Why not extrusion simulation?

Process modeling software can help you or your machinery/material suppliers optimize your process.

Trying out new extrusion equipment or materials too often boils down to trial-and-error processing, where even years of experience can't eliminate all the guesswork. The same holds true for optimizing existing extrusion processes. Increasingly, though, computer-aided engineering (CAE) can help draw a quick, cost-efficient bead on the solutions to all sorts of processing problems and equipment-design challenges.

Yet unlike injection molding, in which computational modeling and analysis have been well-received, extrusion lags behind when it comes to CAE. Most extrusion processors simply pass up simulated extrusion in favor of the real thing. That lack of acceptance, however, rests on some discouraging facts: CAE simulations of dynamic extrusion problems tend to be complex, and the software can be expensive not only to buy but also to run. "Simulations don't always run on a PC," says Joseph Dooley, research leader of Dow Chemical's Polymer Processing Technology Team. "Many times you're talking about some type of mainframe." In fact, several researchers at resin and equipment suppliers tell of "big simulation problems"--such as true three-dimensional modeling of extrusion--that have taken days to set up and hours to run on a supercomputer.

While plastics processors may have relatively little direct experience with extrusion CAE, it does exert a behind-the-scenes influence. Resin companies and equipment manufacturers already use it to design screws, dies and coextrusion feedblocks and to help processors select resins and process set-ups. And judging from its reception on this supplier level, CAE's impact will keep growing. What's more, the seemingly unstoppable growth of affordable computer power may encourage more processors to employ computerized analysis tools in the future. "As the computer industry evolves, the ability to run big problems on small platforms will emerge, and more people will use these tools," predicts Dow's Dooley.

The argument for using CAE as a first step in problem solving appears most compelling from an economic standpoint. "You can run 100 simulations in the time it takes to change the barrel temperatures on a real extruder," notes Ron Klein, v.p. of Scientific Process & Research Inc., which makes a simulation package called Extrud-PC. And as Quantum Chemical Corp. research specialist Harry Mavridis points out, tinkering with a process in a production environment "often wastes valuable material until you arrive at the process settings that meet production goals."

TODAY'S SIMULATION TOOLS

The range of computer analysis tools that can be applied to extrusion problems encompasses everything from general-purpose fluid dynamics software to programs solely for singlescrew extrusion. In general, the latter programs begin by asking operators to input information about a real or potential process configuration, including equipment geometries, barrel-zone temperatures, and hopper conditions. These simulations also require certain rheological and other property data on the resins under analysis. So, the end-user must often derive the resin data experimentally--or hire the software vendor to do it. CAE software packages do, however, often come with a database of information on common materials and provisions for the user to add more. Extrud-PC, for instance, has a database of 5000 commercial materials, says Klein.

Once the user enters the data needed for a simulation, complex mathematical algorithms model the meltviscosity, shear rate, temperature and pressure in order to predict outcomes such as production rate, pressure drop, and output uniformity. When designing equipment, these simulations allow a user to run through many iterations in the effort to tailor equipment design or processing conditions to a set of technical and economic goals. For example, Quantum recently used simulation to narrow down the potential routes to increased output on a high-volume BOPPfilm line. "We needed to increase throughput while maintaining a low temperature," says Mavridis. Simulation helped optimize the barrel temperatures on the first try. Moreover, the solution would seem "counter-intuitive" to some processors because barrel temperatures had to be raised to keep the melt temperature low, according to Mavridis.

Several software packages aimed at single-screw processors are available for users with no more computing power than the a 386-based PC. Scientific Process & Research recently introduced a new version of its software for single-screws. Called Extrud-PC Release 3.9, it allows users to evaluate the production capacity of extruder-and-die combinations by calculating the head pressure needed to pump melt through the die at a given rate. In addition to the more usual question-and-answer interface, Release 3.9 also incorporates a brand-new graphical interface that allows users to call up pressure and temperature information corresponding to any point on a screw profile. It also presents broader slices of data and graphically correlates different processing variables. For example, the program can display the size of the melt pool relative to the amount of unmelted solids at different points along the screw. A related program, called Diedes, automatically calculates die-design and processing considerations such as strain-bar settings and lip openings.

A similar single-screw package comes from the Polymer Processing Institute (PPI). Called Polymer Analysis and Simulation Software (PASS), it looks at extruder configurations and has separate modules that address different die-flow simulations. One such module performs non-isothermal, non-Newtonian analysis of flow through mandrel dies, and the other does non-Newtonian flow analysis through coathanger dies, says PPI manager Ron Rakos. This program, too, addresses strain-bar and lip configuration in dies.

Canada's Center for Advanced Polying and Design (CAPPA-D) offers a suite of extrusion anal-ysis programs. One called ExtruCAD models temperature, pressure and flow rate in ,. a single-screw extruder. FlatCAD simulates flow in flat-film and sheet dies, while CoexCAD is used for flow simulations in fishtail or coathanger coex two-layer dies. SpiralCAD analyzes spiral-mandrel dies, predicting flow, pressure and temperature variations throughout the die body. The programs all predict the thickness variation in the final product, according to the center's director John Vlachopoulos. Except for ExtruCAD, all the programs use the "control volume," or "2 1/2-D" method employed by many injection mold-filling packages.

Lastly, Plastics & Computer Inc. has recently adapted its injection mold-filling package, faBest, to extrusion die dimensioning. The user defines geometry of the die passage, inputs material data and enters a range of processing parameters. Though the package can target any processing variable, company president Anne Bernhardt says the usual objective is pressure balancing at the die lips.

For twin-screw extruders, the modeling work has not yet filtered down to the processor level, so little exists in the way of commercial software dedicated to this application. At least one equipment supplier, however, has come up with a package for use by its customers. Werner & Pfleiderer Corp. sells software called Exco 3 for configuring its twin-screw machines. This package graphically depicts screw and barrel configurations, but does not actually simulate the process. Meanwhile, the PPI nears completion on a control-volume model for twin-screw mixing, says Ron Rakos.

NOT FOR THE UNINITIATED 

Whatever the potential benefits from CAE, no one suggests that this approach can directly benefit every processor. The software vendors themselves all agree that the extrusion-specific packages still need people with a great deal of technical savvy to enter the correct data and interpret the results. PPI's Rakos, for instance, says that the level of expertise needed to correctly analyze the output from PASS and other CAE programs best corresponds to someone with an "advanced engineering degree." And he's not alone in that assessment: "The more accurate the data, the more accurate the simulation," says Klein. "And that sort of accuracy requires someone with processing knowledge."

Even further from the expertise level found on many shop floors are general-purpose analysis programs for solving three-dimensional fluid-dynamic problems common to extrusion, chemical processing, hydrodynamics and aerodynamics. Two such commercial packages have seen use in a plastics setting: Fidap from Fluid Dynamics Inc. (FDI) and Nekton from Fluent Inc. FDI's program employs finite-element modeling (FEM), which divides the object of analysis into a mesh of smaller geometric regions. The finite element method then mathematically calculates the pressure, temperature and flow velocity conditions within each of these regions and assembles them into a description of conditions for the entire object.

Fluent's Nekton software relies on the related spectral-element method (SEM). Its meshed models contain fewer nodes but are described by higher-order equations, rather than more nodes with the low-order equations of FEM, according to product manager S. Subbiah.

At least for extrusion, the bottom line for both methods is often a significant time investment in setting up a problem. "It takes a lot of data to drive these models," says Mark Spalding, research leader of Dow's Polymer Processing Technology Team. To model a screw's metering section using Fidap, Spalding says he had to build a finite-element mesh with 60,000 elements. The problem took months of development time, additional days to actually set up the mesh, and a week-and-a-half to solve on a DEC VAX computer. Spalding also notes tha the can perform only 15 or so of this sort of complex 3-D simulation in a year, compared with up to 100 a day for simpler modeling with PC-based programs.

Mavridis says Quantum has found a use for Fidap: modeling temperature gradients on a melt-delivery pipe for a customer puzzled by anomolous thermocouple readings. But Mavridis, too, concedes that this tool can be tough to use. "If you keep changing the problem, it's very difficult to start over from scratch."

Moreover, these programs present stiff entry requirements not just in terms of computing power but also in the level of technical sophistication needed to apply them successfully--success, in this case, meaning correctly focusing on real-world solutions. The 3-D analysis packages do offer striking graphic displays. But users say that nice pictures can be illusory, hiding the underlying mathematical acumen needed to run the programs. "It's easy to create beautiful graphics that can mislead you," explains Dooley. "The finite-element method is mathematically and numerically intensive."

Owing to the educational overhead and time constraints, even R&D professionals don't necessarily turn to the 3-D programs first. Instead, they work with a hierarchy of analysis tools. At Dow, for instance. Dooley often uses an extrusion-specific package--from CAPPA-D--for quick solutions to relatively simple problems, such as running through several die-design iterations. "We might test a problem on four or five software products," adds Spalding. Starting with the simplest software to get the "gross effects," the researchers progress to more and more complicated programs if necessary. Like any task, the choice comes down to picking the right tool for the right job. "It's important that the method addresses all the variables you're concerned with," he says. So for blown film where flow rate, pressure and temperature can throw off output uniformity, the analysis method must be able to address these variables separately, Dooley says. His choice in this application would be a package using the "control-volume" method.

Twin-screw modeling proceeds on the research level as well. At W&P, senior research engineer Arash Kiani has developed a proprietary model that predicts flow and temperature over the length of a twin-screw extruder. "We've had good success predicting the full machine, including the melting section," Kiani reports. "Though W&P has used both Nekton and Fidap to analyze fluid dynamics over individual elements, the company's own model goes a step further by assembling all the localized information. "The commercial programs predict local behavior, but they don,t predict machine behavior," he says.

Of course, even the high-end analysis packages aren't necessarily out of reach of all processors. Both the Fidap and Nekton packages do come in PC versions. And the more sophisticated processor can make use of the software. According to Fluent's Subbiah, Nekton is used at 54 extrusion sites worldwide, almost exclusively for coextrusion and coating applications. He cites a coex packaging line owned by James River Corp. in Ohio as a good example of how process simulation and analysis can benefit a processor directly. On that line, he says, process engineers were able to redesign the internal elements of an existing flat die to eliminate a degradation problem at the film's edges without re-tooling. Dow also has used Nekton to analyze coextrusion dies.

TECH-SERVICE TOOLS

Complex or not, extrusion simulation does seem to have an indirect benefit: Its use by resin and equipment suppliers has resulted in some real customer-service improvements. "Competition is stiff out there," notes

Quantum's Mavridis. "We offer processors our simulation experience as a way to get our foot in the door." Quantum has developed one such simulation tool of its own for blown- and cast-film coextrusion in an effort to solve layer-uniformity and interfacial flow-instability problems (see PT, Feb. '91, p. 54). Since its launch two years ago, Quantum has added a new user interface and passed the program onto its technical service staff. According to Mavridis, these engineers take the program into the field on portable computers, helping customers select resins or optimize processing conditions. Likewise, Dow uses simulation programs to debug customers, lines.

Process simulations--as distinguished from the already-common CAD/CAM--also influence machine and resin design. In blown film, Davis-Standard uses proprietary simulations tools to tailor each blown-film die to a set of customer-specified functional requirements, says product manager Peter Gates. "I couldn't do some of the designs without simulation," Gates says. And noting that financial considerations would have made design errors ruinous when he first started building dies, Gates says simulations have helped him build over 500 dies without re-cutting even one.

The benefits apply to other components and processes as well. "We use simulation to facilitate the design of screws," says Dow's Spalding. "It's the first place we go to determine dimensions or find out whether the screw geometry is wrong." Computer modeling of solids conveying behavior, for instance, has enabled him to design a special screw for increasing the output of engineering thermoplastics in film applications. Similarly, Klein notes that the Solids Draining Screw sold by Scientific Process & Research would not have even existed if not for simulations. This screw has an opening at the end of its metering section to drain unmelted solids, transporting them back to the melt section via an internal, non-rotating screw.

In pipe extrusion, Fluent's Subbiah adds that Nekton has seen use in modeling swell and shrink of the extrudate after it leaves the die.

A FUTURE ON THE SHOP FLOOR?

Beyond the time and expertise it takes to work these packages, another barrier to widespread CAE use remains. Some vendors and users doubt that the packages will ever make their way into processing plants in significant numbers. "There is always going to be a handful of specialized people to use these tools," says Spalding. "An engineer in the plant is not going to have the desire or the time because his job is to push out the pounds."

At a deeper level, gaps in the understanding of extrusion's underlying physical mechanisms persist. Software vendors and users agree that more work needs to be done to come up with better process models. Unlike injection mold filling, extrusion presents a multi-phase problem that encompasses solids conveying and melting behavior as well as fluid dynamics, notes W&P's Kiani. And acceptable mathematical models for all these phenomena just don't exist yet. W&P's model, for example, has a better handle on melt-fed extrusion processes, which strictly falls within the realm of fluid dynamics. Spalding also cites viscoelastic behavior as an especially problematic nut for the available packages to crack.

At this point, then, overcoming model limitations means turning to old-fashioned know-how. "Personal experience is still key," explains Spalding. And for that reason, experimentation backs up the simulation efforts at many places. "Not all of the information in our model is based on physics and fluid dynamics," W&P's Kiani says. "We have added experimental data and process history."

So as useful as these packages may be, empiricism still rules. "Given the choice between experimental data and a simulation," says Dow's Dooley, "I'll go with experimental every time."

FOR MORE DETAILS ON ITEMS IN THIS ARTICLE, USE READERS' SERVICE CARD Center for Advanced Polymer Processing and Design, McMaster University, Hamilton, Ontario (CIRCLE 60) Davis-Standard, div. of Crompton & Knowles, Pawcatuck, Conn. (CIRCLE 61) Dow Chemical Co., Midland, Mich. (CIRCLE 62) Fluent Inc., Lebanon, N.H. (CIRCLE 63) Fluid Dynamics International Inc., Evanston, Ind. (CIRCLE 64) Plastics & Computer Inc., Montclair, N.J. (CIRCLE 65) Polymer Processing Institute at Stevens Institute of Technology, Hoboken, N.J. (CIRCLE 66) Quantum Chemical Corp., USI Div., Cincinnati (CIRCLE 67) Scientific Process & Research Inc., Somerset, N.J. (CIRCLE 68) Werner & Pfleiderer, Ramsey, N.J. (CIRCLE 90)