Thermoforming FEA Simulations – What Material Property to Use?

A common misconception in the thermoforming community is that, before we can begin to work on a FEA simulation of the thermoforming process, we need to spend considerable time and money to develop a complex visco-elastic material model for every polymer. Because of this apparent hurdle, many believe that it is just too costly and time consuming to make simulations a part of the product development process. However, there is a much simpler, low cost (or no cost) material model that works well for most thermoformable polymers.

Nearly all thermoformable polymers behave in a rubber-like manner when heated to within their forming temperature window and stretched. Rubber-like materials retain their initial volume. As the polymers are stretched in-plane they thin out-of-plane (thickness direction) and the total volume does not change. For FEA simulations we call the rubber-like material hyperelastic.

There is now a considerable body of published research work that shows that for most thermoformable polymers we can use the simple hyperelastic material model. In fact, because of the similarity in the way the majority of thermoformable polymers behave during forming, one usually do not need to run any material testing to develop a material model. We can input a rather wide range of hyperelastic material properties in the FEA simulation and predict about the same final material thickness distribution in the final product. Thus, a one-size-fits-all material model actually works quite well for most thermoforming simulations. The only exception is if we want to accurately predict the vacuum or pressure required to form the part, which is almost never the case.

There are a few polymers that go through a strain hardening as they are stretched and thus we cannot use a hyperelastic model. Most likely, a visco-elastic material model would be required for polymers that do not behave like a hyperelastic material during forming.

The main goal of a thermoforming simulation is to predict the material distribution in the final product. Don’t let it get overly complicated. In fact, I have yet to see a published paper that shows that developing a visco-elastic material model provides more accuracy than a simple hyperelastic material model.