ESTIMATION OF ELONGATIONAL VISCOSITY OF POLYMERS FOR

ACCURATE PREDICTION OF JUNCTURE LOSSES IN INJECTION MOLDING

Mahesh Gupta
Mechanical Engineering-Engineering Mechanics Department
Michigan Technological University
Houghton, MI 49931

Abstract elongation rate [4]. Therefore, Meissner’s technique can be

used to determine elongational viscosity only at very low
A new elongational viscosity model along with the
elongation rates (< 10 s-1). Also, Meissner’s technique can
Carreau-Yasuda model for shear viscosity is used for a
only be used for highly viscous melts. Since the
finite element simulation of the flow in a capillary
deformation rate in injection molding can range from 1,000
rheometer. The entrance pressure loss predicted by the
to 100,000 1/s, this technique cannot provide the
finite element flow simulation is matched with the
elongational viscosity data required for injection molding
corresponding experimental data to predict the parameters
simulation.
in the new elongational viscosity model.
Due to the difficulties associated with direct
Introduction measurement of elongational viscosity of a polymer, the
flow in a channel with abrupt contraction (entrance flow)
Since the thickness of most plastic parts is small,
has often been used for an indirect measurement of
typically the flow during the filling stage of injection
elongational viscosity [4]. Since a polymer going through
molding is highly shear dominated. Therefore, the
an abrupt contraction experiences a large elongational
generalized Newtonian constitutive models with shear
deformation, a steep pressure drop, called entrance pressure
thinning viscosity have been successfully employed to
loss, is encountered near the abrupt contraction. The value
accurately simulate the mold filling process [1]. However,
to entrance pressure loss, which depends upon the flow rate
in certain regions of the mold, such as the runners, gates
in the channel, can be used for an indirect measurement of
and portions of mold cavity with abrupt changes in
the strain-rate dependence of the elongational viscosity of a
thickness, polymer melt goes through a significant
polymer. The advantage of the entrance flow method for
elongational deformation. The pressure gradient in these
measuring elongational viscosity is its applicability at high
regions with elongation-dominated flow, predicted by a
elongation rate. Furthermore, an existing capillary
generalized Newtonian formulation can be significantly
rheometer, which is commonly used for measuring shear
smaller than the pressure gradient encountered in the
viscosity, can also be used for elongational viscosity
molding process [2]. The primary reason for the poor
estimation.
prediction of elongation-dominated flow, is the low
elongational viscosity of the generalized Newtonian models Assuming that the pressure loss in an entrance flow can
[3]. The long chain polymer molecules exhibit stiff be obtained by summing the pressure drops due to shear
resistance to an elongational deformation. Therefore, the and elongational deformations, Cogswell [7] obtained
elongational viscosity of a polymer, which is defined as the separate expressions for these two pressure drops. To
ratio of the elongational stress to elongational strain rate, is further simplify the analysis, Cogswell made several other
very high. An accurate knowledge of the elongational assumptions listed below:
viscosity of a polymer is important for simulating the 1. shear stress and shear rate are related by power-law
elongation-dominated flows. model
2. elongational viscosity is constant
Various techniques such as uniaxial extension,
3. the interface between the recirculation region and the
lubricated compression, fiber spinning, bubble collapse,
main flow is a straight line (funnel-shaped flow)
stagnation flow, etc., have been used in the past to
4. no-slip on the interface between recirculation region
experimentally characterize the elongational viscosity of
and the main flow
polymers. A good review of these techniques and the
5. fully developed flow with zero radial velocity
difficulties associated with each of these techniques is
6. neglect Weissenberg-Robinowich correction [3]
presented by Macosko in reference [4]. Among these, the
7. inertia-less, incompressible flow with no normal
most commonly used technique is the uniaxial extension.
stresses in shear.
An elongational rheometer using uniaxial extension, which
With these assumptions, Cogswell performed a force
is based upon Meissner’s work [5], is currently marketed by
balance on a differential section of the funnel-shaped entry
Rheometric Scientific, Inc. [6]. However, maintaining a
region near abrupt contraction. By integrating the
steady uniaxial extensional deformation is difficult at a high
expressions thus obtained, Cogswell obtained the following in an abrupt contraction, Eqn. (3) can be used to determine
expressions for average elongation rate in entrance flow the power-law model (B, m) for the elongational viscosity
(Eqn. (1)) and the corresponding elongational viscosity of a polymer.
(Eqn. (2)) for an axisymmetric flow.
By using independent power-law models for shear and
4τ w γ̇ a elongational viscosities, Gupta [9] analyzed the effect of
ε̇ = ------------------------------ (1)
3 ( n + 1 )∆ p e elongational viscosity on vortex formation and entrance
3 ( n + 1 )∆ p e pressure loss in an axisymmetric 4:1 entrance flow. Similar
η e = ------------------------------ (2) analysis of the flow in a channel with abrupt contraction for
8ε̇
the planar and the three-dimensional cases were later
where τ w = ∆pR ⁄ 2L is the shear stress at the capillary
presented in references [10] and [11], respectively. In
3
wall, γ̇ a = 4Q ⁄ ( πR ) , ∆ p e is the entrance pressure loss, references, [9 - 11], the truncated power-law model was
used for the shear viscosity as well as for the elongational
∆p is the pressure drop in the capillary, R and L are the
viscosity in the axisymmetric and planar cases. This
radius and length of the capillary, respectively, and ε̇ is the approach of simulating polymeric flows is not only frame
elongation rate. invariant, but predicted velocity and pressure distributions
By removing some of the assumptions in the for a general 3-D flow are also identical to those from the
Cogswell’s analysis, Binding [8] obtained a more accurate generalized Newtonian formulation, if the specified shear
expression for entrance pressure loss in an abrupt and elongational viscosity models are such that
contraction. However, many of the assumptions in η e ( e II ) = 3η s ( e II ) for axisymmetric flows and
Cogswell’s analysis, listed below, are retained in Binding’s η e ( e II ) = 4η s ( e II ) for planar flows, with e II being the
analysis:
n–1
second invariant of the strain rate tensor. More recently,
1. power-law model for shear viscosity ( η s = Aγ̇ ) as using the same approach, Sarkar and Gupta [12] employed
m–1 the Carreau model for shear viscosity and a modification of
well as for elongational viscosity ( η e = Bε̇ )
Carreau model for elongational viscosity.
2. the interface between the recirculation region and the n–1
------------
main flow is a straight line (funnel-shaped flow) 2 2
3. no-slip on the interface between recirculation region η s = η 0 ( 1 + ( λe II ) ) (4)
and the main flow
4. fully developed flow with no radial velocity   m–1
-------------
 1  2 2
5. dR ⁄ dz « 1 , such that the terms involving (dR/dz)2 η e = η 0 3 + δ  1 – ----------------------------------  [ 1 + ( λ 2 e II ) ] (5)
 1 + ( λ 1 e II ) 
2
and d2R/dz2 can be neglected, implying a large recircu-  
lation zone The elongational viscosity model in Eqn. (5), reduces to the
6. energy consumed in recirculation zone can be Carreau model for δ = 0 with the elongational viscosity
neglected parameters λ2 and m replacing λ and n respectively, in the
7. inertia-loss, incompressible flow with no normal shear viscosity model. The two additional parameters λ1
stresses in shear. and δ in Eqn. (5), control the increase in the elongational
With these simplifications, Binding employed energy viscosity in the elongation-thickening region [12].
principles to obtain the following equation for entrance
pressure loss: In the present work, we have used the Carreau model
1 for shear viscosity (Eqn. (4)) and model proposed by Sarkar
m -------------
2 A ( 1 + m ) mB ( 3n + 1 )n I nm
2 m+1 and Gupta for elongational viscosity (Eqn. (5)). Knowing
∆ pe = ------------------------------ ---------------------------------------------
2 2
the shear and elongational viscosity parameters, a finite
3m ( 1 + n ) A
(3) element simulation of the entrance flow is performed. The
m(n + 1) 3m ( n + 1 ) finite element simulation of the entrance flow eliminates
---------------------- -------------------------
m+1 m+1
× γ̇ w 1–α most of the simplifying assumptions of Cogswell and
Binding’s analysis. However, the flow simulation still
where α = R0/R1, with R0 and R1 being the radii of assumes an inertia-less, incompressible flow with no
normal stresses in a shear flow. With a knowledge of the
downstream and upstream channels respectively,
shear viscosity and entrance pressure loss, using the
1+1⁄n m+1
∫0 ( 3n + 1 )ρ
1
I nm = ⁄n–2 ρ dρ , and the shear procedure presented in the next section, the elongational
3 viscosity parameter are optimized such that the difference
rate at capillary wall γ̇ w = ( 3n + 1 )Q ⁄ ( πnR 0 ) . If the between the experimental values of entrance pressure loss
power-law model for shear viscosity of a polymer, and and the corresponding predictions from the finite element
entrance pressure loss is known for two different flow rates simulation is minimized.
 Estimation Procedure for Elongational points, a very large computation time is required for
Viscosity Parameters estimation of elongational viscosity parameters, if all the
data points are used in the optimization scheme. Therefore,
In our earlier publication [13], we reported a global a systematic scheme, described in the next paragraph, is
minimization procedure to optimize the elongational employed in the present work to select the four entrance
viscosity parameters. In particular, starting with an initial loss vs flow rate data points for estimation of the four
guess for the four parameters, the simplex optimization elongational viscosity parameters.
scheme [13] was used to iteratively improve their values
such that the difference between experimental and Estimation procedure of elongational viscosity
predicted values of entrance loss is minimized. Even though parameters starts with an initial guess for the four
such a global minimization procedure is more rigorous, and parameters. The software allows the user to specify an
attempts to minimize the error over the complete range of initial estimate of the four parameters. Otherwise, the
flow rate vs entrance loss data, it requires very large amount software automatically initializes the elongational viscosity
of computational time, and convergence to the optimal parameters based upon the generalized Newtonian
values is prohibitively slow at times. formulation, that is η e ( e II ) = 3η s ( e II ) , or from the values
obtained by Binding’s analysis. It should be noted that in
To improve the efficiency of the estimation software, in
both of these cases δ = 0, and λ1 have no affect on
the present work, only the values of λ2 and m are optimized
elongational viscosity if δ = 0. Therefore, only λ2 and m are
simultaneously, which is followed by individual
optimization of δ and λ1. It should be noted that parameters initialized. Since the experimental data at higher flow rate is
expected to be in the power-law region, the data point at the
λ2 and m affect the elongational viscosity only in the
highest flow rate and another data point with flow rate of
power-law region at high strain rate, whereas δ and λ1 about one-hundredth of the highest flow rate are used for
control the shape of the elongational-thickening portion of estimation of power-law parameters λ2 and m. Once the two
the viscosity curve [12]. In particular, λ1 determines the points for estimation of power-law parameters are
strain rate for onset of elongation thickening region, identified, as shown in Fig. 1, the optimal values of λ2 and
whereas δ controls the total increase in the viscosity in the m are determined by alternating the Newton-Raphson
thickening region. Since δ and λ1, and λ2 and m affect the method for λ2 and m. The function fm and f λ are the
elongational viscosity in different regions, values of δ and 2
differences in experimental and predicted values of
λ1, and those of λ2 and m can be optimized independently.
entrance pressure loss for points with the highest flow rate
Because of this localized effects of δ and λ1, and that of λ2 and one-hundredth of the highest flow rate, respectively. At
and m, if all the entrance pressure loss data is in power-law this point of the algorithm λ2 and m have been identified,
region, optimization of the elongation-thickening however δ is zero and λ1 is also unknown. With δ = 0 and
parameters is difficult and vice versa. The localized effects
using the estimated value of λ2 and m in the elongational
of power-law (λ2 and m) and elongation-thickening (δ and
viscosity curve, the data point having the largest difference
λ1) parameters on elongation viscosity curve, result in the
between experimental and predicted value of entrance loss
inefficiency of the global minimization scheme. is used for optimization of δ. If some of the points in the
To optimize the four elongational viscosity parameters, entrance loss vs flow rate data are in the Newtonian range
the experimental data for entrance loss is required for at of elongational viscosity, experimental value of entrance
least four different flow rates. Two of these points should be loss for these points should match with the predicted
in the power-law region (to determine λ2 and m). If the entrance loss for δ = 0 in the elongational viscosity model
polymer melt exhibits the elongation-thickening region, one (Eqn. (5)). Therefore, starting with the lowest flow rate, the
of the points should be in the elongation-thickening region first experimental point with greater than 10% error
(to determine δ) and the fourth point should be in the between experimental and predicted entrance loss is used to
transition region between Newtonian and elongation optimize the value of λ1. Before starting the optimization of
thickening behavior (to determine λ1). Typically, entrance δ, the parameter λ1 is initialized to a very large value λ2 =
pressure loss vs flow rate data is available at a large number λ1 × 1010, which implies that elongation thickening is
of points ( » 4 ). Since an entrance flow simulation for a assumed to start at a very low elongation rate. Using the
fixed set of elongational viscosity parameters requires point identified above for optimization of δ, as shown in
significant amount of computation time (from 2 - 20 Fig. 1, Newton-Raphson method is used again to estimate
minutes on a Sun workstation, depending upon the number the value δ. The function fδ in Fig. 1 is the difference
of nodes in the finite element mesh and number of iterations between experimental and predicted entrance loss for the
required for convergence), and the entrance flow is point being used for optimization of δ. Finally, the
repeatedly simulated at the entrance loss vs flow rate data parameter λ1 is estimated by minimizing the entrance loss
error for the point identified above for λ1 optimization. The elongational viscosity parameter δ was attempted by using
bracketing and bisection method [14] is used for optimizing the entrance loss data at these two points. However, at both
the value of λ1. the temperatures, the predicted entrance loss at the lowest
flow rates in Fig. 2, was found to be highly insensitive to a
Experimental Data change in δ, indicating that even the two points at the
lowest flow rates in Fig. 2 are in the elongation-thinning
A Goettlfert Rheometer 1000 was used to measure the region of elongational viscosity. Therefore, for Dow 132i
viscosity and entrance pressure loss. The diameter of the elongational-thickening parameters (δ, λ1) in Eqn. (5)
capillary in all the dies used was 1 mm. The pressure loss vs could not be predicted using the available entrance loss
capillary length curves for dies with capillary length of 5, data. Accordingly, the elongational viscosity curves for
20, 30 and 40 mm were extrapolated to obtain the entrance Dow 132i in Fig. 3 do not show any elongation-thickening
pressure loss. The Weissenberg-Robinowitch correction behavior. However, the elongational viscosity for Dow 132i
was applied to account for the non-parabolic velocity
may have an elongation-thickening region for eII < 1 s-1,
profile. A Bohlin VOR with 25 mm parallel plates was used
to measure the shear viscosity at low shear rates. in the which can only be captured using our algorithm if the
experiments, a low density polyethylene (Dow 132i) was entrance loss data is available at smaller flow rates. Since
used. The entrance pressure loss and shear viscosity was the entrance loss at lower flow rates may be too small to
measure with a reasonable accuracy, elongational viscosity
measured for two different temperatures - 160 and 175 oC.
at lower elongation rates may not be predicted accurately
using the current technique. This is not a serious limitation
Results and Discussion
of this technique because the elongation rates in polymer
In this paper, the elongational viscosity for a low processing are typically quite high. Also, if the elongational
density polyethylene (Dow 132i) is estimated at two viscosity at lower elongation rates can be measured by
different temperatures. For Dow 132i at 160 and 175 oC, other experimental techniques mentioned earlier in the
the experimental data for entrance pressure loss at various paper, a composite elongational-viscosity may be
flow rates is shown in Fig. 2. As expected, the entrance loss constructed by finding the elongation-thickening
increases significantly as the flow rate is increased. At a parameters (δ, λ1) using the elongational viscosity
fixed flow rate the entrance loss is higher for the lower measured by direct experimental techniques and the power-
temperature. The shear viscosity for Dow 132i, which is law parameters (λ2, m) from the entrance pressure loss
obtained by fitting the Carreau model to the experimental measurement.
data, is shown in Fig. 3. The Carreau model parameters for
For Dow 132i, the elongational viscosities at 160 and
the shear viscosity of Dow 132i are as follows:
175oC, predicted by Cogswell’s and Binding’s analysis are
160oC: η0 = 2.6 × 105 Pa.s λ = 37.35 s-1 n = 0.3409
also shown in Fig. 3. Interestingly, even though Cogswell’s
175oC: η0 = 2.1 × 105 Pa.s λ = 31.69 s-1 n = 0.3409 and Binding’s predictions are based upon numerous
The entrance loss data in Fig. 2 and the Carreau model simplifying assumptions, the elongational viscosity
for shear viscosity in Fig. 3, was used in the present work to predicted by the two analyses are in good agreement with
predict the elongational viscosity employing the algorithm the predictions from the present work.
discussed earlier in this paper. For the two different
temperatures, the predicted elongational viscosity is shown Acknowledgement
in Fig. 3. It should be noted that the predicted elongational
The author would like to thank Prof. F. A. Morrison for
viscosity in Fig. 3 is plotted against e II = 3ε̇ . The making the experimental data for entrance pressure loss and
power-law parameters for the elongational viscosity were shear viscosity available to us.
predicted using the entrance pressure loss value for flow
rates of 5.65 and 339 mm3/s for 160oC and for 5.66 and 452 References
mm3/s for 175oC. With these control points the 1. H. H. Chiang, C. A. Hieber and K. K. Wang, Polym.
elongational viscosity parameters are as follows: Eng. Sci., 31, 116 (1991).
160oC: λ2 = 1.61 s-1 m = 0.281
2. P. Brincat, C. Friedl and K. Talwar, SPE ANTEC
175oC: λ2 = 2.35 s-1 m = 0.317 Technical Papers, 44, 552 (1998).
With these values of power-law parameters for elongational
3. R. B. Bird, R. C. Armstrong and O. Hassager,
viscosity, the largest difference in the experimental and
Dynamics of Polymeric Liquids, Vol. 1 and 2, John
predicted entrance pressure loss is at the lowest flow rates -
Wiley, New York (1987).
1.13 and 0.565 mm3/s for 160 and 175oC, respectively. For 4. C. W. Macosko, Rheology Principles, Measurements
each of the two temperatures, optimization of the and Applications, VCH, New York (1994).
5. J. Meissner, Rheol. Acta., 8, 78 (1969). 11. M. Gupta, SPE ANTEC Tech. Papers, 45, 83 (1999).
6. Rheometrics Scientific, Inc. One Possumtown Road, 12. D. Sarkar and M. Gupta., “An Investigation of the
Piscataway, NJ 08854, USA (http://www.rheosci.com). Effect of Elongational Viscosity on Entrance Flow”, to
7. F. N. Cogswell, Polym. Eng. Sci., 12, 64 (1972). be presented at ASME-IMECE, November 2000.
8. D. M. Binding, J. Non-Newtonian Fluid Mech., 27, 193 13. D. Sarkar and M. Gupta, “Estimation of Elongational
(1988). Viscosity using Entrance Flow Simulation”, to be
9. M. Gupta, Polym. Eng. Sci., 40, 23 (2000). presented at ASME-IMECE, November 2000.
10. M. Gupta, “Simulation of Planar Entrance Flow Using 14. W. H. Press, B. P. Flanery, S. A. Teukolsky and W. T.
Strain-Rate-Dependent Shear and Elongational Vetterling, Numerical Recipes, Cambridge University
Viscosities”, J. Reinforced Plastics and Composites, Press, Cambridge (1989).
(In press).
8
7
Initial estimate of 6
δ, λ1, λ2, m 5
4

Entrance loss (Mpa)

Identify points for m 3
and λ2 optimization
2

i+1 i
fλ
λ2 = λ 2 – ------------------------
2
-
d f λ ⁄ dλ 2
2 1
Experimental (160oC)
Predicted (160oC)
Experimental (175oC)
i+1 i fm Predicted (175oC)
m = m – ----------------------
d f m ⁄ dm
-9 -8 -7
10 10 10
3
Flowrate (m /s)
No Fig. 2. Entrance loss vs. Flow rate for Dow 132i.
Have λ2 and m converged?

107
Yes
106
Identify points for δ
and λ1 optimization 105
ηs, ηe (Pa.s)

104
fδ ηs ηe
i+1 i
δ = δ – ------------------- 103
d f δ ⁄ dδ
102 ηe optimization (160oC)
ηe optimization (175 C)
o

ηe Binding (160 C)
o
1
10
ηe Binding (175 C)
o
No
Has δ converged? ηe Cogswell (160oC)
100 ηe Cogswell (175oC)
ηs (160oC)
Yes 10-1 ηs (175oC)
Estimate the value of λ1 -3 -2 -1 0 1 2 3 4 5
10 10 10 10 10 10 10 10 10
using Bracketing and . .
Bisection technique eII = γ, √3ε (s ) -1

Fig. 3. Variation of shear (ηs) and elongational (ηe)

Fig. 1. Flow chart for optimization of elongational viscosities of Dow132i with the second invariant
viscosity parameters. of strain-rate tensor (eII).