Virtual thermoforming: ready for prime time?

Process simulation for thermoforming is still new and far from perfected. Yet a small but growing band of processors are discarding trial-and-error methods in favor of simulation as a shortcut to optimizing quality.

Sometimes what is thought to be the right mold, material, and processing set-up for a new thermoformed part can fail in early trials, posing a setback to product development. Thermoformers then may have to go back to the drawing board with no real idea of what went wrong. "Now thermoformers are looking to be more high-tech in their approach to product development and they see simulation as a way to get there," says John Perdikoulias, v.p. of Compuplast International, a supplier of simulation software. "Using simulation up front can help a processor develop a better part faster. There's no room left today for tedious and time-consuming trial and error," Perdikoulias says.

"Trial and error is a post-mortem process. With simulation we can tell the customer what can go wrong with their part, then feed back the results to a new simulation to improve the design," says Yves Rubin, product development manager at simulation software firm Fluent Inc. He says simulation provides knowledge not only of what happens in a process, but why it happens.

Franklyn Berry, v.p. at custom thermoformer Premier Plastics in Waukesha, Wis., concurs: "I'm always looking for a way to cut development, set-up or processing time. Simulation gives me benefits today," he says. "Many customers who submit designs for a part have critical criteria that they don't address until R&D is far along. Simulation helps them to tease out those requirements. We run a simulation to get a 'guesstimate' of how well the part will run and to see where thin spots may appear," Berry says.

Gaining momentum

Thermoformers have been slow to embrace simulation technology, which models the stretching and thinning of sheet, as well as its cooling in contact with the mold. Some formers cite the high price of the software packages and their unfamiliarity with the software platforms as issues that have kept simulation on the sidelines. One thermoformer says he was unwilling to trust a software "black box" in which he entered data and then out popped a result with no explanation of how it was reached.

But acceptance of simulation is now gaining momentum. One factor has been the arrival of software that runs on Windows-based PCs instead of Unix workstations. Software prices down to under $10,000 have also increased its appeal.

There are now at least 60 processors, material suppliers, and machine builders using or evaluating the technology. Simulation software is currently offered by five firms, and another is in a testing phase. Five of the suppliers are updating their packages with new features and functions. Some of the new features include faster data processing and simulation of forming multi-layer sheet or glass-reinforced materials.

Thermoforming process simulation is the topic of six papers being presented at the SPE ANTEC Conference taking place this month in New York City.

Settling arguments

Current programs can model deep draws or show the effects of sharp edges or radii in the corners of a mold. Results can be fed into a finite-element structural-analysis program to determine whether the part can meet mechanical requirements. Without simulation, processors can only guess at actual part thicknesses. Simulation also provides guidance for modifying the mold design, initial sheet thickness, or heater set-up in order to optimize material usage or prevent product failures. Heat-transfer models can also show you how the part cools so you can predict and/or minimize cycle time. Processors say simulation predictions can be within 10% of the optimal part and mold design and process parameters - at least with simple parts.

Simulation helps to settle disagreements between the design team and the production team. "Prior to our use of process-simulation software, we'd sit down in a conference and there would be conflicts of opinion as to which mold design is the best," says Greg Rensch, senior specialist engineer in the Manufacturing Research and Development Group of the Fabrication Div. of Boeing Commercial Airplane Group in Seattle. "Now we show a simulation of an optimized mold or part design, and the production department establishes more credibility with the engineering designer."

Simulation saves time, too. "Changes to the tool meant forming had to stop, the tool had to be removed, and then shipped back to the tool maker," says Dudhi Karfono, project engineer for the Exterior Systems Div. at Visteon, Milan, Mich. "It would take a month before the tool came back, even for minor changes. And then you'd hope the changes you made were right. A nine-month tool trial involving nine back-and-forth tool trips was typical. With process simulation, I can come up with an optimized mold design in two or three months, from which the prototype is made." Other formers say the time to create a good prototype is trimmed by up to 80%.

Simulation also provides processors with a chance to experiment with different process set-ups without consuming valuable machine time or material. "We use simulation aggressively where the part volumes are low, from 20 to 150 pieces," says Berry of Premier Plastics, which forms parts as large as refrigerator liners. "In one part where we were looking for stiffness and rigidity, we initially ruled out the use of polycarbonate because we thought it would be too expensive. Through simulation, we found out that we would have to upgauge to 3/8 in. if we used polyethylene, but we could get the needed performance for one-third the amount of PC."

Premier Plastics also used simulation to show that a recreational-vehicle assembly of five pieces of sheet metal could be made 30% lighter and less expensive with no sacrifice of performance by converting it to thermoformed ABS.

How it works

Simulation starts with a CAD model of a part or mold. Most thermoformers we interviewed prefer to use 3D CAD models, although a less detailed 2D model can generate a simulation faster because fewer data points are created. The CAD model is imported into the simulation package, or it can be generated from scratch using drawing features in the simulation software. (Most thermoformers use the former approach.)

Next, the simulation program prompts the user to enter data on the material behavior, typically extracted from a user-expandable database included in the package. The simulation software also asks the user to set up process parameters of a virtual thermoforming machine, such as heater zone settings, sheet movement, plug assist, draw boxes, and cooling rate of the mold. Once the CAD model, material selection, and machine parameters are set, the simulation package goes to work developing a finite-element mesh of the model.

Thermoformers rarely run just one simulation. In fact, they run several with different numbers of mesh elements in order to hone in on an optimized solution in the most time-efficient manner. It's a general rule that the more mesh elements requested, the longer it takes to generate the simulation.

As an example of how this works, Ford's Visteon Div. starts off with a "coarse" simulation of 1000 elements when it begins designing thermoformed products such as door panels, instrument panels, and bumper fascias. The coarse run shows general characteristics of how the part forms - e.g., thickness distribution, webbing, thinning, or thin spots. A simple model with 1000 elements takes about 4 hr to run, says Dr. Mohammad Usman, supervisor of CAD/CAM/CAE at Visteon. Again, the process isn't 100% accurate. For an instrument panel formed on a male tool, the simulation identified two places where webbing occurred, though actual test runs found four. "When we used a female tool in physical trials, there were two spots of webbing. The software simulation only saw one," says Usman.

Visteon reviews the data from the coarse run, debugs the part or mold model, and plugs in new data for another iteration. This procedure may take two or three days. After two or three coarse iterations, Visteon moves on to a complex simulation model comprising from 10,000 elements up to 40,000 for really sophisticated parts. These simulations take more time but produce more accurate results.

Premier Plastics typically runs about 10 simulations to refine a model. "You get 50% of your answers with the rough-cut analysis," says Berry.

Room for improvement

Despite these advantages, simulation is not a panacea. Thermoformers using it complain that the accuracy of its predictions still needs improvement. Modeling of both sheet heating and cooling is one area they feel needs additional work. Thermoformers want non-isothermal representation of heat transfer so that they can see, for example, the different effects of single-sided or top-and-bottom heating.

Another important issue concerns the use of a hyperelastic model or viscoelastic model of material behavior during processing. Early simulation software used only hyperelastic models, which treat the sheet as an infinitely elastic rubbery membrane. Viscoelastic models used by some later software versions model plastic flow as well as stretching. They take into account strain rate and how stress/strain relationships change with temperature. These help, for example, to predict sheet sag. "You can't tell what will happen if the sheet is heated for 20 seconds versus 30 seconds," says Wayne Shih, principal technical service representative at Eastman Chemical. Sag is important because it affects final thickness distribution, he notes.

"It's not perfect, but you can use simulation get real close to an optimized process," says Rubin. "Even if you don't get 100% accuracy, you'll do a lot less fine-tuning with simulation than without it."

Users and vendors of simulation software agree that one of its weak points is the materials data that it requires as inputs. A number of firms are working to remedy the paucity of accurate materials characterization for thermoformability. They include Eastman and other resin makers, some of which are working with the CAMPUS consortium to develop suitable material models. Materials data can also be developed for processors by commercial testing laboratories such as Datapoint Testing Service or Polymics Corp.

Users' favorite features

All of the current simulation programs display their results graphically as a "contour" map of the part thickness, area stretch ratio, or temperature in gradations of color. Many of the packages also allow you to "map back" the formed-part results to the original fiat sheet. Displaying the colored "map" of final thickness distribution on the fiat sheet reveals where thin spots will occur, so that processors can adjust heating profiles to improve material distribution.

Packaging producers also like the "map-back" feature for positioning graphics so that the desired image or logo will appear correctly in the formed part.

Simulation packages allow you to stop and analyze the forming process at any moment. This can help you to determine when to activate plug-assist movement, how deep a stroke is required, and when to engage the vacuum or shut off forming-air pressure. "When you have a deep-draw part of 6-8 in. or more, you must have simulation to see if you can do it right with the plug assist," says Visteon's Usman.

Simulation packages do a good job of telling the user where the first contact between sheet and mold will occur, users report. They also say the prediction of temperature profiles for formed parts is pretty accurate. Ability to simulate slip or friction of the sheet when it contacts the mold or plug is another useful software feature. It can help to optimize plug shape and movement. Here again, good material data are important. "If the friction between sheet and tool that is used in simulation is different from the real world, the thickness distribution in the final part will be different from the prediction," says Capel English, director of CAE at the design firm Polymer Solutions.

Some simulation packages offer shrinkage and warpage modules that predict final dimensions, stresses, and part distortion. However, the accuracy of these calculations depends heavily on the quality of the material model. Premier Plastics' Berry also warns that accuracy of shrink/warp predictions may be compromised if the mold CAD model has an insufficient number of elements. In general, he is not convinced that there is any way to accurately predict warpage today.

New developments

Among the six suppliers of simulation software, Polydynamics introduced the first package, T-formcad, in 1992. Its features remain essentially unchanged. The software costs $1900 for an annual license or $5700 for a perpetual license. Meanwhile, other firms are coming out with new or upgraded products:

* C-Mold has improved the modeling and visualization portion of the user interface module for its thermoforming simulation package. It also has improved the animation of output results on pressure, temperature, and stretching. The software, available for a Unix workstation or a PC, costs $10,000 for the user interface and $30,000 for the thermoforming simulation package.

* Compuplast will soon offer a 4.0 version of the T-SIM simulation package developed by T-SIM cz in the Czech Republic. The upgraded version is said to make simulation set-up easier, and it allows users to rotate, zoom, and pan images in real time. New display options allow for shading and wire frames. The viscoelastic model reportedly has been improved. New features will include shrinkage and warpage. Perdikoulias also says the package can simulate forming of multi-layer sheets or filled and reinforced materials. Price starts around $15,000.

* Fluent Inc.'s new Polyflow 3.7, coming in a few months, offers enhanced modeling of heating, sheet sagging, prestretching, plug-assist movement, and billow forming. It also has non-isothermal viscoelastic properties. Polyflow 3.7 calculates results twice as fast as before. Other features are integrated pre/post-processing, CAD import, automatic meshing, process animation, and an expandable database of 30 generic materials. Available on Unix and PCs, Polyflow 3.7 is priced around $20,000.

* The Industrial Materials Institute of the Canadian National Research Council (IMI-CNRC), together with a consortium of 15 academic and industrial organizations (including Dow Plastics and Montell Polyolefins), is developing thermoforming simulation based on its current package for blow molding. The new software will contain a viscoelastic model and will display animated process sequences that analyze heating, forming, cooling, and warpage. An optimization feature can automatically feed simulation results back into the program to run additional iterations. PC and workstation versions will both be Windows based. A multi-layer feature will be included. The product can be licensed through an arrangement with the consortium. Robert DiRaddo, group leader of process modeling and optimization, also says this package will be the first to simulate twin-sheet forming.

* Sherwood Technologies is developing four Java-based programs for thermoforming that analyze only sheet heating and cooling (see PT, Oct. '98, p. 55). It will specifically address simulation of single-sided and top-and-bottom heating, as well as patterned heating. Sherwood's programs will cost $150 each or $500 for the set.

* University of Massachusetts at Amherst is developing software for modeling multi-material coextrusions. David Kazmer, assistant professor of mechanical and industrial engineering, is conducting research to create a simulation that combines models for the different materials in the coex structure, He hopes to offer a suite of software tools by year's end.