AIJSTPME (2013) 6(3): 59-68

Circular Die Swell Evaluation of LDPE Using Simplified Viscoelastic Model

Limtrakarn W.
Department of Mechanical Engineering, Faculty of Engineering, Thammasat University, Pathumthani,
Thailand
E-mail address: [email protected]

Pratumwal Y., Krunate J., Prahsarn C., Phompan W., Sooksomsong T., Klinsukhon W.
National Metal and Materials Technology (MTEC), Pathumthani, Thailand

Abstract
Simplified viscoelastic model based finite element method was presented to evaluate circular die swell of
LDPE. Differential viscoelastic models were firstly described for the incompressible viscoelastic flow with
isothermal extrusion, and simplified viscoelastic model (SVM) was then selected for viscoelstic component in
extra stress terms. Discrete elastic viscous stress splitting (DEVSS) method was used to stabilize elliptic term
in momentum equations. Mini–element method for hexahedral element was applied to stabilize equal order
velocity pressure interpolation. Interface–tracking approach was used to detect the position and moving of the
free surface at extrudate region. In experiment, circular die was selected to study LDPE extrusion flow. The
measured viscoelastic properties were employed as input data in LDPE flow simulation. CCD camera was
used to capture extrudate swell. Simplified viscoelstic model was then used to predict extrudate swell in three
flow rate conditions, 100, 264, and 490 mm3/s. Compared with experimental results, the die swell predicted
by SVM were 1.37%, 2.37%, and 5.5% different, for the three flow rate conditions, respectively. The results
showed that simplified viscoelastic model could be employed to predict circular die swell.

Keywords: Die swell, Simplified viscoelastic model, LDPE

1 Introduction
Die swell is a common behavior observed in the and in-plant trials can be reduced by CFD simulation.
polymer extrusion process. Accurate prediction of an Extrusion die is design by several polymer flow
extrudate dimension is important to die design and simulation software such as Ansys/Polyflow,
product quality. Extrusion is used to produce fibers, OpenFoam, Flow2000, Dieflow, HyperXtrude, etc.
films, tubing, and a variety of profiles. Extrusion [2] CFD simulation has great potential to predict
process is a very complex task and requires important interior details of the extrusion process
detailed knowledge of material characteristics, flow such as velocity, shear stress, pressure, and
phenomena, and extrusion processing experience [1]. temperature fields in the region of interest, which is
The appropriate design of an extrusion die and its not possible to do experimentally.
process parameters are extremely important to In extrusion processes, the swelling of the extrudate
achieve the desired shape and accurate dimensions of is controlled by a physical mechanism: relaxation of
the extruded product. Die design for a new product the polymeric liquids originating from the recoiling
and cost reduction is developed on the basis of and relaxing of the stretched polymer chains as they
previous experience and experimentation. Trials and exit from the die. In order to accurately simulate
errors as well as experiment tasks have been the polymer extrusion processes, the computational
common practice to design the properly extrusion die model must account for this mechanism and capture
with expensive cost and time. The proper design of the effect of non–linear polymer or viscoelasticity. Non–
extrusion dies can be facilitated by computational linear viscoelastic models have been developed, including
fluid dynamics (CFD) analysis. Costly experiments the Kay–Bersteinighly branched polymKearsly–Zappa

59
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

(KBKZ) integral model and the exponent Phan–Thien– 2.2 Constitutive equations
Tanner (EPTT) differential model. These two models
Elastic viscous split stress method is applied to
give good predictions of the behavior of polymeric
decompose the stress tensor in 2 components: pure
fluids under rheometric and extrusion flows [3–7].
The die swell of low density polyethylene (LDPE) viscous stress (σs) and viscoelastic stress (σv).
melt was simulated [6] in a long die using four types
of damping functions of the KBKZ model to examine
σ = σs + σv (3)
the cause of the exaggerated swelling ratios. The
pom–pom model is a molecular constitutive model
for modeling the flow behavior of polymeric fluids The pure viscous stress is represented for generalized
such as the LDPE. The double convected pom–pom Newtonian fluids and given as
(DCPP) model [8] was proposed to predict the
behavior of polymeric fluids under rheometric and
complex flows. The results were in good agreement σs = 2ηD (4)
with the experimental data [9]. Simplified
viscoelastic model (SVM) [10] was presented and
included the effect of the first normal stress where η is shear viscosity coefficient and can be
difference to account for one of the key contributions expressed by the Carreau Yasuda model as following,
of polymer viscoelasticity to the die–swell of
polymeric melts. Computationally, it is very cost–
effective in comparison with the other non–linear η = η∞ + (η0 – η∞)(1+(λ γ )a )(n-1)/a (5)
viscoelastic models [11], and gives reasonably good
predictions of complex viscoelastic fluid flows. In the
present work, prediction of the SVM on the die swell where η0 and η∞ are the shear viscosities at the limits
of a well–characterized commercial LDPE through a of zero and infinite shear rate, respectively, λ is the
long die had been described and shown to be in good relaxation time, n describes the slope of the power-
agreement with the result of the other advanced law region, and γ is the local shear rate.
constitutive models.

2 Theoretical formulation ∇V + (∇V)T

D is the rate of deformation tensor =
The equations for the incompressible viscoelastic 2
flow with isothermal extrusion are represented by the (6)
conservation of mass and momentum, as well as a
constitutive equation.
where ∇V and (∇V)T are the velocity gradient and
2.1 Conservation equations transpose of the velocity gradient.
The equations of conservation of mass and
momentum can be expressed as In this paper, the viscoelastic components were
described in the Simplified Viscoelastic model
∇⋅V = 0 (1) (SVM). The SVM or light viscoelastic model is an
extension of the generalized Newtonian fluid models
and and accounts for viscoelastic effects in a simple form.
The viscoelastic stress tensor for the SVM is given as
= −∇p + ∇σ⋅
DV
ρ (2)
Dt
σ xx τ xy 0 ψµ (χ )χ η( γ ) γ 0
where ρ is the density, V is the velocity, p is the σv =  τ yx 0 0 =  η( γ ) γ 0 0 (7)
pressure, t is time, and σ is the stress tensor and is
given by a constitutive equation.
 0 0 0  0 0 0

60
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

The shear stress is related to the shear rate 1.8×104

dependence of shear viscosity η( γ ) and defined in

Complex Modulus (G*) [Pas]

the usual way of the generalized Newtonian fluid
1.6×104
models. The first normal stress σxx is given by
ψµ( χ ) χ , which consist of a weighting factor ψ, a
normal viscosity µ( χ ), and a viscoelastic variable χ . 1.4×104
The normal viscosity µ( χ ) can be defined in the
same way as the shear viscosity η( γ ) by replacing χ 1.2×104
with γ . In this paper Bird – Carreau model is applied
for µ.
1.0×104
0.01 0.1 0.01
η = η∞ + (η0 – η∞)(1+(λ γ ) )2 (n-1)/2
(8) Shear strain
Figure 1: Complex modulus (G*) data of the LDPE
sample obtained from the stress sweep test at 180oC.
The viscoelastic variable χ is governed by a
transport equation with a shear rate dependent 106
relaxation time λ. G’
G”
Modulus G’, G” (Pa)

Dχ 104
λ( χ ) + χ = γ (9)
Dt

102
Equation (9) is based on the idea of a memory effect
from viscoelastic fluids. The weighting factor ψ
controls the swelling intensity against flow rate and
the relaxation time controls the development of the 1
extrudate diameter along the jet. 0.01 0.1 1 10 100 1000
Angular Frequency (rad/s)
3 Experimental methods Figure 2: The experimental G’ and G” data of the
A commercial LDPE (PETLIN LD C150Y) supplied LDPE sample obtained from the dynamic frequency
by PETLIN (Malaysia) SDN BHD, was used in this sweep with strain equal to 5% (0.05) at 180oC.
study. Its melt flow rate is 5 dg/min (ISO 1133).
Some specifications of the resin are as follows:
3.2 Nonlinear Rheology
density = 0.921 g/cm3 (ISO 1183A), tensile strength
at yield = 15 MPa (ISO R527-3), and ultimate Steady shear and step shear experiments were
elongation = 610% (ISO R527-3). performed using Gemini 200HR Nano rotational
rheometer of Malvem-Bohlin Instruments. Steady
shear experiments were performed at 180oC using
3.1 Linear Rheology
parallel plate geometry (25 mm diameter, gap 1 mm)
Storage modulus G’(ω) and loss modulus G”(ω) were by ramping the shear rate from 0.0001 to 1000 s−1.
obtained from small amplitude oscillatory shear The first normal stress difference was also obtained
experiments, performed using a 25mm parallel plate from normal force measurements. Step shear
geometry on Gemini 200HR Nano rotational experiments were performed using 25 mm. parallel
rheometer. Data was obtained at 180oC. The plate geometry at 180oC. For higher shear rates (> 3
equilibration time was set to 5 minutes in order to s−1), the transient viscosity data was not reliable
maintain thermal equilibration before data acquisition since the specimen was forced out from the gap
began. between the parallel plates.

61
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

Figure 3: The steady shear viscosity data of the

LDPE sample obtained from the steady rate
sweep at the temperature of 180oC.
Figure 5: Temperature sensor.
3.3 Experimental setup
The extrusion flow experiments were performed on a
twin-screw extruder (Labtech Engineering, LTE-20-
32). In this work, the effects of test volumetric flow
rate on the extrudate swell of LDPE were considered
for the test temperature range of 160-180oC. The
screw speed (rpm) / feeder speed (rpm) were set to be
25/5, 70/7 and 120/10 to obtain volumetric flow rate
100, 264, and 490 mm3/s, respectively.

Figure 6: Circular Die.

3.4 Extrudate swell measurement

The extudate swell diameters of the polymer melts
were directly measured by determining the size of the
extrudate diameter in the fully swollen (2 in. away
from the die exit) [12]. The pictures were taken from
high speed camera (Nikon D90) couple with
TAMRON SP AF 90mm macro. The experiments
were repeated for three times to assure the obtained
Figure 4: Twin screw machine. results.

62
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

where Q is the flow rate, Um is the mean velocity, A

is the cross sectional area and r is the radius of the
die. For a multimode spectrum, the λm can be
estimated from [9]

M
∑ λ2iG i
λm = i =1 (12)
M
∑ λiG i
i =1

where i represents the ith relaxation mode, M is the

Figure 7: Extrudate swell of LDPE, test temperature total number of different modes, λi (s) is the ith linear
range 160-180oC and Screw speed/ feeder speed25/5 relaxation time and Gi (Pa) is the ith plateau modulus.
(volumetric flow rate 99.96 mm3/s).
4.1 DEVSS method [13]
4 Computational model and technique To restrain and even eliminate the instability from the
A finite element method based on discrete elastic loss of characteristics of the governing equations in
viscous stress splitting (DEVSS) and a streamline symbolizing elliptic behavior, the elastic viscous
upwinding (SU) scheme, was used to solve the stress splitting (EVSS) method was proposed by
governing equations. The DEVSS scheme can be Perera and Walters [14] to retain an elliptic term and
readily applied to all the viscoelastic constitutive consequently the viscous contribution in the
equations [10] as the constitutive equation need not momentum equation. But to achieve this end the
be rewritten. The implementation of the DEVSS/SU constitutive equation has to be modified. Moreover,
scheme permits the use of a lower order interpolation the convected derivative of the rate of strain tensor
scheme for the viscoelastic unknown variables emerges, which requires calculation of a second-
without generating an ill-posed problem. This can order derivative of the velocity field. So, it is
significantly save computational costs. For the SVM difficult to apply this approach to complex
and Bird – Carreau model, Picard iteration is used to constitutive equations.
solve the governing equations with a AMF direct The DEVSS method is proposed as a discrete version
solver as it gives a better convergence for a fluid with of the elastic viscous stress splitting method. In this
the power-law index < 0.7. method a stabilizing elliptic term is introduced into
For the numerical results presented below, the the discretized momentum equation in the absence of
Weissenberg (Wi) number is defined as a purely viscous contribution or as the viscous
contribution is negligible in comparison with the
viscoelastic contribution and can be numerically
Wi = γ λ m (10) implemented in a convenient manner. To construct
the DEVSS method, an auxiliary variable D as a
where λm is the mean relaxation time and γ is the discrete counterpart of the rate of deformation tensor
and an independent variable is introduced and
nominal shear rate and is defined as determined in terms of satisfaction of

Um Q 4Q
γ = = = (11) D –D = 0 (13)
r Ar πr 3

63
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

in a weighted average form, i.e. (W, D – D) = 0, of a grid is proposed. An elastic response corresponds
with W denoting a suitable weighting function. to the angular deformation of the corners, of the
Further, taking the divergence of Eq. (13) and adding diagonal of elements, and of the side of elements.
– 2 ηe ∇·( D – D) into the momentum conservation Energy may be associated with the deformation of the
grid, and the remeshing is the result of the
equation with the use of Eq. (2) result in minimization of that energy: one solves an elastic
problem.

= −∇p + ∇σ⋅
DV
ρ – 2 ηe ∇·( D – D) (14) 4.3 Geometry and boundary conditions
Dt
The computational geometry consists of a quarter of a
cylinder with the dimensionless length z* = 40 and
4.2 Free surface condition the radius r1* = r2* = 1. Figure 5 shows the geometry
Interface – tracking approach is used to detect the used in the simulation and the computational mesh
position and moving of the free surface at extrudate with the boundary conditions. Fully developed
region. The kinematic conditions are approached for velocity and stress profiles are specified at the inlet
moving interface. The large deformation appears on BD1. A vanishing force condition is implemented at
free surface interface. The evolution is implemented the exit BD2. Care is taken to ensure that the
to obtain solution convergence. extrudate is long enough to satisfy this condition at
The following three boundary conditions are defined BD2. A no slip condition is applied at the wall of die
on a free surface. BD3. BD4 is the free surface, which is a priori
unknown. Surface tension is neglected; hence a
vanishing force condition is applied at BD4. Next to
vr nr + vz nz = 0 (15) this, the kinematic condition, v⋅ n = 0 is solved,
where n is the normal vector to the free surface. This
implies that there is no flow through the free surface.
1 For BD5 and BD6, a symmetry condition is specified
1 
tr nr + tz nz = S  +  = 0 (16) by imposing zero values for the normal velocity and
 ρ1 ρ 2  the tangential force.

tr nr – tz nz = 0 (17)

with variables specification of radial velocity (vr),

axial velocity (vz), components of the unit normal to
the free surface (nr, nz), surface force normal to the
surface (tr, tz), principal radii of curvature (r1, r2) and
surface tension coefficient (S). When S equals zero, it
represents free slip condition. While S equal one, it is
equivalent to a no–slip boundary condition. An
evolution method (typical iterative approach) for
modeling a free surface flow involves enforcing the
boundary conditions of equations (16) and (17).
In order to avoid over distortion of elements, a
remeshing technique is used to propagate the motion
of the free surface into the mesh. The mesh is Figure 8: 3–D extrusion flow geometry in a quarter
considered as a deforming elastic grid. A remeshing of a cylinder with origin at r1 = 0, r2 = 0 and z = 0
based on the minimization of the deformation energy with different boundaries.

64
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

5 Results and discussion The swelling ratios obtained from finite element
model were found in good agreement with
Finite element model with 21,952 nodes and 19,206
experimental results. The largest difference of the SW
elements was generated to simulate extrudate. This is
in the range of Wi used was 2%. Similar results were
necessary for simulations involving the discretization
obtained using the SVM in the same range of Wi.
of non-linear viscoelastic models using the combined
DEVSS/SU scheme [9]. The detail of mesh was v*
given in Figure 9. The velocity field obtained from 2.0
SVM was shown in Figures 10a. The dimensionless 1.8
pressure distribution is given as P* = P/pa, where P is 1.6
the pressure difference along the flow direction and 1.4 M1W1
pa is the atmospheric pressure. The dimensionless 1.2 M1W2
velocity is v* = v/Um, where Um is the mean velocity 1.0 M1W3
0.8 M2W1
of the flow. With increasing Weissenberg (Wi), the M2W2
shear-thinning velocity profile is more pronounced. 0.6 M2W3
The swelling ratio, SW could be measured with 0.4 M3W1
respect to either r1 or r2 and the same results are 0.2 M3W2
M3W3
obtained from the two directions in the present work.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
*
z
(a) Velocity profile across the r2–axis
(r2–axis, r1* = 0 and z* = 4)
2.5

1.5
Sw

(a) Mesh 1 (M1) 1 m1

m2
0.5
m3
0
W1 W2 W3
(b) Plot of SW against Wi

Figure 10: The simulation results obtained from

finite element model using SVM: (a) velocity profile
across the r2–axis (r2–axis, r1* = 0 and z* = 4) and
(b) Mesh 2 (M2) (b) a plot of SW against Wi.

5.1 Extrudate swell in viscoelastic fluids

Simulations were carried out for viscoelastic models.
This is to validate the accuracy of the predictive
power of SVM in modeling the die-swell behavior of
LDPE fluid in extrusion flow, through comparison
with Wi from W1, W2, and W3. Figure 11 showed the
contour plots of the shear rate for the three flow
conditions. The plots of the pressure distribution along
(c) Mesh 3 (M3) the flow direction at W1, W2, W3 (Figures 12a–c)
showed that there were also in good agreements
Figure 9: Mesh for simulation of extrusion flow. amongst the three models.

65
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

a) Flow rate 100 mm3/s (a) A plot of pressure distribution along the flow
direction (z-axis, r1* = 0 and r2* = 0) at W1.

(b) Flow rate 264 mm3/s (b) Contour plots of pressure distribution along
the flow direction at W2.

(c) Contour plots of pressure distribution along

(c) Flow rate 490 mm3/s the flow direction at W3.
Figure 12: The simulation results obtained from
Figure 11: The Contour profile images obtained by three flow rate conditions: (a), (b), and (c) a plot of
using SVM at different flow rates (a) 100 mm3/s; pressure distribution along the flow direction (z-axis,
(b) 264 mm3/s and (c) 490 mm3/s. r1* = 0 and r2* = 0) at W1, W2, and W3, respectively.

66
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

Also in Figure 12, the plot of the SW against Wi for [2] Pepliński, K., Mozer, A., 2011, Design of
the three conditions showed that there are reasonable Extrusion Die for Plastic Profile using
agreements in the prediction of the extrudate swelling ANSYS/Polyflow software, Journal of Polish
ratio. Taking the experiments as a reference, the CIMAC, 6(3): 221–226.
percentage relative difference in the prediction of the [3] Luo, X-L., Tanner, R.I., 1986a. A streamline
SW obtained from SVM at W1, W2, and W3 were element scheme for solving viscoelastic flow
1.525%, 0.354%, and 8.017%, respectively. Although problems, part I: differential constitutive
SVM is not a full viscoelastic constitutive model, it models., Non-Newtonian Fluid Mechanics, 21:
still could give a good prediction on the die-swell 179–199.
ratio of polymer extrusion flow from along straight [4] Luo, X-L., Tanner, R.I., 1986b. A streamline
die. This was not so surprising because the element scheme for solving viscoelastic flow
viscoelastic effects were predominately from the problems, part II: integral constitutive models.,
contributions of the non-linear shear viscosity and Non-Newtonian Fluid Mechanics, 22: 61–89.
pressure difference in these particular types of [5] Goublomme, A., Draily, B., Crotchet, M.J.,
extrusion flows. 1992. Numerical prediction of extrudate swell
of a high density polyethylene., Non-Newtonian
6 Conclusions Fluid Mechanics, 44: 171–195.
[6] Huang, S.X., Lu, C.J., 2006. Stress relaxation
Computer simulations of the extrusion flow process characterisitics and extrudate swell of the
had been carried out, using both viscoelastic and non- IUPAC-LDPE Melt., Non-Newtonian Fluid
Newtonian models. Through comparisons of Mechanics, 136: 147–156.
viscoelastic model and experiments, the accuracy of [7] Mitsoulis, E., 2007. Past and new developments
the SVM in predicting the die-swell behavior of in the numerical simulation of polymer flows
polymeric fluid in extrusion flow had been verified with the K-BKZ model. In: ESAFORM
and was found in reasonably good agreement with Conference on Material Forming.
experimental results. The effects of viscoelasticity of [8] Clemeur, N., Rutgers, R.P.G., Debbaut, B.,
a polymer melt in extrusion processes had also been 2003. On the evaluation of some differential
quantified. A surprising cooperative effect of the formulation for the pom–pom constitutive
viscoelasticity had been identified. Although the model., Rheological Acta, 42: 15.
SVM captures some essential physics of polymer [9] Clemeur, N., Rutgers, R.P.G., Debbaut, B.,
extrusion processes, it should be noted that it does not 2004. Numerical simulation of abrupt
account for the full viscoelastic effect of a polymeric contraction flows using the double convected
fluid, such as the second normal stress difference and pom–pom model., Journal of Non-Newtonian
extensional viscosity, which seem to have no Fluid Mechanics, 117 (2–3): 193–209.
significant effects on the particular flow studied here. [10] Fluent, 2008. Polyflow 3.12 User’s Guide. F.
Inc. 3.11, 664.
Acknowledgments [11] Debbaut, B., Marchal, T., 2008. Numerical
simulation of extrusion process and die design
The authors are pleased to acknowledge the Thailand
for industrial profile, using multimode pompom
Commission on Higher Education of Thailand (the
model. Plastics, Rubber and Composites, 37:
National Research University Project), the National
142–150.
Science and Technology (NRCT), the Thailand
Research Fund (TRF), Royal Project Foundation, [12] Brydson, J.A. 1981. Flow Properties of Polymer
Thammasat University, and the National Metal and Melts, 2nd ed, London: Gorage Godwin.
Materials Technology Center (MTEC) for supporting [13] Han, X., Li, X., 2008. A mixed finite element
this research work. scheme for viscoelastic flows with XPP model.
Acta Mech Sin, 24: 671–680.
[14] M.G.N. Perera and K. Walters, 1977, Long-
References
range memory effects in flows involving abrupt
[1] Kostic, M. M., Reifschneider, L. G., 2006, changes in geometry, Part I. Flows associated
Design of Extrusion Dies, Encyclopedia of with L-shaped and T-shaped geometries,
Chemical Processing, 633-649. J. Non-Newtonian Fluid Mech, 2 (2): 49–81.

67
 Limtrakarn W. et al. / AIJSTPME (2013) 6(3): 59-68

[15] Pichelin, E., Coupez, T. 1998. Finite element polyethylene using the Wagner constitutive
solution of the 3D mold filling problem for equation, J. Non-Newtonian Fluid Mech., 69:
viscous incompressible fluid, Comput., Methods 113-136.
Appl. Mech. Engrg, 163: 359-371. [19] Guenette, R., Fortin, M., 1995. A new mixed finite
[16] Yu, H., C. Kietzmann, P. Cook, S. Xu, F. Costa, element method for computing viscoelastic flows,
and P. K. Kennedy. 2004. A New Method for J. Non-Newtonian Fluid Mech., 60: 27-52.
Simulation of Injection Molding, SPE ANTEC [20] Ganvir,V., Lele, A., Thaokar, R., Tautham,
Proceedings, Chicago: 641-644. B.P., 2009. Prediction of extrudate swell in
[17] Liang, J.Z., Yang, J., Tang, C.Y., 2010. Die- polymer melt extrusion using an Arbitrary
swell behavior of PP/Al(OH)3/Mg(OH)2 flame Lagrangian Eulerian (ALE) based finite element
retardant composite melts. Polymer testing, 1-5. method, J. Non-Newtonian Fluid Mech.,
[18] Fulchiron, R., Revenu, P., Kim, B.S., Carrot, C., 156: 21-28.
Guillet, J., 1997. Extrudate swell and isothermal [21] Vaddiruju, S.R., 2005. M.D. Thesis, Northern
melt spinning analysis of linear low density illinois university, USA.

68