Resulting in Material Modeling
I recently read the book “Thinking in Bets” by Annie Duke. It is an excellent book, you should check it out if you have not already. In this book the author talks about how to make decisions when you have limited information. One example that she mentions in the book is how poker players are forced to make decisions with limited information, and how difficult it is for a poker player to evaluate different strategies due to the concept of Resulting. That is, it is very easy to fall into the trap of thinking that a strategy is good just because you won a specific hand. When in fact it may have been a bad strategy, and you were just lucky.
The concept of Resulting also applies to material model selection and calibration. Just because a calibrated material model matches some experimental data does not mean that it is a good material model. You should think carefully so that you don’t fall into the trap of resulting.
Here is an example where I have uniaxial tension data for a polyethylene, and I have calibrated an isotropic hardening plasticity model to the experimental data. The figure shows an excellent agreement between the experimental data and the model predictions.
But, is this a good material model, or are we just resulting (i.e. assuming that the model is good since it agrees with the data)?
It is an example of resulting in material modeling!
I found some more experimental data for the same polymer. The figure to the right shows that our initial material model does NOT agree well with the new more complete set of information. In other words, the elastic-plastics material model was not good.
In this case there are much better material models. This figure shows the predictions of the PolyUMod Three Network model. This model captures the essential physics of the deformation response of the material. It is a good model, and not simply resulting.
Here’s a video with some more info about resulting in material modeling.