You are on page 1of 19

Applied Mathematical Modelling 35 (2011) 257–275

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

Modelling and simulation of moving interfaces in gas-assisted injection


moulding process
Li Qiang, Ouyang Jie *, Yang Binxin, Jiang Tao
Department of Applied Mathematics, Northwestern Polytechnical University, Xi’an 710129, PR China

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

Article history: The mathematical models of gas–liquid two-phase flow are introduced, in which the multi-
Received 5 July 2009 mode eXtended Pom–Pom (XPP) model is selected to predict the viscoelastic behavior of
Received in revised form 23 May 2010 polymer melt. The gas-penetration process is simulated using Level Set/SIMPLEC methods,
Accepted 4 June 2010
which can capture the moving interfaces at different time, including the gas–melt interface
Available online 9 June 2010
and the melt front. The physical features such as velocity, temperature and elasticity are
described at different time. The influences of gas delay time and injection pressure on
Keywords:
gas-penetration time and penetration length are analyzed. The numerical results show that
Gas-assisted
Two-phase flow
the Level Set/SIMPLEC methods can precisely trace the two moving interfaces in gas-pen-
Viscoelastic etration process, the fractional coverage increases at very low Deborah numbers, while at
Level Set higher Deborah numbers the fractional coverage decreases, and the penetration length is
SIMPLEC affected significantly by gas delay time and injection pressure.
Interface Ó 2010 Elsevier Inc. All rights reserved.

1. Introduction

Gas-assisted injection moulding (GAIM) is one of the innovative injection moulding processes recently developed.
Numerical simulation is now playing an important role in GAIM, especially tracing the two moving interfaces which are
the gas–melt interface and melt-front interface. At present, there are mainly two methods for tracing free interfaces based
on Eulerian description: volume of fluid (VOF) and the level set method. The level set method was introduced by Osher and
Fedkiw [1], which has been applied in various areas in recent years. It is a highly robust and accurate method for capturing
moving interfaces under complex motions. The major virtue of the level set method is that it works in any number of space
dimensions, handles topological merging and breaking naturally, and is easy to program.
In most previous papers the melt was simulated based on Hele–Shaw model, in which the gas–melt interface was traced
using VOF method [2,3]. Hele–Shaw model is simple, which can only describe the thin-wall injection moulding, and cannot
describe the effect of the elasticity of complicated fluid flows. Herein we choose the level set method to simulate the two mov-
ing interfaces in GAIM. The multi-mode extended Pom–Pom (XPP) model is implemented to predict the behavior of the vis-
coelastic melt, which is based on molecular theory of rheology and can provide a good fitting to the rheology of polymer melts
and concentrated solutions [4]. The energy equation is used to describe the total viscosity varying with temperature using the
Cross-WLF model. Due to the different physical properties, the dynamic interaction of gas and melt in GAIM become more
complicated. In this paper, the equations for gas–liquid two-phase flow are expressed by the generalized Navier–Stokes
equations, which are solved using the finite-volume method on collocated meshes, and it is easy to implement.
The effect of elasticity on the transient gas penetrating process is another important aspect in GAIM. Huzyak and Koelling
[5] found that the layer thickness increased at higher Deborah numbers. The phenomenon that the layer thickness increased

* Corresponding author. Tel.: +86 29 8849 5234; fax: +86 29 8849 1000.
E-mail address: jieouyang@nwpu.edu.cn (J. Ouyang).

0307-904X/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2010.06.002
258 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

caused by elasticity was also observed by Gauri and Koelling [6], who compared Newtonian and Boger fluids experimentally.
Ro and Homsy [7] analyzed theoretically the injection of air into a Hele–Shaw cell filled with an Oldroyd-B fluid, and pre-
dicted that normal stress thinning raises the layer thickness. Yijie Wang [8] and Henrik and Torbjorn [9] found that an in-
crease in the steady fractional coverage at low Deborah numbers, where the melt behaves like a diluted un-entangled
system. At higher Deborah numbers, where the melt behaves as an entangled melt system, the steady fractional coverage
decreases. In this paper, the effect of elasticity is analyzed in details, and the changing tendency of fractional coverage with
Deborah number are in agreement with the experimental results [8,9].
The remainder of this paper is outlined as follows. In Section 2 the mathematical model for GAIM is established, in which
the level set function for the moving interfaces is introduced. The generalized Navier–Stokes equations for gas–liquid two-
phase flow are given. The total viscosity of the polymer melt is characterized by the Cross-WLF model which varies with tem-
perature. The finite-volume scheme on collocated meshes is given and analyzed in detail in Section 3. The numerical results
are shown and discussed in graphical form in Section 4. The paper ends in Section 5 with some conclusions.

2. Mathematical model for GAIM

Due to the different physical properties, the dynamic interaction of gas and melt in GAIM become more complicated. So
we make the following four assumptions:

1. The gas velocity is much less than 70 m/s, and the pressure of the melt is also not too high, so both of gas and melt can be
treated as the incompressible fluid.
2. The gas viscosity is eight orders of magnitude smaller than the melt viscosity, which means that the Reynolds number is
too large in the gas regions to be solved in the unified equations. In this paper, we make the gas viscosity only three orders
of magnitude smaller than the melt viscosity artificially.
3. The pressure drop is small in the gas regions, so the gas regions can be treated as areas of permanent pressure.
4. The polymer melt has high viscosity, while the surface tension is very small, so the effect of the surface tension can be
neglected.

2.1. Governing equations

For the incompressible fluids including gas and melt, the continuity equation, momentum equation and energy equation
can be written as the generalized equations
r  u ¼ 0; ð1Þ

@ðqð/ÞuÞ
þ r  ðqð/ÞuuÞ  r  ð2gð/ÞDÞ ¼ rp þ He ð/Þr  ð2ðb  1Þgð/ÞDÞ þ He ð/Þr  s; ð2Þ
@t

@ðqð/ÞC v ð/ÞTÞ
þ r  ðqð/ÞC v ð/ÞuTÞ  r  ðjð/ÞrTÞ ¼ He ð/Þsij : rv i ; ð3Þ
@t
where
1
D¼ ðru þ ruT Þ; ð4Þ
2
qð/Þ ¼ qg þ ðqm  qg ÞHe ð/Þ; ð5Þ
gð/Þ ¼ gg þ ðgm  gg ÞHe ð/Þ; ð6Þ
C v ð/Þ ¼ C g þ ðC m  C g ÞHe ð/Þ; ð7Þ
jð/Þ ¼ jg þ ðjm  jg ÞHe ð/Þ; ð8Þ
8
> 0 if / < e
<   
He ð/Þ ¼ 1
1 þ /e þ p1 sin pe/ if j/j 6 e ; ð9Þ
>2
:
1 if / > e
where q, g, C and j are density, viscosity, specific heat and thermal conductivity, respectively. / is the level set function,
which will be detailed in Section 2.3. The subscript g and m represent gas and melt, respectively e = 1.5 Dx. In order to com-
pute conveniently, the governing Eqs. (1)–(3) are nondimensionalized via
x y u v q g p U
x ¼ ; y ¼ ; u ¼ ; v ¼ ; q ¼ ; g ¼ ; p ¼ ; t  ¼ t;
L L U U qm gm qm U 2 L
ð10Þ
j Cv L T
j ¼ ; C v ¼ ; sij ¼ s; T  ¼ ;
jm Cm gm U T0
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 259

where L and U are length scale and velocity scale, respectively. T0 is temperature scale. We define k = qg/qm, l = gg/gm, h = Cg/
Cm, c = jg/jm, then q = q(/) = k + (1  k)  He(/), Cv = Cv(/) = h + (1  h)He(/), j = j(/) = c + (1  c)He(/). For simplicity, the
mark ‘‘” is omitted, and the governing Eqs. (1)–(3) can be written in component form
@u @ v
þ ¼ 0; ð11Þ
@x @y
  !
@ðquÞ @ðquuÞ @ðqv uÞ 1 @ 2 ðguÞ @ 2 ðguÞ
þ þ  þ
@t @x @y Re @x2 @y2
!  
@p b  1 @ 2 ðguÞ @ 2 ðguÞ He ð/Þ @ sxx @ sxy
¼  þ He ð/Þ 2
þ 2
þ þ ; ð12Þ
@x Re @x @y Re @x @y

  ! !  
@ðqv Þ @ðquv Þ @ðqvv Þ 1 @ 2 ðgv Þ @ 2 ðgv Þ @p b  1 @ 2 ðgv Þ @ 2 ðgv Þ He ð/Þ @ sxy @ syy
þ þ  þ ¼  þ H e ð/Þ þ þ þ ;
@t @x @y Re @x2 @y2 @x Re @x2 @y2 Re @x @y
ð13Þ
  !
@ ðqC v T Þ @ðqC v TÞ @ðqC v TÞ @ 2 ðjTÞ @ 2 ðjTÞ
Pe þu þv  þ
@t @x @y @x2 @y2
   
@u @v @u @ v
¼ Br  He ð/Þ sxx þ syy þ sxy þ ; ð14Þ
@x @y @y @x
where Reynolds number Re = qmUL/gm, Peclet number Pe = qmCm  UL/jm, Brinkman number Br = gm U2/jmT0. When Eq. (14)
is solved in the gas regions the right hand of that equals zero since the gas viscosity is very small.
The extra tensor stress s is defined by XPP constitutive equation
r a
f ðk; sÞs þ k0b s þ G0 ðf ðk; sÞ  1ÞI þ s  s ¼ 2k0b G0 d; ð15Þ
G0
where d is the rate of deformation tensor, which is given by d = 1/2(ru + (ru)T),
  " # sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k0b mðk1Þ 1 1 aIss jIs j 2
f ðk; sÞ ¼ 2 e 1 þ 2 1 ; k¼ 1þ ; m¼ :
k0s k k 3G20 3G0 q

In these equations k0b and k0s are the orientation and backbone stretch relaxation times, respectively, and G0 is the linear
relaxation modulus. The constitutive equation possesses features of the Giesekus model since a non-zero second normal
stress difference is predicted when a – 0. g = ls + lp is the total viscosity. b = ls/(ls + lp). I refers to the trace of a tensor,
q is the number of arms, and k is tensile modulus. f(k, s), k, v are expressed by the following formula
  "  2 # sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 1 tðk1Þ 1 De a De 2
f ðk; sÞ ¼ 1 e þ 2 1 I ss ; k¼ 1þ Is ; t¼ :
e k k 1b 3 1b q

where the Deborah number De = k0bU/L.


In order to improve stability of calculation, (2/3)g(@uk/oxk)dij (dij is a d function) is added to the constitutive equation
[10,11], which approaches to zero in the numerical approximation. Then each of the governing equations is written using
the following general conservative form
   
@U @ @U @ @U
d1 þ uUd2  C þ v Ud2  C ¼ S/ ; ð16Þ
@t @x @x @y @y
For Cartesian coordinates, d1, d2 and C are constants, and U and S/ are functions that are defined depending on the particular
equation under consideration (see Table 1).

2.2. Cross-WLF model

The Cross model, which is the most appropriate model to study both filling and packing phases, has been chosen to assess
the total viscosity g of the polymer melt. Cross approach has been selected in order to adjust better the temperature and
pressure sensitivities of zero-shear-rate viscosity [12].
The expression of the Cross model is as follows
g0 ðT; pÞ
gðT; c_ ; pÞ ¼ ; ð17Þ
1 þ ðg0 c_ =s Þ1n
260 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

Table 1
Definition of the constants and functions in the general Eq. (16).

Equation U d1 d2 C S/
Continuity 1 0 1 0 0
g
u-momentum u q q  @p b1 @ 2 ðguÞ 2
ðguÞ
þ @ @y sxx þ @ sxy
þ HeReð/Þ @@x
Re @x þ He ð/Þ Re @x2 2 @y

g
v-momentum v q q  @p b1 @ 2 ðgv Þ 2
ðgv Þ
þ @ @y
@s @s
þ HeReð/Þ @xxy þ @yyy
Re @x þ He ð/Þ Re @x2 2

4  @u
sxx normal stress sxx De De 0 f ðk; sÞsxx þ 2De @u aDe s2xx þ s2xy þ  1b 2 @v
@x sxx þ 2De @y sxy  1b De ðf ðk; sÞ  1Þ  3 @x
@u
3  2b @x
4 
syy normal stress syy De De 0 f ðk; sÞsyy þ 2De @@yv syy þ 2De @@xv sxy  1b
aDe s2xy þ s2yy þ  2b @@xv  1b 2 @u
3 De ðf ðk; sÞ  1Þ  3 @x
 
sxy normal stress sxy De De 0 f ðk; sÞsxy þ De þ @@yv sxy þ De @@xv sxx þ De @u aDe þ @@xv
@y syy  1b sxx þ syy þ ð1  bÞ
@u @u
@x @y

szz normal stress szz De De 0 aDe s2  1b ðf ðk; sÞ  1Þ


f ðk; sÞszz  1b zz De
h i
Energy equation T PeqCv PeqCv j Br  He ð/Þ sxx @u þ s @v @v
yy @y þ sxy @y þ @x
@u
@x

where g0(T, p) is the melt viscosity under zero-shear-rate conditions, s* is the model constant that shows the shear stress
rate, from which the pseudoplastic behavior of the melt starts. n is the model constant which symbolizes the pseudoplastic
behavior slope of the melt as (1  n).
The WLF expression was considered to determine the viscosity of the melt under zero-shear-rate conditions
 
A1 ðT  T  Þ
g0 ¼ D1 exp  ; ð18Þ
A2 þ ðT  T Þ
e 2 þ D3  P; A1 is the model constant that shows the temperature dependence of melt glass
where T* = D2 + D3  P; A2 ¼ A
transition temperature under zero-shear-rate conditions. D1 is the model constant which registers the melt viscosity,
under zero-shear-rate conditions, at melt glass transition temperature and at atmospheric pressure. T* is the glass
transition temperature of the melt, depending on the pressure. D2 is the model constant which registers the glass tran-
sition temperature. D3 is the model constant which symbolizes the variation of the glass transition temperature of the
melt according to the pressure. A e 2 is a model parameter that depends on the type of polymer melt that has been
considered.

2.3. Level set method

The primary mathematical thought of the level set method is: the moving boundary is viewed as zero-level set of the
function /(x, t). At each time t, the value of the level set function /(x, t) is computed to determine its zero-level set, which
can track the interfaces implicitly and accurately. For the conventional injection moulding, the melt front can be defined C(t)
as zero-level set of the level set function / [13]. There are two moving interfaces C1 and C2 in GAIM, which are shown in
Fig. 1. According to [1], the level set function /(x, t) can be defined using the unified form, which makes j/(x,t)j always
the minimum distance between xand the interface C1 and C2. And the sign of /(x, t) is that / > 0 in the melt region X1
and / < 0 in the gas regions X2, respectively. That is, at any time /(x, t) satisfies the following formula
8
< minðdistanceðx; C1 ðtÞÞ; distanceðx; C2 ðtÞÞÞ
> x 2 X1
/ðx; tÞ ¼ 0 x 2 C1 ðtÞ [ C2 ðtÞ : ð19Þ
>
:
 minðdistanceðx; C1 ðtÞÞ; distanceðx; C2 ðtÞÞÞ x 2 X2

The level set equation in the Eulerian coordinates is [1]


d/ @/
¼ þ u  r/ ¼ 0: ð20Þ
dt @t

Ω2

Ω1 Γ2 Ω2
Γ1

Fig. 1. Level Set function and initial condition.


Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 261

However, after several or even one time step iteration, the level set function would not be a distance function any longer
because of numerical errors. In order to prevent the position of the zero-level set (i.e. the interfaces) from changing, we can
solve the following initial problem numerically [1], called ‘‘reinitialization”
(
@/
@s
¼ signð/0 Þð1  jr/jÞ
; ð21Þ
/ðx; 0Þ ¼ /0 ðxÞ

where
/0
signð/0 Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi: ð22Þ
/0 þ minðDx; DyÞ2
2

3. Numerical implementation

The governing equations are discretized by the finite-volume method and SIMPLEC algorithm on the non-staggered
meshes, and all quantities are stored on the same nodes (see Fig. 2). To solve the problem of the pressure–velocity decou-
pling, we use Momentum Interpolation (MI) method on the collocated grids [14]. Both of the level set function and the reini-
tialization function belong to the Hamilton–Jacobi equations, which are discretized in this paper by the fifth-order WENO
(weighted essentially non-oscillatory) scheme in space and third-order TVD R–K (total variation diminishing Runge–Kutta)
scheme in time, respectively [15–17].

3.1. Discretization by the finite-volume method

In Fig. 2, nodes for the physical quantities lie in the center of the controlling volume limited by the dashed [14,18]. The
continuum equation (1) can be expressed as the following formula by integrating in the controlling volume and get
ðue  uw ÞDy þ ðv n  v s ÞDx ¼ 0 ð23aÞ
or
X
F ¼ F e  F w þ F n  F s ¼ 0: ð23bÞ

And the discretization of the momentum Eq. (2) can be written as a generalized quantity U, that is,
aP UP ¼ aE UE þ aW UW þ aN UN þ aS US þ S/ ; ð24Þ
where S/ is the source term in the momentum equation (see Table 1), and the coefficients aE, aW, aN, aS, aP can be expressed as
aE ¼ De AðjPe jÞ þ maxðF e ; 0Þ; aW ¼ Dw AðjPw jÞ þ maxðF w ; 0Þ;
aN ¼ Dn AðjPn jÞ þ maxðF n ; 0Þ; aS ¼ Ds AðjPs jÞ þ maxðF s ; 0Þ; ð25Þ
DxDy
aP ¼ aE þ aW þ aN þ aS þ d1 ;
Dt

Fig. 2. Sketch map of the non-staggered meshes.


262 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

where Pe, Ps, Pw, Pn are the Peclet numbers on the cell faces; Fe, Fs, Fw, Fn are the cell faces flux; and De, Ds, Dw, Dn denote diffuse
derivatives on the cell faces. The form A(jPDj) can be different according to the method by which the convection terms are
discretized. Here the central difference scheme A(jPDj) = 1  0.5jPDj is adopted. And all the above coefficients are formulated
as follows:
Dy Fe
F e ¼ d1 ufe Dy; De ¼ C; Pe ¼ ;
xE  xP De
Dy Fw
F w ¼ d1 ufw Dy; Dw ¼ C ; Pw ¼ ;
xP  xW Dw
ð26Þ
Dx Fn
F n ¼ d2 v fn Dx; Dn ¼ C ; Pn ¼ ;
yN  yP Dn
Dx Fs
F s ¼ d2 v fs Dx; Ds ¼ C ; Ps ¼ :
yP  yS Ds
And the discretization of the constitutive equation can be expressed as the following generalized quantity U (where U de-
notes sxx, sxy or syy, or szz)
asP UP ¼ asE UE þ asW UW þ asN UN þ asS US þ S/ ; ð27Þ
where the coefficients asE , asW , asN , asS , asp can be formulated as follows:
asE ¼ maxðF e ; 0Þ; asW ¼ maxðF w ; 0Þ;
s
aN ¼ maxðF n ; 0Þ; asS ¼ maxðF s ; 0Þ; ð28Þ
DxDy
asP ¼ asE þ asW þ asN þ asS þ d1 ;
Dt
where Fe, Fs, Fw, Fn are the same as (26), and S/ equals S/ in Table 1.

3.2. Momentum Interpolation method on the collocated meshes

Though the discretized Eqs. (23)–(25) are non-linear systems, the numerical solutions can be obtained by iterative pro-
cess. Actually, most difficulty of all lies in how to discretize the convection term (first order derivative term). In this paper,
momentum interpolation methods on the collocated are implemented to solve the pressure–velocity and stress-velocity
couplings problems. We take the u-momentum equation as an example for treating pressure–velocity using the MI method
[18] and the XPP model as an example for coupling stress (sxy)-velocity on the E–W direction.

3.2.1. Pressure–velocity coupling


When the momentum equation is numerically solved, the under-relaxation method is often used during the iteration,
that is
P 
anb uinb þ Sui
uiP ¼ u0iP þ ai  u0iP ; ð29aÞ
aiP
where ai is the relaxation factor.
Eq. (29a) can be rewritten as
P
anb uinb þ ð1  ai Þa0iP u0iP þ Sui
uiP ¼ ; ð29bÞ
a0iP
aiP
a0iP ¼ :
ai
And separates the pressure from the source term, that is
@p
Sui ¼ S0ui  D; ð30aÞ
@xi
where D is the area of the control volume and D = DxDy.
Substituting the formula (30a) into the formula (29b), we get
"P #
anb uinb þ ð1  ai Þa0iP u0iP þ S0ui 1 @p
uiP ¼  D: ð30bÞ
a0iP a0iP @xi

When the pressure gradient term is discretized, that is (op/@x)P = (pe  pw)/(xe  xw), the pressure on the cell face is re-
quired. The quantity pe at the cell face is determined by line interpolation on the collocated grid, that is
pe ¼ fPx pE þ ð1  fPx ÞpP :
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 263

The interpolation coefficient is


xe  xp
fpx ¼ :
xE  xP
The pressure–velocity decoupling will happen possibly if the pressure is obtained by linear interpolation. For this reason,
the adjacent grid point pressures are used for the pressure discretization according to the idea of the Rhie and Chow’s [19].
For the grid point P, we have
"P # 
anb uinb þ ð1  ai Þa0iP u0iP þ S0ui 1 @p
ðuiP ÞP ¼  D : ð31aÞ
a0iP 0
aiP @xi P
P

And for the grid point adjacent to P (each of the grid points E, W, N and S, denoted by NB), we have
"P # 
anb uinb þ ð1  ai Þa0iP u0iP þ S0ui 1 @p
ðuiP ÞNB ¼  D : ð31bÞ
a0iP a0iP @xi NB
NB

^ i as follows:
For simplicity, we define u
P
anb uinb þ ð1  ai Þa0iP u0iP þ S0ui
^i ¼
u :
a0iP
And

1 @p
^ iP 
ðuiP ÞP ¼ u D ; ð32aÞ
a0iP @xi P

1 @p
ðuiP ÞNB ^ iNB  0
¼u D : ð32bÞ
aiP @xi NB
We assume that f is the cell face between the grid point P and NB, and then the following formula can be obtained
  
1 @p D pNB  pP
^if 
ðuiP Þf ¼ u D ^ if 
¼u ; ð33Þ
a0iP @xi f
a0iP f
xiNB  xiP

^ if is defined as an arithmetic mean of the adjacent quantities as follows:


where u

^if ¼ u
u ^ iP þ u^iNB ¼ u^ iP fpx þ u
^ iNB 1  fpx
         
1 1 1 1 x 1
and ¼ þ ¼ f þ 1  fpx :
a0iP f a0iP P a0iP NB a0iP P p a0iP NB

The interpolation coefficient fpx is


xif  xiP
fpx ¼
xiNB  xiP
From formula (33) we can see that the velocity on the cell face is computed by velocities and pressures on the two adja-
cent node points P and NB, and so the pressure–velocity decoupling problem is avoided successfully. However, it will occupy
much computer storage for storing the quantity u^ if . For saving the computer CPU time, the formula (32a) and (32b) can be
rewritten as follows:

1 @p
^ iP ¼ ðuiP ÞP þ
u D ; ð34aÞ
a0iP @xi P

1 @p
^ iNB
u ¼ ðuiP ÞNB þ 0 D : ð34bÞ
aiP @xi NB
And the arithmetic mean of the formula (31) is
 
1 @p 1 @p
u ^ iNB ¼ ðuiP ÞP þ ðuiP ÞNB þ
^ iP þ u D þ D : ð35Þ
a0iP @xi P
a0iP @xi NB

Substituting the formula (35) into the formula (33) and we get
  
1 @p D pNB  pP
^if 
ðuiP Þf ¼ u D ^ if 
¼u
a0iP @xif
a0iP f xiNB  xiP
   
1 @p 1 @p D pNB  pP
¼ ðuiP ÞP þ ðuiP ÞNB þ 0 D þ 0 D  0 : ð36Þ
aiP @xi P aiP @xi NB aiP f xiNB  xiP
264 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

The formula (36) can be changed as


    "     #
1 @p D pNB  pP D pNB  pP @p @p
^ if 
ðuiP Þf ¼ u D ^ if 
¼u ¼ ðuiP ÞP þ ðuiP ÞNB  0  þ : ð37Þ
a0iP @xi f
a0iP f
xiNB  xiP aiP f xiNB  xiP @xi P @xi NB

The pressure correction scheme can be obtained using the SIMPLEC-type method [11,19]. We take the pressure correction
scheme on the E–W direction as an example. According to (37), the velocity on the east face is
"P #  
anb unb þ ð1  au Þa0uP u0P þ S0ui Dy  
ufe ¼  pE  pP : ð38Þ
a0uP a0uP fe
fe

The corrected velocity on the east cell face still satisfies the formula (38), therefore
"P #  
anb ðunb þ u0nb Þ þ ð1  au Þa0uP u0P þ S0u Dy  
ufe þ u0fe ¼  pE þ p0E  pP  p0P ; ð39Þ
a0uP a0uP fe
fe

where the quantities with a prime denote the corrected values, The formula (38) is subtracted from (39), and the following is
obtained
 
Dy  0   
u0fe ¼  pE  p0P ¼ de p0E  p0P : ð40aÞ
a0uP fe

Analogously, the corrected values on the other cell faces can be expressed as
 
Dy  0   
u0fw ¼  0 p  p0W ¼ dw p0P  p0W ; ð40bÞ
auP fw P
 
Dx    
v 0fn ¼  p0N  p0P ¼ dn p0N  p0P ; ð40cÞ
a0v P fn

 
Dx  0   
v 0fs ¼  pP  p0S ¼ ds p0P  p0S : ð40dÞ
a0v P fs

Then the velocity on the cell face can be corrected as


uif ¼ uif þ u0if : ð41Þ

The pressure correction can be attained from the upper formula, that is
X
aP p0P ¼ anb p0nb þ S/ ; ð42Þ

where aE ¼ de Dy; aW ¼ dw Dy; aN ¼ dn Dx; aS ¼ ds Dx; S/ ¼  þ F fe F fw F fn  F fs .


And the velocity corrections on the grid point are as follows:
Dy  0 
u0P ¼  0
pe  p0w ; ð43aÞ
aup
Dx  0 
v 0P ¼  p  p0s : ð43bÞ
a0v p n

The velocity is given by the following formula

uiP ¼ uiP þ u0iP : ð44Þ

And the pressure is


p ¼ p þ ap p0 : ð45Þ

3.2.2. Stress–velocity coupling


By integrating the corresponding constitutive equation on the control volume, the stress sxy is discretized as
follows:
X     
@ v @u @u @ v @v @u De
asp ðsxy ÞP ¼ asnb ðsxy Þnb þ ð1  bÞ þ þ Desxy þ þ Desxx þ Desyy a sxy ðsxx þ syy Þ D:
@x @y @x @y @x @y 1b P
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 265

The under-relaxation technique is often implemented during the iteration, that is


P s !   "  0  0 !
anb ðsxy Þnb þ ð1  as Þs0xy D @v @u
ðsxy ÞP ¼ þ 0s ð1  bÞ þ
a0Ps aP P @x @y
P

 0  0 !  0  0 #
@u @v @v @u De 0 0
þ Des0xy þ 0
þDesxx 0
þ Desyy a 0
s s þ syy ; ð46aÞ
@x @y @x @y 1  b xy xx
P

P s !   "  0  0 !
anb ðsxy Þnb þ ð1  as Þs0xy D @v @u
ðsxy ÞE ¼ þ 0s ð1  bÞ þ
a0Ps aP E @x @y
E
 0  0 !  0  0 #
0 @u @v 0 @v 0 @u De 0 0 0

þ Desxy þ þDesxx þ Desyy a s s þ syy : ð46bÞ
@x @y @x @y 1  b xy xx
E

We define the following form for simplicity


P s   !
anb sxy nb þ ð1  as Þs0xy
ðs
^xy ÞP ¼ ;
a0Ps
P
P s   !
anb sxy nb þ ð1  as Þs0xy
ðs
^xy ÞE ¼ :
a0Ps
E

And then the formula (46a) and (46b) can be changed into the following form

  "  0  0 !  0  0 !  0
D @v @u @u @v @v
ðs
^xy ÞP ¼ ðsxy ÞP  ð1  bÞ þ þ Des0xy þ þDes0xx
a0Ps P @x @y @x @y @x
 0 #
0 @u De 0 0 0

þ Desyy a s s þ syy ð47aÞ
@y 1  b xy xx
P

  "  0  0 !  0  0 !
D @v @u 0 @u @v
ðs
^xy ÞE ¼ ðsxy ÞE  ð1  bÞ þ þ Des þ
a0Ps E @x @y xy
@x @y
 0  0 #
@v @u De 0 0
þ Des0xx þ Des0yy a s s þ s0yy ð47bÞ
@x @y 1  b xy xx
E

Similarly, the stress sxy on the east cell face, e, can be expressed as

 "
  0  0 !  0  0 !
D @v @u 0 @u @v
ðsxy Þfe ¼ s
^xyfe þ 0s ð1  bÞ þ þ Desxy þ
aP fe @x @y @x @y
 0  0 #
@v @u De 0 0
þ Des0xx þ Des0yy a sxy sxx þ s0yy ; ð48Þ
@x @y 1b
fe

where f denotes the face between the grid point P and the grid point NB.
Using the linear interpolation for all the terms except ð@u=@xÞ0 and ð@ v =@xÞ0 in the formula (47a) and (47b),
we get

       
D D D D
s^xyP þ s^xyE ¼ ðsxy ÞP þ ðsxy ÞE  ðax ÞP þ ðax ÞE  ðay ÞP þ ðay ÞE ;
a0Ps P a0Ps E a0Ps P a0Ps E

where the long bar denotes the linear interpolation formula, that is

Q P þ Q E ¼ 1  fpx Q P þ fpx Q E , where fpx is the interpolation factor expressed as

xe  xP
fpx ¼
xE  xP
266 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

and
"  0   #
@u   @v 0
ðax ÞP ¼ Des0xy þ ð1  bÞ þ Des0xx ;
@x @x
P
"  0   #
0 @u  0
 @v 0
ðax ÞE ¼ Desxy þ ð1  bÞ þ Desxx ;
@x @x
E
" #
  @u0  0
@v De 0 0
0 0 0
ay P ¼ ð1  bÞ þ Desyy þ Desxy a s ðs þ syy Þ ;
@y @y 1  b xy xx
"  0  0 # P
  @u @v De 0 0
ay E ¼ ðð1  bÞ þ Des0yy Þ þ Des0xy a s ðs þ s0yy Þ :
@y @y 1  b xy xx
E

Similarly, we have
  "  0   #    
D 0 @u  0
 @v 0 D   D  
s^xyfe ¼ s^xyP þ s^xyE ¼ ðsxy ÞP þ ðsxy ÞE  0s Desxy þ ð1  bÞ þ Desxx  0s ay P þ 0s ay E :
aP fe @x @x aP P aP E fe

We can obtain the following form by substituting the upper formula into the one (48)
 "   0   #
D 0 uE  uP  0
 vE  vP 0
ðsxy Þfe ¼ ðsxy ÞP þ ðsxy ÞE þ De s þ ð 1  bÞ þ Des
a0Ps fe xy
xE  xP xx
xE  xP
fe
"  0 # "  0 #
D @u D @u
 0s Des0xy þ 0s Des0xy
aP @x aP @x
P E
"  0 # "   #
D  @v D  @v 0
 0s ð1  bÞ þ Des0xx þ 0s ð1  bÞ þ Des0xx ; ð49Þ
aP @x aP @x
P E
   
where s
0xy ¼ s0xys0xy ; s0xx ¼ s0xx P þ s0xx E .
þ
P E
From (49) we can see that the stress sxy on the east cell face, e, is computed only by the stress sxy on the two adjacent node
points P and E, and hence the stress-velocity decoupling problem is avoided successfully.
The stresses on other cell faces can be deduced completely similarly, here only the expressions are given as follows:
    0
4 D uE  uP
ðsxx Þfe ¼ ðsxx ÞP þ ðsxx ÞE þ  2b þ 2Des0xx
fe 3 a0Ps xE  xP
"    0 # "    0 #
D 4 @u D 4 @u
 0s  2b þ 2Desxx
0 þ 0s  2b þ 2Desxx
0 ; ð50Þ
aP 3 @x aP 3 @x
P E

where s
0xx ¼ ðs0xx ÞP þ ðs0xx ÞE
 "    u  u 0
#
D 0 v N  uP 0 0 N P
ðsxy Þfe ¼ ðsxy ÞP þ ðsxy ÞN þ 0s Desxy
 þ ð1  bÞ þ Desyy

aP fn yN  yP yN  yP
"  0 # "  0 #
D @v D @v
 0s Des0xy þ 0s Des0xy
aP @y aP @y
P E
"   # " #
D @u
0
D @u0
 0s ð1  bÞ þ Des0yy þ 0s ð1  bÞ þ Des0yy ; ð51Þ
aP @y aP @y
P N

where s
0xy ¼ ðs0xy ÞP þ ðs0xy ÞN ; s
0yy ¼ ðs0yy ÞP þ ðs0yy ÞN .
     
4 D
0yy
vN  vP 0
ðsyy Þfn ¼ ðsyy ÞP þ ðsyy ÞN þ  2b þ 2Des
fn 3 a0Ps yN  yP
"    0 # "    0 #
D 4 @v D 4 @v
 0s  2b þ 2ksyy0 þ 0s  2b þ 2ksyy
0 ; ð52Þ
aP 3 @y aP 3 @y
P N

where s
0yy ¼ ðs0yy ÞP þ ðs0yy ÞN .

3.2.3. The numerical step of the MI method


For convenience, the present numerical step is summarized as follows:
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 267

Step 1. Give the initial velocity (u0, v0), the initial stress s0xx , s0xy , s0yy and the initial pressure p* on the grid points to obtain
the velocity u0f , v 0f and the pressure on the cell face.
Step 2. Based on the assumption or the computed velocity at the time level t  Dt on the cell face, solve the coefficients
and the source term in the constitutive equation, and obtain the stress sxx , sxy , syy on the grid points from the con-
stitutive equation, and determine the pressure on the cell face using the MI method.
Step 3. Solve the coefficients in the momentum equation based on the assumption or the computed velocity at the time
level t  Dt on the cell face, and the source term in the momentum equation based on the pressure and the stress
on the cell face.
Step 4. Determine the velocity (u*, v*) from the momentum equation.
Step 5. Solve the velocity (uf , v f Þ on the cell face using the MI method and then determine the source term and the coef-
ficients in the pressure correction equation.
Step 6. Determine the pressure correction value from the pressure correction equation.
Step 7. Solve the velocity correction value ðu0f ; v 0f Þ on the cell face, and the velocity correction value (u0 , v0 ) on the grid
point.
Step 8. Update the velocity on the cell face, the velocity and the pressure on the grid points.
Step 9. Replace the pressure p*, the velocity (u0, v0) on the grid point and (u0f , v 0f Þ on the cell face with the updated ones,
respectively.
Step 10. Stop the computation if the convergent results have been reached. Otherwise, return to Step 2 and repeat the
computations.

4. Results and analysis

4.1. Verification examples

4.1.1. The dam break problem


The dam break problem is adopted to validate the level set method for predicting free surface hydrodynamics by com-
paring the numerical result with the experimental data. The initially prescribed height of the water column schematic is
h = 1. See [20] for details on the simulation parameters. Fig. 3 shows the moving interface of dam break at different time.
Fig. 4 shows the history of water front marching along the ground surface. As shown in these figures, the numerical results
calculated by level set method agree very well with the experimental data in [20].

4.1.2. Gas-penetration in a tube completely filled with viscoelastic fluid


In this test, we choose the example in [21] to validate the level set method for tracing the free surface of viscoelastic flu-
ids, where they used the Phan–Thien and Tanner (PTT) constitutive equation model to describing the viscoelastic fluid, here-
in we adopt the XPP equation, which is similar to PTT. We examine the transient displacement of viscoelastic fluids by a gas
in a straight tube, in which the dimensionless length is 12, and the radius is 1. In order to compute conveniently, the initial
gas–melt interface C1 is set to a semicircle (see Fig. 1), and the melt front C2 is a section of line. The GAIM is simulated in a
tube with overflow cavity in the right wall, i.e. the shot size is 100% of the mould volume (the melt front C2 is just the right
mould wall). According to [21], the following parameters are adopted in the numerical simulation: De = 0.02, b = 0.5, Re = 0,
P = 8333, q = 2.0, e = 0.3. The space and time step size are 0.1 and 0.01, respectively. The program was written in Fortran 90
and run on a PC, with Pentium (R) 4 CPU at 2.80 GHz, and it took about 6869 s to complete a run.
From Fig. 5 we can see the evolution of the gas/melt interface. When the bubble has traveled dimensionless distances of 2,
6, and 10 or at elapsed dimensionless times t = 0.52 (0.54, [21]), 1.36 (1.335, [21]), and 1.79 (1.772, [21]), which are in good
agreement with the results in [21] except that the different entrance boundaries were imposed (see Fig. 5b). The finite-vol-
ume method on collocated meshes adopted in this paper is easy to implement, and for obtaining the same results it takes
much less time to complete a run than the finite element method which they used [21].

4.2. Gas-penetration in a tube partially filled viscoelastic fluid

Gas-penetration in a tube partially filled viscoelastic fluid is more common in GAIM, herein we mainly focus on this case.
The numerical simulation of GAIM is performed for a polymer, namely Polycarbonate (PC) flow within a rectangular model

Fig. 3. The moving interface of dam break at different dimensionless time: t = 0, 0.4, 2.7.
268 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

Fig. 4. The history of water front marching along the ground surface: (a) height of the water wall and (b) location of the water front.

t=0.52 t=1.36 t=1.79

(a)

(b)
Fig. 5. The bubble appearance at different dimensionless time: (a) t = 0.52, t = 1.36, t = 1.79 and (b) t = 0.548, t = 1.335, t = 1.772 [12].

(18  6) in dimensionless form (same as follows). The material properties of PC are shown in Table 2 [22]. The temperature
of polymer melt, the mould wall and the gas are 525, 323 and 513, respectively. The pressure of the gas injection is 3921.57.
The shot size is 72.2% of the mould volume. The following parameters are used in the numerical simulation: De = 1, b = 1.0/
9.0, a = 0.15, q = 2.0, e = 1.0/3.0, and the space and time step size are 0.1 and 0.005, respectively. It took about 16,453 s to
complete a run on a PC. The bubble appearances and the velocity of melt are shown in Figs. 6 and 7.
In Fig. 6 the shadow and the blank area denotes the melt flow field and the gas region, respectively. The figure gives the
process that the gas penetrates the melt, in which the bubble develops equably, and the melt is pushed forward. At first the
bubble is a semicircle (see Fig. 1). As time goes by, the bubble goes ahead meanwhile extends to both sides until it reaches a
proper width at some time, then it begins to move straight ahead instead of extending to both sides. At last, the bubble is in
the cylindrical shape. Due to the high melt viscosity, the melt front is convex with most the same curvature radius until it
reaches the right mould wall.
Fig. 7 presents the velocity vectors of the melt at different time. And the time shown in the figure represents the time
when the melt front reaches the position. From Fig. 7, it is seen that the velocity become smaller and smaller from the

Table 2
The material properties of PC.

Parameter Value Parameter Value


q (kg/m3) 1020 D1 (Pa s) 1.9  1011
Cv (J/kg K) 1700 D2 (K) 417.5
j (W/m K) 0.173 D3 (Pa/K) 0.0
n 0.574 A1 27.396
s* (Pa) 182,680 e2
A ðKÞ 51.6
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 269

(a) (b)

(c) (d)
Fig. 6. The bubble appearance and the melt front at different times for De = 1: (a) t = 400Dt; (b) t = 850Dt; (c) t = 1300Dt; and (d) t = 1375Dt.

(a) (b)
Fig. 7. Velocity vectors of the melt at different times: (a) t = 400Dt and (b) t = 850Dt.

1.5
0.5
0

0
0.5
1.5
Fig. 8. Stress sxx contours at the last time.

-0.517241
-0.298885

0.358892
0.517241

Fig. 9. Stress sxy contours at the last time.

Fig. 10. Stress syy contours at the last time.

centerline to the wall. The velocity of the melt front become larger and larger with increasing time until the mould cavity is
filled.
The stresses sxx, sxy, syy, the first normal stress difference and the second normal stress difference are shown in Figs. 8–12
at the last time step, the absolute values of which fall down uniformly from the two horizontal mould walls to the centerline.
270 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

1.5
0.5
0

0
0.5
1.5

Fig. 11. First normal stress difference at the last time.

-0.0976149
0.0660149

0.0660149
-0.0976149
Fig. 12. Second normal stress difference at the last time.

488 488

513 513 513 513

488 488

(a) (b)
488 488

513 513 513 513

488 488

(c) (d)
Fig. 13. Temperature contours of the melt at different time: (a) t = 400Dt; (b) t = 850Dt; (c) t = 1300Dt and (d) t = 1375Dt.

Fig. 14. Effect of Deborah number on the thickness fraction.


Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 271

Fig. 15. Effect of Deborah number on the shear rate near the mould wall.

Fig. 13 shows the temperature contours at different times during the gas injection stage. The melt temperature becomes
close to the gas temperature in a short time due to the convective heat transfer of the gas. In the whole gas-penetration pro-
cess the layer thickness of temperature field under melting point increases slightly.

4.2.1. Effect of the elasticity


The viscoelastic melts are simulated during the gas-penetration process. Herein we first define the fractional coverage m:

m ¼ R20  R2b =R20 , where R0 is the radius of the circular tube, and Rb is the radius of the penetrating gas bubble. Fig. 14 shows
the effect of Deborah number on the thickness fraction, from which we can see that an increase in the fractional coverage at
very low Deborah numbers, and at higher Deborah numbers the fractional coverage decreases. This phenomenon has been
observed experimentally by Henrik and Torbjorn [9]. At the case of the same length of the bubble, and at some point of the
mould wall where the steady bubble appearance exists, the different shear rate can be attained according to different Deb-
orah numbers. From Fig. 15 we can see that shear rate increases with increasing Deborah numbers, and when the Deborah

Fig. 16. Effect of Deborah number on gas-penetration time.


272 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

1 1 5 43
32 2 2
4 1
1 2
5 3 4
34 1------gas delay time: 0 5 1------gas delay time: 0
5 2------gas delay time: 100 2------gas delay time: 100
3------gas delay time: 200 3------gas delay time: 200
4------gas delay time: 300 4------gas delay time: 300
5------gas delay time: 400 5------gas delay time: 400

0 18

(a) (b)
Fig. 17. The bubble appearance and the melt front for different gas delay time: (a) t = 400Dt and (b) t = 850Dt.

1---gas delay time: 0


2---gas delay time: 100
1 2 3 4 5 3---gas delay time: 200
4---gas delay time: 300
5---gas delay time: 400

Fig. 18. The final bubble appearances for different gas delay time.

number reaches certain value the shear rate increases quickly. Because XPP fluids exhibit both shear thinning and elastic
behavior, the tendency of elasticity to increase the residual thickness is counterbalanced by shear thinning phenomena.
Deborah number can also affect the gas-penetration time. Fig. 16 illustrates the relationship between Deborah number
and the gas penetration time. Gas-penetration time decreases as increasing Deborah number, which also can be explained
by the dual effect of fluid elasticity and shear thinning [21].

4.2.2. Influence of the gas delay time on the bubble appearance


The gas delay time is the time interval between ending melt injection and starting gas penetration. Since the temperature
of the mould cavity is very low the melt will cool and solidify rapidly. So the bubble appearances and the final material prop-
erties are affected significantly by the gas delay time [23]. We take De = 10 as an example to explain the influence of the gas
delay time on the bubble appearance, where the gas delay time is chosen as 0, 50, 100, 150, 200, 250, 300 and 400, respec-
tively. For simplicity, we only give the bubble appearances and the melt fronts for gas delay time of 0, 100, 200, 300 and 400.
Fig. 17 shows the bubble appearance and the melt front for different gas delay time, from which we can see that the bub-
ble width becomes more and more narrow and the curvature of the melt front is more and more large as the gas delay time
increases.

13

12
gas penetration length

11

10

7
0 100 200 300 400 500
gas delay time
Fig. 19. The relation between gas-penetration time and gas delay time.
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 273

8.5

gas injection time 7.5

6.5
0 100 200 300 400 500
gas delay time
Fig. 20. The relation between gas-penetration length and gas delay time.

Fig. 18 shows the final bubble appearances for different gas delay time, from which we see that the bubble appearances
are affected by the gas delay time significantly.
From Figs. 19 and 20, we see that the penetration time is prolonged with increasing gas delay time, meanwhile the pen-
etration length is lengthened. This is because the melt velocity becomes higher and the gas moving resistance becomes larger
with increasing delay time. At the same time, the cooling and solidifying time increases, and the solidified layer becomes
larger near the mould cavity wall. Therefore, the gas penetration length gets longer [23].

4.2.3. Influence of the gas injection pressure on the bubble appearance


In this paper, the gas injection pressure is chosen as 980.39, 1960.78, 3921.57, 5882.35, 7843.14, 9803.92, 11764.71 and
13,725.49. Figs. 21 and 22 reveal that the penetration time becomes shorter and shorter as the injection pressure increases.
When the injection pressure reaches some values the gas-penetration length is almost the same, which agrees well with the
results in [23] where the results were obtained by the Moldflow software.

14

13

12

11
penetration time

10

3
2000 4000 6000 8000 10000 12000 14000
injection pressure
Fig. 21. The relation between gas-penetration time and gas injection pressure.
274 Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275

7.8

7.75

penetration length
7.7

7.65

7.6

7.55

7.5
2000 4000 6000 8000 10000 12000 14000
injection pressure

Fig. 22. The relation between gas-penetration length and gas injection pressure.

5. Conclusions

The 2D constitutive equation, XPP model, is introduced to simulate the viscoelastic melt in GAIM. And the corresponding
physical quantities are analyzed and discussed. The conclusions are given as follows:

1. The level set/SIMPLEC algorithms can capture accurately the gas–melt interface and the melt front by defining the inter-
faces as the zero-level set. The melt velocity becomes smaller and smaller from the centerline to the wall. Due to the con-
vective heat transfer of the gas, the melt temperature becomes close to the gas temperature in a short time. In the whole
gas injection process, the layer thickness of temperature field under melting point increases slightly.
2. The numerical results show an increase in the fractional coverage at very low Deborah numbers, while at higher Deborah
numbers the fractional coverage decreases. The shear rate increases monotonously with the Deborah numbers near the
mould wall. The tendency of elasticity to increase the residual thickness is counterbalanced by shear thinning
phenomena.
3. The shorter the gas delay time is, the shorter the gas penetration length becomes. The higher the gas injection pressure is,
the shorter the injection time becomes, and within some scope, the shorter the gas-penetration length will be.

Acknowledgements

The support from the National Natural Science Foundation of China (NSFC) (Nos. 10590353 and 10891159) and National
Basic Research Program of China (No. 2005CB321704) are fully acknowledged.

References

[1] A. Osher, R. Fedkiw, Level Set Methods and Dynamic Implicit Surfaces, Springer, 2003.
[2] M. Sun, G.F. Zhou, X.S. Zhang, J.M. Pang, 3D simulation for filling process in gas-assisted co-injection molding, China Plastics 17 (2003) 86–91.
[3] G.Q. Zheng, The Morphology Structure and Properties of the Parts Molded by Gas-assisted Injection Molding, Sichuan University, Chengdu, 2007.
[4] M.H.V. Wilco, W.M.P. Gerrit, P.T. Frank, Differential constitutive equations for polymer melts: the extended Pom–Pom model, J. Rheol. 45 (2001) 823–
843.
[5] P.C. Huzyak, K.W. Koelling, The penetration of a long bubble through a viscoelastic fluid in a tube, J. Non-Newtonian Fluid Mech. 71 (1997) 1525–1532.
[6] V. Gauri, K.W. Koelling, The motion of long bubbles through viscoelastic fluids in capillary tubes, Rheol. Acta 38 (1999) 458–470.
[7] J.S. Ro, G.M. Homsy, Viscoelastic free surface flows: thin film hydrodynamics of Hele–Shaw and dip coating flows, J. Non-Newtonian Fluid Mech. 57
(1995) 203–225.
[8] B.S. Yijie Wang, The Effect of Non-Newtonian Rheology on Gas-Assisted Injection Molding Process, The Ohio State University, 2003.
[9] K.R. Henrik, E. Torbjorn, Gas displacement of polymer melts in a cylinder: experiments and viscoelastic simulations, J. Non-Newtonian Fluid Mech. 143
(2007) 1–9.
[10] P.J. Oliveira, F.T. Pinho, G.A. Pinto, Numerical simulation of non-linear elastic flows with a general collocated finite-volume method, J. Non-Newtonian
Fluid Mech. 79 (1998) 1–43.
[11] P.J. Oliveira, F.T. Pinho, Plane contraction flows of upper convected Maxwell and Phan-Thien-Tanner Fluids as predicted by a finite-volume method, J.
Non-Newtonian Fluid Mech. 88 (1999) 63–88.
Q. Li et al. / Applied Mathematical Modelling 35 (2011) 257–275 275

[12] T. Boronat, V.J. Segui, M.A. Peydro, M.J. Reig, Influence of temperature and shear rate on the rheology and process ability of reprocessed ABS in injection
molding process, J. Mater. Process. Technol. 209 (2009) 2735–2745.
[13] S.P. Zheng, J. Ouyang, P.H. Zhang, Dynamic simulation for weldlines in thin mold with rectangle cylinder, J. Chem. Ind. Eng. 59 (2008) 232–238.
[14] M. Sussman, S. Smereka, P. Osher, A level set approach to computing solutions to incompressible two-phase flow, J. Comput. Phys. 114 (1999) 146–159.
[15] W.Q. Tao, Numerical Heat Transfer, second ed., Xi’an Jiaotong University Press, Xi’an, 2004.
[16] S.P. Zheng, J. Ouyang, L. Zhang, Z.F. Zhao, Research on a numerical scheme for capturing free front during injection molding, Polymer-Plastics
Technology and Engineering 48 (2009) 446–454.
[17] G.S. Jiang, D.P. Peng, Weighted ENO Schemes for Hamilton–Jacobi equations, SIAM J. Sci. Comput. 21 (2000) 2126–2143.
[18] D.Y. Song, The Algorithm of Collacated-Grid Finite Volume Method and Its Application of Numerical Simulation in a Contraction Flow for Viscoelastic
Fluids, East China University of Science and Technology, Shanghai, 2002.
[19] C.M. Rhie, W.L. Chow, Numerical study of the turbulent flow past an airfoil with trailing edge separation, AIAA J. 21 (1983) 1525–1532.
[20] F.J. Kelecy, R.H. Pletcher, The development of a free surface capturing approach for multidimensional free surface flows in closed containers, J. Comput.
Phys. 138 (1997) 939–980.
[21] D. Yannis, T. John, On the gas-penetration in straight tubes completely filled with a viscoelastic fluid, J. Non-Newtonian Fluid Mech. 117 (2004) 117–
139.
[22] C.T. Liu, Simulation-based Studies on Processing Optimization and Part Performance in Injection Molding, Zhengzhou University, Zhengzhou, 2003.
[23] H.S. Liu, C.W. Wei, G.F. Zhou, Research on effect of technological conditions on gas-assisted injection molding process, Modern Plastics Process. Appl.
20 (2008) 46–49.

You might also like