shear flow of Giesekus fluids by SPH. Int J Numer Methods Heat Fluid Flow
[27] Chen JY, Feng DL, Liu JH, et al. Numerical modeling of the damage mechanism of 2023;33(1):263–81.
concrete-soil multilayered medium subjected to underground explosion using the [48] O’connor J, Domínguez JM, Rogers BD, et al. Eulerian incompressible smoothed
GPU-accelerated SPH. Eng Anal Bound Elem 2023;151:265–74. particle hydrodynamics on multiple GPUs. Comput Phys Commun 2022;273:
[28] Zhang C, Zhu Y, Yu Y, et al. An artificial damping method for total Lagrangian SPH 108263.
method with application in biomechanics. Eng Anal Bound Elem 2022;143:1–13. [49] Xu X, Deng XL. An improved weakly compressible SPH method for simulating free
[29] Zhang C, Gao H, Hu XY. A multi-order smoothed particle hydrodynamics method surface flows of viscous and viscoelastic fluids. Comput Phys Commun 2016;201:
for cardiac electromechanics with the Purkinje network. Comput Methods Appl 43–62.
Mech Eng 2023;407:115885. [50] Liu GR, Liu MB. Smoothed particle hydrodynamics: a mesh-free particle method.
[30] He F, Zhang H, Huang C, Liu M. A stable SPH model with large CFL numbers for Singapore: World Scientific; 2003.
multi-phase flows with large density ratios. J Comput Phys 2022;453:110944. [51] Liu WK, Jun S, Zhang YF. Reproducing kernel particle methods. Int J Numer
[31] Yang Q, Xu F, Yang Y, et al. A GPU-accelerated adaptive particle refinement for Methods Fluid 1995;20:1081–106.
multi-phase flow and fluid-structure coupling SPH. Ocean Eng 2023;279:114514. [52] Liu MB, Xie WP, Liu GR. Modeling incompressible flows using a finite particle
[32] Zöller C, Adams NA, Adami S. A partitioned continuous surface stress model for method. Appl Math Model 2005;29:1252–70.
multiphase smoothed particle hydrodynamics. J Comput Phys 2023;472:111716. [53] Liu MB, Liu GR. Restoring particle consistency in smoothed particle
[33] Li L, Yang X, Sohag MMA, Wang X, Liu Q. SPH-ASR study of drop impact on a hydrodynamics. Appl Num Math 2006;56:19–36.
heated surface with consideration of inclined angle and evaporation. Eng Anal [54] Zhang GM, Batra RC. Modified smoothed particle hydrodynamics method and its
Bound Elem 2022;141:235–49. applications to transient problems. Comput Mech 2004;34:137–46.
[34] Troconis J, Sánchez-Silva F, Carvajal-Mariscal I, et al. Simulation of van der Waals [55] Zhang GM, Batra RC. Wave propagation in functionally graded materials by
liquid droplets within a hot air atmosphere using the smoothed particle modified smoothed particle hydrodynamics (MSPH) method. J Comp Phys 2007;
hydrodynamics method. Int J Heat Mass Transf 2023;202:123749. 222:374–90.
[35] Yoo HS, Jo YB, Kim JW, et al. A simple Eulerian–Lagrangian weakly compressible [56] Batra RC, Zhang GM. SSPH basis functions for meshless methods, and comparison
smoothed particle hydrodynamics method for fluid flow and heat transfer. Int J of solution of solutions with strong and weak formulations. Comput Mech 2008;41:
Number Methods Eng 2023;124(4):928–58. 527–45.
[36] Meng ZF, Zhang AM, Yan JL, et al. A hydroelastic fluid–structure interaction solver [57] Zhang GM, Batra RC. Symmetric smoothed particle hydrodynamics (SSPH) method
based on the Riemann-SPH method. Comput Methods Appl Mech Eng 2022;390: and its application to elastic problems. Comput Mech 2009;43:321–40.
114522. [58] Fang J, Parriaux A, Rentschler M, Ancey C. Improved SPH methods for simulating
[37] Zhang C, Zhu Y, Lyu X, et al. An efficient and generalized solid boundary condition free surface flows of viscous fluids. Appl Num Math 2009;59(2):251–71.
for SPH: applications to multi-phase flow and fluid–structure interaction. Eur J [59] Monaghan JJ. SPH without a tensile instability. J Comput Phys 2000;159:290–311.
Mech B Fluids 2022;94:276–92. [60] Monaghan JJ. Smoothed particle hydrodynamics. Annu Rev Astron Astrophys
[38] Sun WK, Zhang LW, Liew KM. A coupled SPH-PD model for fluid–structure 1992;30:543–74.
interaction in an irregular channel flow considering the structural failure. Comput [61] Colagrossi A, Antuono M, Souto-Iglesias A, Touzé DLe. Theoretical analysis and
Methods Appl Mech Eng 2022;401:115573. numerical verification of the consistency of viscous smoothed-particle-
[39] Ellero M, Tanner RI. SPH simulations of transient viscoelastic flows at low hydrodynamics formulations in simulating free-surface flows. Phys Rev E 2011;84:
Reynolds number. J Non-Newton Fluid Mech 2005;132:61–72. 026705.
[40] Hashemi MR, Fatehi R, Manzari MT. SPH simulation of interacting solid bodies [62] Zhuang X, Ouyang J, Li W, Li Y. Three-dimensional simulations of non-isothermal
suspended in a shear flow of an Oldroyd-B fluid. J Non-Newton Fluid Mech 2011; transient flow and flow-induced stresses during the viscoelastic fluid filling
166:1239–52. process. Int J Heat Mass Transf 2017;104:374–91.
[41] Xu X, Ouyang J, Li W, et al. SPH simulations of 2D transient viscoelastic flows [63] Rafiee A, Manzari MT, Hosseini M. An incompressible SPH method for simulation
using Brownian configuration fields. J Non-Newton Fluid Mech 2014;208:59–71. of unsteady viscoelastic free-surface flows. Int J Non-Linear Mech 2007;42:
[42] Xu X, Yu P. A multiscale SPH method for simulating transient viscoelastic flows 1210–23.
using bead-spring chain model. J Non-Newton Fluid Mech 2016;229:27–42. [64] Oishi CM, Martins FP, Tomé MF, Alves MA. Numerical simulation of drop impact
and jet buckling problems using the extended Pom-Pom model. J Non-Newton
Fluid Mech 2012;169:91–103.
485