Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105

Contents lists available at ScienceDirect

Journal of Non-Newtonian Fluid Mechanics

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

Extrudate swell of a high-density polyethylene melt: II. Modeling using

integral and differential constitutive equations
Vinod Kumar Konagantia, Mahmoud Ansaria, Evan Mitsoulisb, Savvas G. Hatzikiriakosa,∗
a
Department of Chemical and Biological Engineering, The University of British Columbia, Vancouver V6T 1Z3, BC, Canada
b
School of Mining Engineering and Metallurgy, National Technical University of Athens, Zografou 157 80, Athens, Greece

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

Article history: The extrudate swell phenomenon of a high-molecular-weight HDPE is modeled using the following viscoelas-
Received 26 February 2015 tic constitutive equations: the multi-mode Kaye-Bernstein–Kearsley–Zapas (K-BKZ) integral model with the
Revised 24 June 2015
Wagner and the Papanastasiou–Scriven–Macosko (PSM) damping functions, and multi-mode Phan-Thien–
Accepted 15 July 2015
Tanner (PTT), Giesekus and Double Convected Pom-Pom (DCPP) differential models. The high-molecular-
Available online 30 July 2015
weight high density polyethylene (HDPE) is rheologically characterized and the experimental data have been
Keywords: used to fit the parameters of the constitutive models. Experimental results on extrudate swell are compared
Extrudate swell with the numerical predictions. The simulation results showed that the integral K-BKZ model highly over-
Capillary flow predicts and the differential PTT, Giesekus and DCPP models slightly under-predict the experimental mea-
K-BKZ model surements. When the capillary reservoir is excluded from the simulations, the predictions of the integral
Phan-Thien Tanner (PTT) model models are significantly reduced close to the experimental data, while those of the differential models re-
Papanastasiou–Scriven–Macosko (PSM)
main practically unaffected. Although both integral and differential models used represent the rheological
damping function
Integral and differential constitutive
data equally well, an explanation of the dramatically different predictions of extrudate swell by the integral
equations and differential models remains elusive.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction to decouple the computation of free surface position. Different as-

Extrudate swell is of interest to several polymer processing ap-
plications such as profile extrusion, blow molding, film blowing and
fiber spinning [1]. Although it has been studied extensively both ex-
perimentally and numerically, it is not fully understood [2–9]. Pre-
cise prediction of extrudate swell is very crucial for the optimization
of die design, material and operating parameters. The early numer-
ical studies on the extrudate swell problem were mainly concerned
with Newtonian fluids and showed good agreement with the experi-
mental results [10,11]. However, no such systematic studies exist for
viscoelastic fluids. The advent of advanced numerical techniques and
constitutive models made extrudate swell simulation possible for vis-
coelastic fluids, such as polymer melts and/or solutions [12–15].
Extrudate swell simulation of a highly viscoelastic IUPAC-LDPE
melt was successfully demonstrated by Luo and Tanner [13,14,16]
using the integral K-BKZ model for long and short capillary dies.
Luo and Tanner [13,14] have used the streamline finite element
method (SFEM) with a particle tracking scheme to simulate the
swell problem. This was later modified by Luo and Mitsoulis [6]
∗ sagging/gravity, and non-isothermal effects as well as the model itself
Corresponding author. Tel.: +1 604 822 3107; fax: +1 604 822 6003.
E-mail address: savvas.hatzi@ubc.ca (S.G. Hatzikiriakos).
pects of Luo and Mitsoulis’s numerical method have been modified by
Goublomme et al. [17] and Goublomme and Crochet [18] in order to
simulate the flow at high shear rates (high Weissenberg numbers). In
another study, Barakos and Mitsoulis [19] have simulated extrudate
swell of the IUPAC-LDPE polymer melt using capillary dies of various
aspect ratios, ranging from orifice dies (i.e., L/D = 0) to infinite length
dies (i.e., L/D → ∞), using the standard Galerkin weak formulation for
velocities and pressure by also including the viscoelastic stress tensor
as a body force in the momentum balance equations. The above de-
scribed method seems to overcome the difficulties associated with
SFEM and accounts for the effects of upstream contraction and recir-
culation on extrudate swell. However, the predictions were observed
to be in agreement with experiments for short dies (0  L/D  2) only
at low apparent shear rates and overpredicted the actual values for
long dies. Furthermore, the extrudate swell of a HDPE melt exhibit-
ing significant amount of slip has been studied by Ansari et al. [9],
both experimentally and numerically, using capillary dies of differ-
ent L/D ratios. It was observed that the model overpredicts extrudate
swell even after applying the slip boundary condition. The overpre-
diction was attributed to effects such as flow-induced crystallization,
[9,19,20].

http://dx.doi.org/10.1016/j.jnnfm.2015.07.005
0377-0257/© 2015 Elsevier B.V. All rights reserved.
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105
In the case of extrudate swell simulation using differential mod-
els, most of the early efforts were mainly concerned with the high
Weissenberg number (Wi) problem [12]. However, this problem was
solved by using numerical schemes best suited for hyperbolic equa-
tions [12,21–23]. Subsequently, Elastic-Viscous Stress Split (EVSS)
[24] and/or Discontinuous Elastic-Viscous Stress Split (DEVSS) [25]
formulations on extra-stress tensor were introduced to obtain con-
vergent and numerically accurate solutions. Ganvir et al. [7,26] have
used an arbitrary Lagrangian–Eulerian (ALE) based finite element
method and simulated transient extrudate swell of LLDPE and HDPE
polymer melts. The extrudate swell is simulated for planar and ax-
isymmetric dies with abrupt contraction. The results showed good
agreement with reported experimental results on polymer melts of
moderate elasticity (swell ratio of up to 1.4), however apparent shear
rates as high as those used in practical applications could not be
reached [7,26].
No or limited studies exist on extrudate swell using both inte-
gral and differential rheological models on the same polymer melt
of high elasticity, typically used in blow molding applications. More-
over, there is no systematic study of the extrudate swell of a viscoelas-
tic model where flow model predictions are compared with experi-
mental results for various operating conditions relevant to polymer
processing and for various geometrical parameters of capillary dies,
such as diameter, D, length-to-diameter ratio, L/D, contraction angle,
2α , and the presence of the reservoir. The latter plays a significant
role in the numerical predictions and their consistency with exper-
imental results on extrudate swell as the present study attempts to
demonstrate.
In this study, a high-molecular-weight HDPE is rheologically
characterized using multimode integral (Kaye-Bernstein–Kearsley–
Zapas referred to as the K-BKZ) and differential Phan-Thien–
Tanner (referred to as the PTT), Giesekus and Double Convected
POM-POM (referred to as the DCPP) models. The simulation re-
sults using ANSYS POLYFLOW® are compared with experimen-
tal extrudate swell measurements studying the effects of ap-
parent shear rate, temperature, L/D, 2α , and the presence of
reservoir. In addition, various rheological/material functions of the
polymer melt under different types of deformation are computed
using both K-BKZ and PTT constitutive equations to understand
the origin of significant differences on the extrudate swell predic-
tions between these two different classes of models (integral versus
differential).
2. Experimental
2.1. Rheological measurements
The viscoelastic properties of a high-molecular-weight HDPE used
in this work have been determined using a stress/strain controlled
rotational rheometer (Anton Paar MCR501), and the details are dis-
cussed in Ansari et al. [27]. Briefly, the results from three types of
tests have been used to fit the rheological parameters of the con-
stitutive equations used; small-amplitude oscillatory shear (SAOS),
start-up of steady shear, and uniaxial extension. The frequency sweep
(SAOS) tests were carried out at different temperatures ranging from
160° to 220 °C with the 25-mm parallel disk geometry and a gap
of about 1 mm. Further, the time-temperature superposition (TTS)
principle was applied to obtain the master curves at the reference
temperature of 200 °C. The steady shear experiments have also been
done using the 25-mm parallel disk geometry with a gap of 1 mm at
the reference temperature of 200 °C, at various shear rates ranging
from 0.05 to 1 s−1 . The polymer was also characterized in uniaxial
extension using the SER-2 universal testing platform fixture suitable
for the Anton Paar MCR501 rheometer. The SER measurements were
95
performed at different Hencky strain rates ranging from 0.05 to 5 s−1
also at 200 °C.
2.2. Extrudate swell measurements
A pressure-driven Instron capillary rheometer of constant pis-
ton speed has been used to study the extrudate swell and process-
ing/slip behavior of the present polymer melt at 200 °C. More details
on the swell measurements can be found elsewhere [27]. Capillary
extrusion experiments using dies of various length-to-diameter ra-
tios (L/D = 5, 16, and 33, with D = 0.79 mm) and diameters (such as
D = 0.43, 0.79, and 2.11 mm with L/D = 16) were carried out to ap-
ply all the necessary corrections such as Bagley, Rabinowitsch, and
slip (Mooney analysis) to the raw data [28,29]. The extrudate swell
®
measurements were carried out by using a Keyence LS-7030 M op-
tical micrometer placed at the die exit. The experimental set-up is de-
vised such that the optical micrometers can be moved up and down
to capture the complete extrudate swell profile, allowing us to mea-
sure the time dependency of the extrudate swell phenomenon [27].
Moreover, a set of radiation heaters keeps the ambient temperature
the same as that of the die and therefore these measurements are
isothermal.
3. Mathematical modeling
3.1. Governing equations
The governing equations for the flow of incompressible fluids such
as polymer melts under isothermal, creeping and steady flow condi-
tions are given as [30,31]:
∇ ·u=0 (1)
−∇ p + ∇ · τ = 0 (2)
where u is the velocity vector, p is the pressure, τ is the extra stress
tensor of the polymer melt. For viscoelastic fluids, such as polymer
melts, constitutive equations that relate the stress tensor in terms
of velocity/velocity gradients are required to complete the system of
equations.
3.2. Constitutive equations
In the present work, mainly the integral K-BKZ (with Wagner and
PSM damping functions) [32,33] and the differential Phan-Thien Tan-
ner (PTT) [34] models were used for the extrudate swell simulations.
In addition, the differential Giesekus and DCPP models were also used
for few computations to evaluate their ability in predicting extrudate
swell in comparison with the K-BKZ and PTT constitutive equations.
A detailed review comparing various constitutive equations for rep-
resenting the rheological data of polymer melts has been reported by
Larson [35], which provides useful insights on selecting appropriate
viscoelastic rheological models.
The K-BKZ model proposed by Papanastasiou et al. [32] and mod-
ified by Luo and Tanner [16] used in the simulations, is given by:

1 t 
N
gi (−(t−t  )/λi )   
 
τ= e h(IC−1 , IC ) × C−1
t (t ) + θ Ct (t ) dt (3)
1−θ −∞ n=1 λi
where λi and gi are the relaxation times and relaxation moduli,
respectively, N is the number of relaxation modes, IC and IC −1
are the first invariants of the Cauchy–Green tensor Ct , and its in-
verse Finger strain tensor Ct −1 , θ is a material constant given by
N2 /N1 = θ /(1 − θ ), N1 (≡τ 11 − τ 22 ) and N2 (≡τ 22 − τ 33 ) are first and
96 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105


second normal stress differences, respectively, with the value of θ be- αi λi ∇
tween −0.2 and −0.1 observed from experimental measurements. I+ τi · τi + λi τi = 2ηi D (11)
ηi
The Cauchy–Green tensor Ct is calculated by using the deforma-
tion gradient tensor Ft as follows [14,16]: I is the unit tensor and α i is a material constant for each mode, which
Ct = FTt Ft (4) indicates the anisotropic mobility stimulated by flow characteristics.
The differential POM-POM model (based on a molecular theory)
The deformation gradient tensor Ft is obtained from the following best suited for branched polymers, is also used in the present study
relation [14,16]: for comparison. The original differential form of POM-POM model
was developed by McLeish and Larson [45]. The proposed model has
DFt (s)
= −Lt (s)Ft (s) (5) undergone several modifications and changes to make it appropri-
Ds ate for numerical schemes/formulations and to represent viscoelas-

Ft (s) =I (6) tic flow under complex deformations. The formulation developed by
s=0
Clemeur et al. [46] referred to as Double Convected Pom-Pom (DCPP)
where Lt (s) is the velocity gradient tensor, s = t − t is a dummy vari- model is used in the present paper. For a multi-mode DCPP model, the
able used to integrate along the path of the particle and I is the unit total extra stress tensor τ is defined as the sum of individual compo-
tensor. Further, the inverse Finger strain tensor Ct −1 is simply de- nents and the extra stress tensor contribution by each mode is given
termined by inverting Ct . More particulars on the determination of in terms of an orientation tensor S, and a stretching scalar  (state
Cauchy–Green tensor Ct and Finger strain tensor Ct −1 are available in variables) as shown by the equation below:
the literature [14,16].
Gi  2
The function h is a strain-dependent memory (or damping) func- τi = 3 S − I (12)
tion. We have considered two damping functions, namely the Wag- 1 − ζi
ner and the PSM functions. The Wagner damping function is given by
where Gi and ζ i are shear modulus and nonlinear material constants
[33]:
 of each mode, respectively. The state variables, S (orientation tensor)
 0.5
and  (stretch variable) are obtained from the following differential
h(IC−1 , ICt ) = exp −n β ICt−1 + (1 − β)ICt − 3 (7)
t
equations:
where n and β are nonlinear model constants to be determined from 
   
shear and elongational flow data, respectively. The PSM [32] damping ζi ∇ ζi ∇ 1 I
λi 1− S+ S + λi (1 − ζ )[2D : S]S + S− =0 (13)
function is given by: 2 2 2 3
α D 2(−1)
h(IC−1 , IC ) = (8) λsi − λsi (∇ v : S) + ( − 1)e qi = 0 (14)
(α − 3) + β IC−1 + (1 − β)IC Dt
where α and β are model constants to be determined from shear and where λi and λsi are the relaxation times related to the orientation
elongational flow data, respectively. and stretching mechanisms, respectively and q represents the num-
The PTT model has also extensively been used in the simulation ber of dangling arms at the ends of the pom-pom molecule.
of complex viscoelastic flows and was found to be one of the sim-
plest differential models to represent both the shear and elonga-
tional properties of polymer solutions in entry flows [34]. Several re- 3.3. Boundary conditions
searches have shown that the multi-mode PTT model is an excellent
candidate for simulation of flow of polymer melts in practical applica- The conservation equations (Eqs. (1) and (2)) combined with the
tions [34,36–38]. For a spectrum of N relaxation modes (six are used constitutive equation (either the integral or the differential mod-
in the present simulation), the viscoelastic extra-stress tensor τ is els) need to be solved along with suitable boundary conditions. The
obtained as the sum of N individual contributions τ i : present problem is solved by applying well-known boundary condi-
tions such as described in the literature [1,9,38,47]. The boundary

N
τ= τi, (9) conditions are: (i) fully developed velocity profile at the inlet, (ii)
i=1 the radial velocity component and the shear stresses are zero along
the symmetry axis, (ii) slip at the solid walls that has the form, Vs =
Each individual contribution τ i is described by the PTT constitutive
−β sl τ w b [9,48], (iv) zero surface tractions and zero transverse veloc-
equation [39,40]:
  
  ity along the extrudate (domain) exit, and (v) zero surface tractions
εi λi ξi ∇ ξi
and no flow normal to the surface, n · u = 0, at the free surface.
exp tr(τ i ) τ i + λi 1− τ i + τ i = 2 ηi D
ηi 2 2
The rate of deformation tensor D is defined as: D = (∇ v + ∇ vT )/2,
where v is the velocity gradient tensor. The symbols  and
over
the stress tensor (τ ) represent the upper and lower convective time
derivatives, while ξ i and ε i are non-dimensional shear and elonga-
tional material parameters, respectively.
The Giesekus model is also one of the most realistic differential
viscoelastic models and is best suited for shear flows, which can pre-
dict shear thinning viscosity and also normal stresses in all directions
[41]. Many studies have been reported in the literature validating
its ability to represent viscoelastic polymer melts [42–44]. Like PTT
model, a multi-mode Giesekus model is used in the present work and
the viscoelastic extra-stress tensor τ is obtained as the sum of N indi-
vidual contributions (Eq. (9)). The individual contribution of the extra
stress tensor for the Giesekus model is defined by:
(10)
3.4. Method of solution
The finite element method (FEM) based flow simulations use
u–v–p formulation (primitive variable approach) for integral mod-
els, such as the K-BKZ Wagner/PSM as in viscous flows, and
EVSS (mixed variable approach) [24] or DEVSS formulation (en-
hanced mixed variable approach) [25] for differential models, such
as the multi-mode PTT, Giesekus and DCPP models [1,42,60], com-
bined with special streamline-upwind/Petrov–Galerkin (SUPG) for-
mulation for both integral and differential constitutive equations
[19,38]. Further, the integral models include an appropriate nu-
merical integration method (such as Gauss-Laguerre quadrature
method) to determine the stress components along the stream-
lines [1,9,49]. The above numerical schemes have been used in the
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105 97

numerical package ANSYS POLYFLOW® , which has been licensed for 106
use in the present work. Independent numerical results using the in- HDPE MFI 5.72g/10min
tegral method outlined by Luo and Tanner [13,14] and Luo and Mit- T=200°C

Dynamic moduli, G', G'' (Pa)

soulis [6] have confirmed the numerical results obtained from ANSYS
POLYFLOW® .
105
A typical FEM grid for a capillary die with length-to-diameter ra-
tio of L/D = 16 is presented in Fig. 1. The flow domain shows a 12:1
abrupt contraction with the diameter of the reservoir (barrel) (Db )
being 9.52 mm and the diameter of the die being 0.79 mm (D). The G'
length-to-diameter ratio is 16, long enough to ensure fully developed 104 G''
K-BKZ
flow conditions. The extrudate length (Lext ) considered in the simu-
PTT
lations is 25 mm (Lext /D = 31.64), necessary to capture the memory Giesekus
effects associated with the viscoelastic flow simulations. The flow DCPP
domain is meshed using a mapped meshing technique and the ar-
103
rangement of meshes/grids is such that the grids become denser to- 10-2 10-1 100 101 102 103
wards the entry singularity, i.e. near the entrance of the die, and Frequency, ω (rad/s)
larger grids are used away from the entry singularity. Similarly, a finer
grid or more mesh elements are used towards the die exit to bet- Fig. 2. The master curves of storage (G ) and loss (G ) moduli and the correspond-
ing fitting using a six-mode Maxwell spectrum at the reference temperature of
ter capture the exit singularity. The mesh is chosen based on the ex-
Tref = 200 °C (see Tables 1 and 2).
perience from earlier studies (both viscous and viscoelastic) of our
group [9,48]. The grid consists of 1900 elements and a mesh four
times denser is also used by dividing each element into four sub-
N
elements to check the mesh independency of the numerical results. gk λ2k
The mesh independency is checked such that the differences be- λ = k=1 , (15)
N
k=1 gk λk
tween the overall pressures and swell ratios from both the meshes
are <1%.


N
4. Results and discussion η0 = gk λk . (16)
k=1
4.1. Rheological characterization

The values of these parameters are λ = 21.46 s, η0 = 215,196 Pa s, in-

The master curves of the dynamic moduli G (ω) and G (ω) of the
HDPE resin (with a Melt Flow Index, MFI = 5.72 g/10 min) at the ref-
erence temperature of 200 °C along with the Maxwell model predic-
tions (using a six-mode relaxation spectrum) are shown in Fig. 2.
The fitted Maxwell parameters, relaxation times, λi , and the relax-
ation moduli, gi , are listed in Table 1 [9]. Furthermore, the average
relaxation time, λ, and the zero shear viscosity, η0 , can be calculated
from:
Fig. 1. A typical finite element grid for the simulation of the 12:1 abrupt circular con-
traction with L/D = 16, Lext /D = 31.64, D = 0.79 mm and 2α = 180°.
dicating a highly viscoelastic polymer melt.
The nonlinear model parameters n and β of the Wagner damp-
ing function, and the material parameters α and β of the PSM damp-
ing function are obtained by fitting the shear and elongational rhe-
ological data [16,32,33], such as the start-up of steady stress growth
coefficient (ηs + ) and uniaxial stress growth coefficient (ηE + ). A non-
zero second normal stress difference is found to decrease the swell
only about 2–7% (depending on shear rate) and thus θ is not con-
sidered in the present simulations. This observation is in agreement
with previous studies [9,16,18,47]. The corresponding model fittings
along with the experimental rheological data are shown in Figs. 3
and 4. The values of the relaxation spectrum and the material con-
stants of the HDPE polymer melt at 200 °C for both the Wagner
and PSM damping functions of the K-BKZ equations are listed in
Table 1.
The relaxation spectrum and the nonlinear material constants for
the differential models (six-mode PTT, Giesekus and DCPP models)
were obtained by fitting the corresponding constitutive equations si-
multaneously to the dynamic moduli, start-up of steady shear, and
uniaxial elongation data. The detailed procedure for the fitting of PTT
[50], Giesekus [41,42] and DCPP [46] model parameters are available
in the literature and the calculated model parameters are listed in
Tables 2–4, respectively. The nonlinear materials parameters ξ i of PTT
model are set to zero as the effect of N2 is not significant on extrudate
swell and also to reduce the number of material parameters and thus
the complexity of calculations associated with the model. The cor-
responding model fittings of differential models PTT, Giesekus and
DCPP are also included in Figs. 2–4. As it can be seen from Figs. 2–4,
the rheological data (linear and nonlinear) are represented well by all
constitutive equations.
Tables 1 and 2 further include the relaxation spectrums at tem-
peratures 160 °C and 180 °C that are necessary to study the effect of
extrusion temperature on extrudate swell by using the integral K-BKZ
and differential PTT models.
98 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105

Table 1
Relaxation spectrum at temperatures 160 °C, 180 °C, and 200 °C for HDPE melt with K-BKZ model parameters (Wagner and PSM damping functions, θ = 0).

Mode # Relaxation spectrum K-BKZ parameters

160 °C 180 °C 200 °C

gi (Pa) λi (s) gi (Pa) λi (s) gi (Pa) λi (s) Wagner

1 387,808 0.0017 387,808 0.0012 387,808 0.00086 n β

2 185,307 0.015 185,307 0.011 185,307 0.0075 0.2145 0.9
3 93,338 0.108 93,338 0.074 93,338 0.0548
4 37,766 0.796 37,766 0.547 37,766 0.403 PSM

5 12,934 5.92 12,934 4.07 12,934 2.99 α β

6 5025 60.76 5025 41.75 5025 30.78 7.6 0.6

determine the slip velocity of this polymer as a function of wall shear

S (Pa.s)

HDPE MFI 5.72g/10min stress using the well-established Mooney analysis.

T=200°C 0.05 s-1
The calculated slip velocity from the Mooney analysis is plotted as
105
a function of the wall shear stress in Fig. 6. The slip data is fitted with a
nonlinear slip law, Vs = −β sl τ w b , with β sl = 1565.7 mm/s/MPab and
Shear stress growth coefficient,

0.5 s-1
b = 3.4, indicating considerable slip for the present polymer melt.

1 s-1
4.3. Flow pressure drop comparison
4
10
K-BKZ (Wagner)
K-BKZ (PSM) The pressure drop predictions obtained from the FEM flow sim-
PTT ulations using all the constitutive equations, i.e., both integral and
Giesekus differential models, are compared with the experimental data of cap-
DCPP
illary extrusion in Fig. 7. All the constitutive equations predict the to-
103 tal pressure drop of capillary extrusion reasonably well (within ±10%
10-1 100 101 102
error for most cases).
Time (s)

Fig. 3. Experimental data (symbols) and model fits (lines) of the shear stress growth
coefficient ηS+ , using both integral (K-BKZ/Wagner and K-BKZ/PSM) and differential
4.4. Extrudate swell simulations
(PTT, Giesekus and DCPP) constitutive models.
Extrudate swell simulations were carried out to study its depen-
dence on the capillary die geometric parameters, such as length-to-
diameter ratio, L/D, die entrance angle (2a), presence of reservoir, and
on the operating conditions, such as the apparent shear rate, γ̇A , and
temperature. The presence of reservoir is important as it will show
the effect of memory at least for the case of integral models.

4.5. Integral versus differential constitutive models

Fig. 8 shows the variation of extrudate swell ratio, BD (defined as

Fig. 4. Experimental data (symbols) and model fits (lines) of uniaxial stress growth co-
efficient ηE+ , using both integral (K-BKZ/Wagner and K-BKZ/PSM) and differential (PTT,
Giesekus and DCPP) constitutive models.
4.2. Slip modeling
The Bagley-corrected flow curves of the HDPE polymer melt at
200 °C, using capillary dies of three different diameters D = 0.43,
0.79 and 2.11 mm (all having the same L/D = 16) along with LVE
data are presented in Fig. 5. The diameter dependence of the flow
curve and the significant deviation of the capillary data from the
LVE curve (no-slip) indicate the presence of slip behavior for the
present polymer melt [48]. The data in Fig. 5 can be used to
the ratio of the extrudate diameter, Dext , to the die diameter, D) with
apparent shear rate, simulated using the aforementioned constitutive
equations, with reservoir, for the die with L/D = 16 and D = 0.79 mm.
The values of the swell ratio, BD , reported in Fig. 8 as well as the cor-
responding predicted values by the various models are those at a dis-
tance 25 mm (Lext /D = 31.64) from the die exit. At this distance the
extrudate swell has been well developed and takes values close to
about 90% of its ultimate value as can be seen below when the com-
plete extrudate swell profiles are presented.
The swell ratio predicted using both integral (K-BKZ/Wagner and
K-BKZ/PSM) and differential (PTT, Giesekus and DCPP) models in-
creases with an increase of the apparent shear rate conforming to
the general trend of experimental data. Similar results were re-
ported in the literature [7,9,11,15,18,26,45]. However, the integral K-
BKZ model significantly overpredicts the extrudate swell (≈250% at
high shear rates), while the differential models slightly underpre-
dict the results. In the case of the integral K-BKZ model, due to
the memory effects, the melt attempts to attain its original shape
in the reservoir as it exits the die [9,17]. This is intriguing in a
sense that while both integral and differential constitutive models
represent the rheological data equally well, their predictions of ex-
trudate swell are so distinctly different. This can be attributed to
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105 99

Table 2
Relaxation spectrum at temperatures 160 °C, 180 °C, and 200 °C for HDPE melt with PTT model parameters.

Mode # Relaxation spectrum PTT parameters

160 °C 180 °C 200 °C

ηi (Pa s) λi (s) ηi (Pa s) λi (s) ηi (Pa s) λi (s) εi ξi

1 658.26 0.0017 452.32 0.0012 336.66 0.0009 8.29 0

2 2736.25 0.015 1880.2 0.011 1395.33 0.0075 1.88 0
3 10101.7 0.108 6941.31 0.074 5473.37 0.0548 9.42 0
4 30052.3 0.796 20650.3 0.547 13400.3 0.403 0.21 0
5 76606.2 5.92 52639.6 4.07 36232.6 2.99 6.57 0
6 305,387 60.76 209,845 41.75 155,795 30.78 0.23 0

Table 3
The relaxation spectrum along with Giesekus model parameters at 200 °C. 101 HDPE MFI 5.72g/10min
T=200 oC
Mode # ηi (Pa s) λi (s) αi

Slip Velocity, Vs (mm/s)

1 348.90 0.0009 0.01
2 1371.21 0.0075 0.39 100
3 5484.34 0.055 0.48
4 15264.5 0.403 0.45
5 37988.7 2.99 0.35
6 154,673 30.78 0.40 D = 0.43 mm
10-1
D = 0.79 mm
D = 2.11 mm
Mooney
Table 4 Vs(mm/s)=1565.7 [ τ (MPa)] 3.4
w
The relaxation spectrum along with DCPP model parameters at 200 °C. 10-2
0.06 0.08 0.1 0.15 0.2 0.25
Mode # Gi (Pa) λi (s) λsi (Pa) qi ζi
Wall Shear Stress, τ (MPa)
w
1 372,942 0.0009 0.0006 2 0.62
2 184,712 0.0075 0.0055 2 0.02 Fig. 6. The slip velocity (Vs ) as a function of wall shear stress (τ w ) for the HDPE melt
3 93,454 0.055 0.0372 7 0.04 at 200 °C. The continuous line is a fit to the experimental data indicating a slope of 3.4.
4 37,761 0.403 0.01 8 0.08
5 12,862 2.98 0.113 11 0.02
6 5007 30.77 24.81 14 0.82
50
HDPE MFI 5.72g/10min
T=200°C, L/D=16
Total pressure drop, ΔP(MPa)

0.50
HDPE MFI 5.72g/10min
o
T=200 C
0.30
Wall Shear Stress (MPa)

10
Complex Modulus or

0.20

0.10

0.07
0.05
0.03
100 101 102
Apparent Shear Rate (s -1 ) or Frequency (rad/s)
Fig. 5. The Bagley-corrected flow curves using three different capillary dies of diame-
ters D = 0.43, 0.79, and 2.11 mm (with L/D = 16) along with the LVE plotted as a flow
curve (no-slip). The diameter dependence of the flow curves implies the presence of
slip, and the Mooney method can be used to determine slip as a function of the wall
shear stress.
the high elastic energy/memory effect associated with the integral
K-BKZ model compared to that of differential models. Similar behav-
ior was observed in several earlier studies [9,15,17,18]. On the other
hand all differential models used resulted into similar predictions. To
understand this problem, a more detailed analysis of the predictions
of material functions by the two constitutive equations should be per-
formed in order to identify its origin (see below).
LVE
D = 0.43 mm
1
D = 0.79 mm
4 10 100 200
D = 2.11 mm .
Apparent shear rate, γA (s-1)
103
Fig. 7. Experimental data (solid symbols) and model predictions of the total pressure
drop for flow in a capillary having L/D = 16 and D = 0.79 mm, using the K-BKZ/Wagner,
K-BKZ/PSM, PTT, Giesekus, DCPP models for the HDPE polymer melt at 200 °C.
Extrudate swell simulations were carried without reservoir us-
ing the integral K-BKZ (Wagner and PSM) and differential PTT mod-
els and the results are presented in Fig. 9. When the reservoir is ex-
cluded from the simulations, the predictions of the PTT, Giesekus and
DCPP models only slightly change compared to those with the reser-
voir. On the other hand, for the K-BKZ case, excluding the reservoir
causes significant decrease to the predictions. In fact the K-BKZ pre-
dictions agree reasonably well with the experimental results now.
Therefore, memory effects are significant in flow simulations using
integral models.
100 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105

6 it is exiting the capillary die. The experimental extrudate swell pro-

K-BKZ (Wagner) files are compared with the simulated profiles using the K-BKZ (PSM)
and PTT models at low and high apparent shear rates (5 s−1 and
K-BKZ (PSM)
PTT
5
Giesekus 160 s−1 ), in Fig. 10(a) and (b), respectively. The predicted extrudate
Extrudate Swell, BD

DCPP
Experimental swell profiles keep developing along the length of the extrudate (i.e.,
4 distance from the die exit) in a similar fashion with the corresponding
experimental ones. It is also seen that the use of the integral K-BKZ
(PSM) model without the presence of the reservoir results into good
3 prediction of the whole extrudate profile. The extrudate swell pre-
HDPE MFI 5.72g/10min, T=200°C
dicted using the PTT model reaches its ultimate value very close to
D=0.79 mm, L/D=16
the die exit itself, indicating that the memory effects fade away fast.
2
Similar swell profiles were reported in the literature using a multi-
mode PTT model [38,51,54].
1
4 10 100 200
. -1 4.7. Pressure and stress profiles comparison
Apparent shear rate, A (s )

Fig. 8. Experimental data (solid symbols) and model predictions of extrudate swell The model predictions of pressure p, shear stress T12 , and first
ratios predicted by using integral K-BKZ models with Wagner and PSM damping func- normal stress difference N1, w along the wall and free-surface and
tions and differential PTT, Giesekus and DCPP models with reservoir. first normal stress difference N1, sym, along the axis of symmetry
using K-BKZ (PSM) and PTT models, at an apparent shear rate of
26 s−1 , are depicted in Fig. 11(a)–(d), respectively. These profiles
6
K-BKZ (Wagner) are in agreement with the general trends reported in earlier stud-
K-BKZ (PSM) ies [26,38,55,56]. As can been from Fig. 11(a)–(d), all these material
K-BKZ (Wagner) without reservoir
5 K-BKZ (PSM) without reservoir function profiles along the flow domain predicted by integral K-BKZ
PTT (PSM) and differential PTT model are fairly similar. Some small dif-
Extrudate Swell, BD

PTT without reservoir

Expt ferences exist near the die entry and exit singularities (especially for
4 first normal stress difference along the symmetry axis) which can-
not explain the significant differences in the extrudate swell pre-
dictions of the two models. Also, the first normal stress difference
3 HDPE MFI 5.72g/10min, T=200°C
N1, sym along the axis of symmetry, predicted using the differential PTT
D=0.79 mm, L/D=16
model relaxes slightly faster than that predicted by integral K-BKZ
2 model (Fig. 11(d)). These differences could be attributed to the vari-
ations in the numerical schemes/formulations associated with these
two types of models.
1
4 10 100 200
. -1
Apparent shear rate, A (s ) 4.8. The effect of temperature

Fig. 9. Experimental data (solid symbols) and model predictions of extrudate swell The effect of extrusion temperature (isothermal case where
ratios predicted by using K-BKZ models with Wagner and PSM damping functions and
the die and ambient temperature are the same) on extrudate
the PTT model with and without reservoir.
swell is also studied by using both integral K-BKZ (PSM) (with no
reservoir) and differential PTT models; for the capillary die with
L/D = 16, D = 0.79 mm. The corresponding simulation results us-
4.6. Swell profiles comparison ing K-BKZ (PSM) and PTT models along with experimental mea-
surements are presented in Fig. 12(a) and (b), respectively. The
As discussed above, a single value of BD cannot characterize simulation results (with no reservoir) show that the swell ratio
completely the extrudate behavior of the polymer as its profile decreases with increasing temperature of the polymer melt, consis-
is continuously been developed in the downstream direction once tent with experimental observations. The elasticity or modulus of the

2.6 5
K-BKZ (PSM) (a) (b)
2.4 K-BKZ (PSM) without reservoir
PTT 4
Extrudate swell, BD
Extrudate swell, BD

2.2
Experimental
2.0 K-BKZ (PSM)
K-BKZ (PSM) without reservoir
1.8 3
PTT
1.6 Experimental

1.4 2

1.2
1.0 1
0 5 10 15 20 25 0 5 10 15 20 25
Distance from die exit or Extrudate length (mm) Distance from die exit or Extrudate length (mm)

Fig. 10. Comparison of experimental extrudate profiles with simulations for the capillary die having L/D = 16 and D = 0.79 mm at (a) a low shear rate, 5 s−1 and (b) a high shear
rate, 160 s−1 .
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105 101

14 0.4
(a) K-BKZ (PSM) (b)
12 PTT 0.3

(MPa)
10

Pressure (MPa)
0.2

Die Entry

Die Entry
Die Exit

12
8

Shear stress,
0.1
6

Die Exit
0.0
4

2 -0.1

0 -0.2
0 10 20 30 40 50 0 10 20 30 40 50
Distance from flow domain inlet (mm) Distance from flow domain inlet (mm)
1 normal stress difference, N 1,W (MPa)

1st normal stress difference, N1,sym (MPa)

1.5
(c) 0.8
(d)
1.0 0.6
Die Entry

0.5 0.4

0.2
0.0
Die Exit

0.0

Die Entry

Die Exit
-0.5
-0.2

-1.0 -0.4
0 10 20 30 40 50 0 10 20 30 40 50
st

Distance from flow domain inlet (mm) Distance from flow domain inlet (mm)

Fig. 11. Comparison of (a) Pressure p (b) Shear stress τ 12 (c) First normal stress differences N1, W along the wall and free-surface and (d) First normal stress differences N1, sym , along
the axis of symmetry, using integral K-BKZ (PSM) and differential PTT models, at an apparent shear rate of γ̇A = 26 s−1 .

2.6 2.6
-1

2.4
(a) L/D= 16, D=0.79 mm
2.4
(b) 11 s PTT
-1
26 s PTT
Extrudate Swell, BD

-1
Extrudate Swell, BD

64 s PTT
2.2 2.2 -1
11 s Expt
-1
26 s Expt
2.0 2.0 -1
64 s Expt

1.8 1.8

1.6 11 s-1 K-BKZ (PSM) 1.6

26 s-1 K-BKZ (PSM)
1.4 64 s-1 K-BKZ (PSM) 1.4
11 s-1 Expt
1.2 26 s-1 Expt 1.2
64 s-1 Expt L/D= 16, D=0.79 mm
1.0 1.0
150 160 170 180 190 200 210 150 160 170 180 190 200 210

Temperature (oC) Temperature ( oC)

Fig. 12. Variation of extrudate swell with temperature simulated using (a) K-BKZ (PSM) and (b) PTT models compared with experimental data (solid symbols).

polymer melt increases and the corresponding recoverable strains increasing the L/D ratio, in qualitative agreement with the exper-
and the extrudate swell increases with decrease of temperature [52]. imental observations [9,15,16,19]. Flow in a longer die makes the
The decrease in swell with temperature could be further attributed memory effects fade away in the reservoir and causes a decrease
to the change in velocity profile. The slip velocity increases with tem- in the extrudate swell. Excluding the upstream reservoir elimi-
perature of the polymer melt [9,48], and thus reduces the extrudate nates the effect of die length on extrudate swell as can be seen in
swell to a certain degree. Fig. 13(b).
A similar trend was observed for extrudate swell simulations us-
4.9. The effect of die length ing the PTT model as well. The results are plotted in Fig. 14. Here
there is no need to do the simulations with and without reservoir
The effect of length-to-diameter ratio (L/D) on extrudate swell as the results for both cases are similar. The comparison is consis-
simulations using the K-BKZ model with the PSM damping func- tent qualitatively, although there are significant quantitative differ-
tion is studied using three capillary dies of different L/D ra- ences (20% for the long die to 50% for the short die). The decrease
tios (L/D = 5, 16, and 33 all with diameter D = 0.79 mm). The re- of the extrudate swell ratio with length of the die (seen also in
sults are compared with experimental measurements in Fig. 13(a). Fig. 13(a) is related to the fading memory of the viscoelastic polymer
The simulation results show that the swell ratio decreases with melt [19].
102 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105

6 3.0
(a) L/D = 5 Expt
L/D =16 Expt
D = 0.79mm (b) L/D = 5 Expt D = 0.79mm
L/D =16 Expt
L/D = 33 Expt
5 L/D =33 Expt

Extrudate swell, BD

Extrudate swell, B D
L/D = 5 K-BKZ (PSM)
L/D = 5 K-BKZ (PSM)
L/D =16 K-BKZ (PSM) 2.5 L/D =16 K-BKZ (PSM)
L/D =33 K-BKZ (PSM)
L/D =33 K-BKZ (PSM)
4
2.0
3

1.5
2

1 1.0
4 10 100 200 4 10 100 200
. -1 .
Apparent shear rate, A (s ) Apparent shear rate, A (s-1 )

Fig. 13. Extrudate swell simulations using the K-BKZ (PSM) model for three different dies with L/D = 5, 16, and 33 and their comparison with experimental results (a) including
the reservoir in the flow simulation (b) excluding the reservoir in the flow simulation.

2.4 effects) with die entrance angle (entry flow becomes more elonga-
D = 0.79mm tional). Similar results were reported in the literature related to an-
2.2 nular dies with converging or diverging die heads [6,20,53]. However,
there are no or limited studies available on the effect of die entrance
2.0 angle on extrudate swell simulation in circular dies, using both inte-
Extrudate swell, B D

gral and differential models. The significant elastic energy/memory

1.8
effect pertaining to integral K-BKZ model in comparison with the dif-
ferential PTT model is demonstrated again with these results. While
1.6
the predictions are qualitatively consistent with the experimental ob-
servations, there are significant quantitative differences.
1.4 L/D = 5 Expt
L/D = 16 Expt
L/D = 33 Expt
1.2 L/D = 5 PTT
5. Evaluation of material functions
L/D = 16 PTT
L/D = 33 PTT
1.0 As mentioned earlier, to explain the significant differences
10 100 between the extrudate swell predictions of the integral K-BKZ
. -1
Apparent shear rate, (s ) (Wagner or PSM) with those of the differential PTT model, several
material functions were calculated, particularly those not measured
Fig. 14. Extrudate swell simulations using the 6-mode PTT model for three different
dies with L/D = 5, 16, and 33 and their comparison with experimental results. Similar
experimentally. The material functions in steady shear, such as the
simulation results with or without the reservoir. shear viscosity (η), and the first normal stress difference coefficient
(1 = N1 /γ̇ 2 ) are presented in Fig. 17(a) and (b), respectively. All
three constitutive equations, namely the K-BKZ with Wagner, K-BKZ
It would be very interesting to calculate the die length necessary with PSM, and PTT models, represent the variation of shear viscosity
for complete relaxation of memory effects associated with the inte- and first normal stress difference coefficient with shear rate nearly
gral K-BKZ model, when the reservoir is included in the flow sim- identically.
ulations. For this purpose, the extrudate swell is calculated in cap- In addition, the start-up and relaxation/decay material func-
illary dies having various length-to-diameter ratios in the range of tions are also calculated using the three constitutive equa-
5 < L/D < 500 (D = 0.79 mm) at three different shear rates using the tions. The transient material functions calculated and plotted in
K-BKZ (PSM) model. The results are presented in Fig. 15. An L/D ra-
tio as high as 100 is needed for nearly complete relaxation of the
memory effects. In this case, the swell ratio reaches the values ob- 6
tained for the case ‘without reservoir’. The L/D value needed for com- D=0.79 mm
plete relaxation becomes higher as the apparent shear rate is in- 11 s-1
5 26 s-1
creased. These results show the strong memory effects associated
Extrudate Swell, BD

with the use of integral constitutive equations such as the K-BKZ 64 s-1
(PSM) model. 4

4.10. The effect of entrance angle 3

The effect of entrance angle (2a) on extrudate swell simula-

2
tion is also studied by using K-BKZ PSM and multi-mode PTT mod-
els, for a short capillary die having D = 1.27 mm and L/D = 4, at
different apparent shear rates. The corresponding simulation re- 1
0 100 200 300 400
sults are presented in Fig. 16(a) and (b), using K-BKZ (PSM) and
PTT models, respectively, compared with the experimental observa- L/D ratio
tions. The swell ratio increases slightly with the die entrance an- Fig. 15. Variation of extrudate swell with length to diameter ratio, L/D, using the inte-
gle for both cases, consistent with the experimental measurements. gral K-BKZ (PSM) model with reservoir at three different shear rates 11 s−1 , 26 s−1 , and
This could be due to the increase of normal stresses (elasticity 64 s−1 .
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105 103

6
(a) L/D = 4, D=1.27 mm
5 s-1 Expt
5

Extrudate Swell, BD
-1
11 s Expt
-1
26 s Expt
4 40 s-1 Expt
64 s-1 Expt
5 s-1
3
11 s-1
26 s-1
2 40 s
-1

64 s-1
1
0 50 100 150 200
Entrance angle, 2 °)

3.0
(b) L/D = 4, D=1.27 mm
-1
5s
-1
11 s
2.5
Extrudate swell, B D

-1
26 s
-1
40 s
-1
64 s
2.0 -1
5 s - Expt
-1
11 s - Expt
-1
1.5 26 s - Expt
-1
40 s - Expt
-1
64 s - Expt
1.0
0 50 100 150 200
Entrance angle, 2a °)

Fig. 16. Variation of extrudate swell with die entrance angle using (a) K-BKZ (PSM) and (b) PTT models, at different apparent shear rates.

106 108
First Normal stress difference

(a) 107 (b)

(Pa.s)

106
105
(Pa.s)

105
104
Shear Viscosity,

104
103
coefficient,

102
103 PTT 101 PTT
K-BKZ (PSM) K-BKZ (PSM)
100 K-BKZ (Wagner)
K-BKZ (Wagner)
102 10-1
10-3 10-2 10-1 100 101 102 103 104 10-3 10-2 10-1 100 101 102 103 104
-1 -1
Shear rate (s ) Shear rate (s )

Fig. 17. Variation of (a) shear viscosity (η) and (b) first normal stress difference coefficient (1 ) with shear rate predicted by using the K-BKZ constitutive model with Wagner and
PSM damping functions and the PTT constitutive model.

Fig. 18(a)–(d) are the shear stress growth coefficient (ηS+ ), and the using the integral K-BKZ model with a simple exponential Wagner
shear stress decay coefficient (ηS− ), the first normal stress difference damping function, allows simple analytical expressions as described
coefficient (1+ ), and first normal stress difference decay coefficient by Laun [57]. However, when the PSM damping function [32] is used,
(1− ). As clearly shown in Fig. 18(a)–(d), the three constitutive equa- the material functions cannot be expressed in simple analytical forms
tions predict similar results, and thus their distinctly different behav- and thus the steady recoverable deformations (both shear and elon-
ior in extrudate swell prediction cannot be explained. gational) were determined numerically by using a MATLAB program.
The elastic recoil/recoverable strain predictions (in both shear The recoverable shear and elongational strains using the PTT model
and elongation) are of great importance in polymer processing phe- can also be determined by the methods described in Tanner and
nomena, such as extrudate swell in extrusion and spinning, dis- Nasseri [58] and Tanner et al. [59].
tortion in injection molding, and shrinkage in film blowing [50]. The steady recoverable shear and elongational strains using the
The recovery is also closely associated with memory effects and three constitutive equations calculated (K-BKZ Wagner, K-BKZ PSM
thus it seems appropriate to compare the predictions from the and multi-mode PTT models) are depicted in Fig. 19(a) and (b), re-
two models in an attempt to understand the discrepancy between spectively. All three constitutive equations predict similar steady
the swell predictions of integral and differential rheological mod- elastic strains/recoverable deformations. Elastic recovery (and
els. The steady recoverable strains during shear and elongation, stress relaxation) upon a change in flow kinetics, considered to
104 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105

106 106
(a) 0.01 s-1
(b)
0.01 s-1

coefficient, S- (Pa.s)
S (Pa.s)
Shear stress growth 105

Shear stress decay

105 0.1 s-1
0.1 s-1
-1
104 1 s
1 s-1
coeffcient,

104
10 s-1
10 s-1 103

103
102
10-2 10-1 100 101 102 10-2 10-1 100 101 102
Time (s) Time (s)

First Normal stress difference

0.01 s -1 0.01 s -1
First Normal stress difference

(Pa.s)
(Pa.s)

107 (c) 10 7 (d)

K-BKZ (Wagner) 0.1 s -1
106 K-BKZ (PSM) 10 6
PTT 0.1 s -1
1 s -1
105 10 5

decay coefficient,
growth coefficient,

1 s-1
104 10 4 10 s -1

103 10 3
10 s -1
102 10 2

101 10 1
10-2 10-1 100 101 102 10 -2 10 -1 10 0 10 1 10 2
Time (s) Time (s)

Fig. 18. Predictions of several material functions of the HDPE polyethylene melt at 200 °C using the K-BKZ models with Wagner and PSM damping functions and the multi-mode
PTT model (a) shear stress growth coefficient (ηS+ ) (b) shear stress decay coefficient (ηS− ) (c) first normal stress difference coefficient (1+ ), and (d) first normal stress difference
decay coefficient (1− ).
Steady recoverable elastic strain, r,s

102 101
Steady recoverable shear strain, r,s

K-BKZ (Wagner) (a) K-BKZ (Wagner) (b)

K-BKZ (PSM)
101 K-BKZ (PSM)
100 PTT
PTT
100
10-1
10-1

10-2
10-2

10-3 10-3
10-4 10-3 10-2 10-1 100 101 102 103 10-4 10-3 10-2 10-1 100 101
. .
Shear rate, (s -1 ) Hencky strain rate, (s-1 )

Fig. 19. Predictions of (a) steady recoverable shear strain (γ r,s ) and (b) steady recoverable elongational strain (ε r,s ) using the K-BKZ model with Wagner and PSM damping functions
and the multi-mode PTT model.

be strongly related to extrudate swell, also showed no dependence extrudate swell were studied using integral K-BKZ and differential
on the rheological model/constitutive equation, at least under sim- PTT constitutive models. The predictions were compared with avail-
ple deformations such as steady shear and elongation. As shown in able experimental observations. The results showed that the ex-
Figs. 17–19, it was observed that all the material functions during trudate swell decreases with temperature and increases with ap-
steady shear (dynamic), start-up flow (transient), relaxation and re- parent shear rate and die entrance angle (2a) consistent with ex-
covery were represented quite well using the three constitutive equa- perimental observations. Similarly, the simulation results (of both
tions (described in this paper), leaving the significant difference be- integral K-BKZ and differential PTT models) obtained using dies
tween swell predictions made by integral K-BKZ and differential PTT of different lengths (i.e. different lengths with constant diame-
models as an intriguing problem to explain. ter in the present study) show that the extrudate swell decreases
with increasing the L/D ratio, indicating the diminishing mem-
6. Conclusions ory of the polymer melt with die length. However, the integral K-
BKZ models (i.e., with Wagner and PSM damping functions) highly
The extrudate swell simulation of a high-molecular-weight HDPE overpredict the experimental measurements, while the differen-
was performed using 6-mode integral K-BKZ (with both Wag- tial PTT, Giesekus and DCPP models underpredict, leading to an
ner and PSM damping functions) and differential PTT, Giesekus intriguing problem to explain. In an attempt to understand this
and DCPP models. The effects of apparent shear rate, tempera- huge difference between the swell predictions of integral versus
ture, presence of reservoir, and die geometrical parameters on differential models, several material functions, including dynamic,
 V.K. Konaganti et al. / Journal of Non-Newtonian Fluid Mechanics 225 (2015) 94–105
transient, relaxation and recovery material functions were deter-
mined using K-BKZ and PTT models. However, all the material func-
tions calculated turned out to be the same by these models, leaving
the problem unexplained.
Acknowledgements
Financial assistance from NSERC (Automotive Partnership
Canada), the federal funding agency of Canada, is acknowledged.
References
[1] E. Mitsoulis, Computational polymer processing, in: P.D. Gujrati, A.I. Leonov
(Eds.), Modeling and Simulation in Polymers, Wiley-VCH Verlag, Weinheim, Ger-
many, 2010, pp. 127–195. (chapter 4).
[2] R.I. Tanner, A theory of die swell, J. Polym. Sci. Part A-2: Polym. Physics 8 (1970)
2067–2078.
[3] R.I. Tanner, R.E. Nickell, R. Bilger, Finite element methods for the solution of
some incompressible non-Newtonian fluid mechanics problems with free sur-
faces, Comp. Meth. Appl. Mech. Eng. 6 (1975) 155–174.
[4] D. Kalyon, V. Tan, M.R. Kamal, The dynamics of parison development in blow
molding, Polym. Eng. Sci. 20 (1980) 773–777.
[5] A. Dutta, M.E. Ryan, A study of parison development in extrusion blow molding,
J. Non-Newton. Fluid Mech. 10 (1982) 235–256.
[6] X.-L. Luo, E. Mitsoulis, Memory phenomena in extrudate swell simulations for
annular dies, J. Rheol. 33 (1989) 1307–1327.
[7] V. Ganvir, B.P. Gautham, H. Pol, M.S. Bhamla, L. Sclesi, R. Thaokar, A. Lele, M.
Mackley, Extrudate swell of linear and branched polyethylenes: ALE simula-
tions and comparison with experiments, J. Non-Newton. Fluid Mech. 166 (2011)
12–24.
[8] Y. Béreaux, J.-Y. Charmeau, J. Balcaen, Optical measurement and modelling of pari-
son sag and swell in blow moulding, Int. J. Mater. Form. 5 (2012) 199–211.
[9] M. Ansari, E. Mitsoulis, S.G. Hatzikiriakos, Capillary extrusion and swell of a HDPE
melt exhibiting slip, Adv. Polym. Technol. 32 (2013) E369–E385.
[10] R.I. Tanner, Die-swell reconsidered: some numerical solutions using a finite ele-
ment program, Appl. Polym. Symp. 20 (1973) 201–208.
[11] R.E. Nickell, R.I. Tanner, B. Caswell, The solution of viscous incompressible jet and
frees-surface flows using finite-element methods, J. Fluid Mech. 65 (1974) 189–
206.
[12] M.J. Crochet, R. Keunings, Finite element analysis of die swell of a highly elastic
fluid, J. Non-Newton. Fluid Mech. 10 (1982) 339–356.
[13] X.-L. Luo, R.I. Tanner, A streamline element scheme for solving viscoelastic flow
problems. Part I. Differential constitutive equations, J. Non-Newton. Fluid Mech.
21 (1986) 179–199.
[14] X.-L. Luo, R.I. Tanner, A streamline element scheme for solving viscoelastic flow
problems part II: integral constitutive models, J. Non-Newton. Fluid Mech. 22
(1986) 61–89.
[15] D.G. Kiriakidis, E. Mitsoulis, Viscoelastic simulations of extrudate swell for an
HDPE melt through slit and capillary dies, Adv. Polym. Technol. 12 (1993) 107–
117.
[16] X.-L. Luo, R.I. Tanner, Finite element simulation of long and short circular die ex-
trusion experiments using integral models, Int. J. Numer. Meth. Eng. 25 (1988)
9–22.
[17] A. Goublomme, B. Draily, M.J. Crochet, Numerical prediction of extrudate swell of
a high-density polyethylene, J. Non-Newton. Fluid Mech. 44 (1992) 171–195.
[18] A. Goublomme, M.J. Crochet, Numerical prediction of extrudate swell of a high-
density polyethylene: further results, J. Non-Newton. Fluid Mech. 47 (1993) 281–
287.
[19] G. Barakos, E. Mitsoulis, Numerical simulation of extrusion through orifice dies
and prediction of Bagley correction for an IUPAC-LDPE melt, J. Rheol. 39 (1995)
193–209.
[20] K. Housiadas, J. Tsamopoulos, Unsteady extrusion of a viscoelastic annular film: I.
General model and its numerical solution, J. Non-Newton. Fluid Mech. 88 (2000)
229–259.
[21] M. Bush, R.I. Tanner, N. Phan-Thien, A boundary element investigation of extru-
date swell, J. Non-Newton. Fluid Mech. 18 (1985) 143–162.
[22] J.M. Marchal, M.J. Crochet, A new mixed finite element for calculating viscoelastic
flow, J. Non-Newton. Fluid Mech. 26 (1987) 77–114.
[23] M.J. Crochet, Numerical simulation of viscoelastic flow: a review, Rubb. Chem.
Tech. 62 (1989) 426–455.
[24] D. Rajagopalan, J.A. Byars, R.C. Armstrong, R.A. Brown, J.S. Lee, G.G. Fuller, Com-
parison of numerical simulations and birefringence measurements in viscoelastic
flow between eccentric rotating cylinders, J. Rheol. 36 (1992) 1349–1376.
[25] R. Guénette, M. Fortin, A new mixed finite element method for computing vis-
coelastic flows, J. Non-Newton. Fluid Mech. 60 (1995) 27–52.
[26] V. Ganvir, A. Lele, R. Thaokar, B.P. Gautham, Prediction of extrudate swell in poly-
mer melt extrusion using an Arbitrary Lagrangian Eulerian (ALE) based finite el-
ement method, J. Non-Newton. Fluid Mech. 156 (2009) 21–28.
105
[27] E. Behzadfar, M. Ansari, V.K. Konaganti, S.G. Hatzikiriakos, Extrudate swell of
HDPE melts: I. Experimental, J. Non-Newton. Fluid Mech. (2015) (in press).
[28] M. Mooney, Explicit formulas for slip and fluidity, J. Rheol. 2 (1931) 210–222.
[29] M. Ansari, S.G. Hatzikiriakos, A.M. Sukhadia, D.C. Rohlfing, Rheology of Ziegler-
Natta and metallocene high-density polyethylenes: broad molecular weight dis-
tribution effects, Rheol. Acta 50 (2011) 17–27.
[30] R.I. Tanner, Engineering Rheology, Oxford University Press, Oxford, 2000.
[31] E. Mitsoulis, S.G. Hatzikiriakos, Steady flow simulations of compressible PTFE
paste extrusion under severe wall slip, J. Non-Newton. Fluid Mech. 157 (2009)
26–33.
[32] A.C. Papanastasiou, L.E. Scriven, C.W. Macosko, An integral constitutive equation
for mixed flows: viscoelastic characterization, J. Rheol. 27 (1983) 387–410.
[33] M.H. Wagner, Analysis of stress-growth data for simple extension of a low-density
branched polyethylene melt, Rheol. Acta 15 (1976) 133–135.
[34] L. Quinzani, R.C. Armstrong, R.A. Brown, Use of coupled birefringence and LDV
studies of flow through a planar contraction to test constitutive equations for
concentrated polymer solutions, J. Rheol. 39 (1995) 1201–1228.
[35] R.G. Larson, A critical comparison of constitutive equations for polymer melts, J.
Non-Newton. Fluid Mech. 23 (1987) 249–269.
[36] F.P.T. Baaijens, S.H.A. Selen, H.P.W. Baaijens, G.W.M. Peters, Viscoelastic flow past a
confined cylinder of a low density polyethylene melt, J. Non-Newton. Fluid Mech.
68 (1997) 173–203.
[37] A. Debbaut, T. Avalosse, J. Dooley, K. Hughes, On the development of secondary
motions in straight channels induced by the second normal stress difference. Ex-
periments and simulations, J. Non-Newton. Fluid Mech. 69 (1997) 255–271.
[38] A. Béraudo, A. Fortin, T. Coupez, Y. Demay, B. Vergnes, J.F. Agassant, A finite ele-
ment method for computing the flow of multi-mode viscoelastic fluids: compar-
ison with experiments, J. Non-Newton. Fluid Mech. 75 (1998) 1–23.
[39] N. Phan-Thien, R.I. Tanner, A new constitutive equation derived from network
theory, J. Non-Newton. Fluid Mech. 2 (1977) 353–365.
[40] N. Phan-Thien, A nonlinear network viscoelastic model, J. Rheol. 22 (1978) 259–
283.
[41] H. Giesekus, A simple constitutive equation for polymer fluids based on the con-
cept of deformation-dependent tensorial mobility, J. Non-Newton. Fluid Mech. 11
(1982) 69–109.
[42] M.A. Hulsen, J.V.D. Zanden, Numerical simulation of contraction flows using a
multi-mode Giesekus model, J. Non-Newton. Fluid Mech. 38 (1991) 183–221.
[43] M. Mostafaiyan, K. Khodabandehlou, F. Sharif, Analysis of a viscoelastic fluid in an
annulus using Giesekus model, J. Non-Newton. Fluid Mech. 118 (2004) 49–55.
[44] WC. Tsai, G.H. Miller, Numerical simulations of viscoelastic flow in complex ge-
ometries using a multi-mode Giesekus model, J. Non-Newton. Fluid Mech. 210
(2014) 29–40.
[45] T.C.B. McLeish, R.G. Larson, Molecular constitutive equations for a class of
branched polymers: the pom-pom polymer, J. Rheol. 42 (1998) 81–110.
[46] N. Clemeur, R.P.G. Rutgers, B. Debbaut, On the evaluation of some differential for-
mulations for the pom-pom constitutive model, Rheol. Acta 42 (2003) 217–231.
[47] G. Barakos, E. Mitsoulis, Non-isothermal viscoelastic simulations of extrusion
through dies and prediction of the bending phenomenon, J. Non-Newton. Fluid
Mech. 62 (1996) 55–79.
[48] M. Ansari, S.G. Hatzikiriakos, E. Mitsoulis, Slip effects in HDPE flows, J. Non-
Newton. Fluid Mech. (2012).
[49] A. Mitsoulis, Annular extrudate swell of pseudoplastic and viscoplastic fluids, J.
Non-Newton. Fluid Mech. 141 (2007) 138–147.
[50] A. Langouche, B. Debbaut, Rheological characterisation of a high-density
polyethylene with a multi-mode differential viscoelastic model and numerical
simulation of transient elongational recovery experiments, Rheol. Acta 38 (1999)
48–64.
[51] R. Keunings, An algorithm for the simulation of transient viscoelastic flows with
free surfaces, J. Comp. Phys. 62 (1986) 199–220.
[52] A.M. Henderson, A. Rudin, Effects of die temperature on extrudate swell in screw
extrusion, J. Appl. Polym. Sci. 31 (1986) 353–365.
[53] Y. Mu, G. Zhao, C. Zhang, Numerical investigation of die geometry effect on LDPE
annular extrudate swell, J. Appl. Polym. Sci. 1 (2010) 91–109.
[54] J. Guillet, C. Carrot, B.S. Kim, J.F. Agassant, B. Vergnes, C. Béraudo, J.R. Clermont,
M. Normandin, Y. Béreaux, Rheology for Polymer Melt Processing, in: J.M. Piau,
J.F. Agassant (Eds.), Elsevier, 1996.
[55] P. Saramito, J.M. Piau, Flow characteristics of viscoelastic fluids in an abrupt con-
traction by using numerical modeling, J. Non-Newton. Fluid Mech. 52 (1994) 263–
288.
[56] R. Ahmed, R.F. Liang, M.R. Mackley, The experimental observation and numerical
prediction of planar entry flow and die swell for molten polyethylenes, J. Non-
Newton. Fluid Mech. 59 (1995) 129–153.
[57] H.M. Laun, Prediction of elastic strains of polymer melts in shear and elongation,
J. Rheol. 30 (1986) 459–501.
[58] R.I. Tanner, S. Nasseri, Simple constitutive models for linear and branched poly-
mers, J. Non-Newton. Fluid Mech. 116 (2003) 1–17.
[59] R.I. Tanner, A.M. Zdilar, S. Nasseri, Recoil from elongation using general network
models, Rheol. Acta 44 (2005) 513–520.
[60] W. Wang, X. Li, X. Han, A numerical study of constitutive models endowed with
POM-POM molecular attributes, J. Non-Newton. Fluid Mech. 165 (2010) 1480–
1493.