SPE Extrusion Division 1-0-Wiki

RSS

Navigation

Extrusion 1-0-Wiki Pages

Search the Wiki
»

Viewing/Creating

Account Management

Administration

Brought to you by:
The SPE Extrusion Division
Board of Directors


PoweredBy

Comparison of the Flow Performance Between Internal and External Deckling in Flat Film Extrusion Die Systems

Olivier Catherine, Cloeren Incorporated, Orange, TX

Abstract

Deckles are devices useful to adjust the slot width of extrusion dies, and hence the extruded product width. There are basically two types of deckles: (i) internal deckles are made of components placed inside the die flow channel to block the flow channel to a specified width upstream the die lips, and (ii) external deckles typically comprise blockage devices positioned directly on and external to the die lips, at the exit orifice. To understand the fundamental differences in flow performance between the two technologies, 3D Computational Fluid Dynamics (CFD) models were built. A comparison of the original die width to the internally and externally deckled dies is carried out by evaluating the flow characteristics such as velocity uniformity at the die exit and Residence Time Distribution (RTD) in the die flow channel. The flow models show that the performance of an internally deckled die is close to that of a nondeckled die, while the external deckle system results in non-uniform flow distribution and broad RTD due to the occurrence of a large stagnation area upstream of the external deckle.

Introduction

High performance flat film dies are typically tailored to specific process conditions, a specific polymer rheology, and a specific slot gap. However, flexibility is often needed in the production process. The extrusion equipment may be required to process different outputs, temperature conditions or even different materials from the original design target. Along those lines, it is not unusual for processors to have requirements for different product widths. In order to minimize the edge trim, while the ideal technical solution would be to use a different die width for each product width, a more practical solution may be to use flow width restriction devices, also known as deckles, to change the width of the extrudate flowing form/exiting the die.

Deckle systems can be fairly complex assemblies but, in principle, they are relatively simple in concept as they consist for the most part of devices that block off the flow channel portion to reduce flow channel to a desired width. The first type of deckle is the internal system (Figure 1). It is located in the flow channel and is known to be the most streamlined of deckle systems. A more simple and cost competitive system is the external deckle, shown in Figure 2, which basically consists of an seal placed on the die lips to create a local flow blocking device to regulate the die slot width.

Performance differences are expected between the two deckle solutions and 3D CFD studies are run to understand and teach the fundamental differences between the two solutions. The study uses the finite element CFD program, Solid Works Flow Simulation 2011 SP1.0, which was previously validated for non-Newtonian, nonisothermal flow and used for a previously published study[1].

This flow simulation study is divided in three main sections: (i) a model of the original (full-width) die; (ii) two models corresponding to 250 mm and 500 mm external deckles; and finally (iii) two models corresponding to 250 mm and 500 mm internal deckles.

Die geometry

This work deals with a flat film extrusion scenario. For comparison sake, the specific output was kept constant for all three slot widths. The initial die slot width is 2050 mm and the deckled scenarios correspond to a slot width of 1550 mm and 1050 mm for the 250 mm deckles and 500 mm deckles respectively. Table 1 details the basic process parameters, Figure 3 and Figure 4 show the die flow channel geometry used for this study. The die designed by Cloeren Incorporated features rectangular entrance and manifold channels, as well as a two stage preland.


Table 1: Process parameters

Polymer and rheology

The polymer used for this study is a compounded blend of two grades of ethylene-vinyl acetate copolymer (EVA):

  • 75% of EVA-1
  • 25% of EVA-2

Where EVA-1 is an EVA manufactured by DuPont which has a vinyl acetate comonomer content of 18 wt%, a density of 0.94 g/cm3 and a melt flow index (MFI) of 0.7 g/10 min (190°C/2.16 kg). EVA-2 is also produced by DuPont and also has a vinyl acetate comonomer content of 18 wt%, a density of 0.94 g/cm3 and a MFI of 2.5 g/10 min (190°C/2.16 kg).

The shear viscosity behavior of this EVA compound is modeled by the combination of the Cross model[2] for the shear dependency and the WLF model[3] for the temperature dependency, according to equations (1) and (2).




Where ή is the shear viscosity (Pa.s), y is the shear rate (1/s), T is the temperature (K), ή0 is the zero shear viscosity (Pa.s), τ* is the characteristic shear stress (Pa) and m is the pseudo-plastic index. ή0, τ* and m are the Cross model parameters. The time-temperature superposition principle (TTS)[4] is included in the Cross model through the shift factor aT, which follows a WLF model, which uses the C1 and C2 (K) parameters. The parameter values used for this work are summarized in Table 2 and illustrated in Figure 5.


Table 2: Cross and WLF parameters

Finally, the following thermal properties were used: melt density ρ= 0.795 g.cm-3, thermal conductivity k=0.25 W.m-1.K-1 and specific heat Cp= 3.14 J.g-1.K-1.

Flow simulations

The Non-Newtonian, non-isothermal flow simulation method and governing equations are similar to those previously published [1].

SolidWorks Flow Simulation 2011 SP1.0[5] was used to solve the coupled thermal-flow problem and a mesh with brick elements was built. The boundary conditions for this model are: (i) a symmetry condition at the centerline for flow and temperature; (ii) a fully developed flow at the entrance ; (iii) an initial fluid temperature and a flow rate are imposed at the entrance; (iv) at the outlet surface (lip opening/exit orifice), the exit pressure is set to 1 atm and; (v) finally, on the flow channel walls, we assume a non-slip condition (v=0 mm/s) and a uniform temperature (210°C) is applied.

1 - Full-width die

The die flow channel design was optimized for the full width configuration, which corresponds to a slot width of 2050 mm for a total extrusion output of 748.3 kg/h. A flow simulation was run for these conditions with the geometry illustrated in Figure 3. The calculated pressure drop is about 15.5 MPa, and Figure 6 shows the evolution of the melt pressure as a function of the position along the main flow direction (MD) at the centerline of the die. Figure 7 shows the velocity contour plot on the plane defined at mid-plane of the height of the lip land gap. As can be seen in Figure 8, the exit velocity across the die width is very uniform.

Because of the relatively high melt viscosity, viscous dissipation in the die can be a legitimate concern. However, as shown in Figure 9 the predicted temperature elevation at die exit due to viscous dissipation is controlled. The plot shows a maximum temperature of 211.8°C at the mid-plane of the flow stream. By way of general information, the peak temperature proximate to the die wall, in the lip land area, was calculated at 214.6°C. As usual for non-isothermal flow assuming a uniform die wall temperature, the warmer portion of the melt is observed near the die ends of the die width.

Finally, another critical parameter in this study is the residence time distribution in the die. For this study, the numerical estimation of the residence time is very challenging due to the existence of low flow rate regions. The best method in this case is to perform a virtual injection of particles in the flow. Residence time is evaluated by tracking the particles from the entrance to the exit. This numerical method is relatively common and many publications can be found for RTD and mixing performance estimation in twin screw extrusion[6 - 8]. This method provides a relative comparison of residence time distribution for different die configurations. There are many parameters that define a particle injection study, like for example the type of particle material, the particle size, the initial temperature etc… These parameters will not be detailed here, however, in order to compare the die deckle variations, these parameters are kept constant throughout the study. One thousand particles were injected in a plane near the entrance of the manifold channel. On each resulting particle trajectory, the trajectory time is plotted. The resulting plot is shown in Figure 10. This particle injection study yields to statistics of major interest, like for example, the cumulative residence time distribution curve shown in Figure 11, and the residence time envelope, shown in Figure 12.

2 - Externally deckled die

Two flow simulations were carried out for the situation where external deckles are used, corresponding to deckles of 250 mm per side (total slot width of 1550 mm) - see Figure 13 - and deckles of 500 mm per side (total slot width of 1050 mm) – see Figure 14.

The total extrusion output was adjusted with the slot width to keep the specific extrusion output constant, at 3.65 kg/hr/cm. Consequently, the calculated melt pressure drop in the die for both deckle configurations is close to that of the full die width. Specifically, the 250 mm deckle configuration resulted in a melt pressure drop through the die of 14.7 MPa and the 500 mm deckle configuration in a melt pressure drop through the die of 14.6 MPa. For illustrative purposes, the melt pressure contour plot through the die flow channel using the 500 mm deckles is shown in Figure 15.

Velocity contour plots for the externally deckled die are shown in Figure 16 and Figure 17 for the 250 mm deckles and the 500 mm deckles respectively. These plots show the occurrence of a large area upstream the deckles where the flow velocities are low. These zones can be considered stagnation areas. In order to understand the flow patterns better, velocity vectors are plotted on top of the contour plots in the deckle area, near the exit. In particular, Figure 18 shows the existence of a strong velocity gradient in the area of the slot proximate to the deckle extremity. This gradient is caused by the excess polymer flowing from a portion of the area upstream and behind the deckle. This phenomenon is further characterized by the exit velocity profiles shown in Figure 19.

The thermal characteristics of the flow in the externally deckled dies are illustrated by the temperature contour plot shown in Figure 20. This shows a classic shear heating effect on the side walls of the entrance channel, along the back wall of the manifold channel. However, the melt is dissipating heat in the area upstream and behind the deckled region to match the die wall temperature owing to the very low flow velocities. As a result, this cold area provides a good correlation to the stagnant melt flow regimes.

Capturing the residence time distribution in the deckled configuration proved challenging; and two steps are necessary to really evaluate the whole die flow pattern. When using the particle injection near the manifold channel entrance similarly to what is done for the full width die, it is possible to compare the residence time figures to that of the full width reference. However, this particle injection method does not fully capture the high residence time area at the very ends of the flow channel and immediately upstream the deckles, as shown by the particle trajectories in Figure 21 and Figure 22. These trajectories however may be seen as the measurable majority of the flow reaching the die exit and therefore, it is interesting to study their statistics compared to the full width configuration. For example the cloud plots of trajectory time versus trajectory flow length for all three configurations (Figure 23) show interesting results. The larger the deckles, the longer the residence time, despite the shorter flow lengths associated with the active width of the die. Interestingly, the maximum flow length for the 250 mm deckle configuration is very close to that of the full width configuration because of the flow trajectories inside the low velocity region ultimately exit the die near the deckle inner end defining the exit slot. The RTD curves (Figure 24) characterize residence time associated with external deckles.

Nevertheless, these figures are not complete because as mentioned before, the trajectories do not capture the regions near the end of the manifold channel. It is possible to obtain a representation of the residence time in these regions when injecting particles along the back of manifold channel, as close as the entrance as possible. Figure 25 shows that the residence time in these regions can be well in excess of 300 s for the 250 mm external deckle configuration, while Figure 26 illustrates a residence time greater than 1,500 s for the 500 mm deckle configuration. While these numbers should not be interpreted as absolute values, they provide a good indication of the flow pattern and RTD in the externally deckled dies. As can be appreciated from this modeling, residence time increases exponentially with external deckle width.

3 - Internally deckled die

Similar to the flow simulations for the external deckle system, two flow simulations were run for the internal deckle scenarios with 250 mm and 500 mm deckles respectively. The principles applied for the boundary conditions for the internal deckle models were the same as for the external deckle flow simulations.

Calculated pressure drop is about 14.8 MPa for both conditions, which is comparable to the full width model. This is expected because the specific output is kept constant relative to the effective slot width for all the simulations. For illustration purposes, the pressure contour plot for the 500 mm internal deckle scenario is shown in Figure 27.

Velocity contour and vector plots are shown in Figure 28 and Figure 29 respectively, for the 250 mm internal deckles; and in Figures 30, 31 respectively, for the 500 mm internal deckles. The internal deckle configuration results in higher overall flow velocities compared to the corresponding external deckle widths. The internal deckle device beneficially replaces this low velocity regimes present in the flow channel when using the external deckles. Also, the flow velocities at the die exit are much improved. Compared to the external deckle scenario, internal deckles result in a much more uniform velocity profile at the exit of the die, particularly at the end regions of the deckled die width. The velocity surge that was present with external deckles at the die exit near the deckle is nonexistent with internal deckles. Figure 32 shows the flow profile at the die lips. It exhibits slightly more end flow when wider deckles are used. The die preland, which is designed to uniformly distribute the flow across the full slot width, becomes less adapted for narrower slot widths. However, this behavior is acceptable because even when using 500 mm deckles, the flow distribution is satisfactory and well within the range of correction from typical lip adjustment systems.

The melt temperature pattern is very similar to that of the full width die, with temperature gradients from the wall to the mid-plane of the channel. The temperature contour for the 250 mm internal deckle configuration is shown in Figure 33. The temperature profile at the die lips exhibits slightly warmer melt near the ends, which is a signature of shear heating in the flow channel. The intensity of this phenomenon however is largely insignificant as demonstrated by the modest temperature increase (1.8°C).

The virtual particle injection method was used to evaluate the RTD for internal deckle configurations. The resulting particle tracings are shown in Figure 34 and Figure 35 when injecting particles near the entrance of the manifold channel for the 250 mm internal deckle configuration and the 500 mm internal deckle configuration respectively. This analysis shows a relatively streamlined flow pattern, equivalent to that of the full width die in both cases. The relationship between residence time and flow length is shown in cloud graph in Figure 36. The shapes of the RTD clouds show that internal deckle configurations behave very similarly to a full width die. Correspondingly, the shapes of the RTD curves (shown in Figure 37) exhibit a behavior of similarity between internally deckled and the full width configuration dies. The main different lays in the smaller values of maximum residence time when using wider deckles (narrower slot width). Finally, a direct comparison of RTD curves between the internal style and external style deckles is shown in Figure 38 for the 250 mm deckle configurations and Figure 39 for the 500 mm deckle configurations.

Conclusion

CFD was successfully used to compare two styles of deckle systems commonly used in flat film or sheet extrusion. The use of advanced CFD tools like virtual particle tracking allows for the performance evaluation of the polymer flow between two different deckle configurations. For instance, it is possible to illustrate and quantify the flow stagnation upstream of external deckles. Flow non-uniformity with this system is also demonstrated. By comparison, it was found that internal deckle systems compare very similarly to full width dies as the flow at the exit of the die remains fairly uniform and unaffected by the die width.

[1] O. Catherine, SPE-ANTEC Tech Papers (2011)
[2] M. M. Cross, Polymer Systems: Deformation and Flow (1968)
[3] M. L. Williams, R. F. Landel and J. D. Ferry, J. Am. Chem. Soc., 77, 3701 (1955)
[4] J. D. Ferry, Viscoelastic properties of polymers. 3rd edit. (1980)
[5] Dassault Systèmes, S.A.
[6] V. L. Bravo, A. N. Hrymak, and J. D. Wright, Polym. Eng. Sci., 44: 779–793 . (2004)
[7] X. –M Zhang, L.-F Feng, W.-X Chen and G.-H Hu, Polym. Eng. Sci., 49: 1772–1783 (2009)
[8] F. Ilinca, J. F. Hetu, Intern. Polym. Proc., XXV 4: 275-285 (2010)


Figure 1. Schematic of a fixed width internal deckle system


Figure 2: External deckle system consisting of (1) a seal, (2) the deckle seal plate and (3) retaining fingers with (3) set screws


Figure 3: Die geometry used for the flow studies – plane view


Figure 4: Die geometry used the flow studies – side view


Figure 5: Viscosity curves of the selected Cross WLF model


Figure 6: Pressure drop profile at the centerline in the main flow direction (machine direction), from entrance (0 mm) to exit (400 mm), showing the contribution of the flow channel portions


Figure 7: Velocity contour plot on the plane defined a mid-height of the lip land


Figure 8: Computed exit velocity profile


Figure 9: Temperature contour plot on the plan defined at mid-height of the lip land gap


Figure 10: Estimation of residence time distribution by virtual particle injection


Figure 11: RTD (Residence Time Distribution) curve for full width die


Figure 12: Trajectory time across the die width as a function of trajectory length for full width configuration


Figure 13: Die geometry for 250 mm external deckles


Figure 14: Die geometry for 500 mm external deckles


Figure 15: Melt pressure distribution map in the die using the 500 mm external deckles


Figure 16: Velocity contour plot through the flow channel with 250 mm external deckle


Figure 17: Velocity contour plot through the flow channel with 500 mm external deckles


Figure 18: Velocity contour plot superimposed with velocity vectors representing the direction and the intensity of the velocity field


Figure 19: Exit velocity profiles for the externally deckled die


Figure 20: Temperature contour plot on midplane for the 500 mm external deckle configuration


Figure 21: Result of virtual particle injection for the die with the 250 mm external deckle configuration - Particles injected near the entrance to the manifold channel


Figure 22: Result of virtual particle injection for the die with the 500 mm external deckle configuration - Particles injected near the entrance to the manifold channel


Figure 23: Comparison of trajectory time to trajectory flow length clouds for external deckle configurations


Figure 24: Comparison of cumulative residence time distribution curves for external deckle configurations


Figure 25: Result of virtual particle injection for the die with the 250 mm external deckle configuration - Particles injected at the midplane near the back wall of the manifold channel


Figure 26: Result of virtual particle injection for the die with the 500 mm external deckle configuration - Particles injected at the midplane near the back wall of the manifold channel


Figure 27: Melt pressure contour plot for the 500 mm internal deckle configuration


Figure 28: Velocity contour plot for the 250 mm internal deckle configuration


Figure 29: Velocity contour plot superimposed with velocity vectors representing the direction and the intensity of the velocity field for the 250 mm internal deckle configuration


Figure 30: Velocity contour plot for the 500mm internal deckle configuration


Figure 31: Velocity contour plot superimposed with velocity vectors representing the direction and the intensity of the velocity field for the 250 mm internal deckle configuration


Figure 32: Velocity profile at the die lips (at mid plane)


Figure 33: Melt temperature contour plot on the mid plane of the die for the 250 mm internal deckle configuration


Figure 34: Result of virtual particle injection for the die with the 250 mm internal deckle configuration - Particles injected at the midplane near the entrance of the manifold channel


Figure 35: Result of virtual particle injection for the die with the500 mm internal deckle configuration - Particles injected at the midplane near the entrance of the manifold channel


Figure 36: Comparison of trajectory time to trajectory flow length clouds for internal deckle configurations


Figure 37: Comparison of cumulative residence time distribution curves for internal deckle configurations


Figure 38: Comparison of RTD curves for internal and external deckles (250 mm)


Figure 39: Comparison of RTD curves for internal and external deckles (500 mm)

Return to Paper of the Month.
Some of the icons were created by FamFamFam.