Introduction
Polycarbonate (PC) is a thermoplastic material that is strong and tough. In this article I will examine different material model options for this important material. I will specifically examine the Johnson-Cook model, the Bergstrom-Boyce (BB) model, the Ansys TNM model, the Abaqus PRF model, and the PolyUMod TNV model. I picked these models since they are often used for this class of materials.
Experimental Data for PC
The figure below shows the experimental data that I used for the study. The data consists of uniaxial tension and uniaxial compression data at different strain rates. The tension tests were performed to failure, but I did not attempt to calibrate a failure model for this study.
I calibrated the following material models using the MCalibration software. Get your own free trial license.
Results: Ansys BB Model
The figure below compares the predictions from the Ansys Bergstrom-Boyce (BB) model to the experimental data. The average error in the model predictions is 13.4%. This material model was created for rubber-like materials, and it is no surprise that it does not work that well for thermoplastics.
Results: Abaqus 3 Network PRF (Yeoh + Power Flow)
The figure below compares the predictions from the Abaqus PRF model to the experimental data. The average error in the model predictions (NMAD) is 11.9%. The predictions are not very impressive.
Results: Abaqus Johnson-Cook
The Johnson-Cook model predictions are shown below. The average error in the model predictions is 11.7%. The predictions do basically bi-linear, which is not in agreement with the experimental data. I do not recommend using this model for PC.
Results: PolyUMod TN
The PolyUMod Three Network (TN) model is able to predict the response of the PC significantly more accurately. The average error of the model predictions is 6.24%. The predictions still do not allow for softening after yielding in compression.
Results: PolyUMod TNV
The most accurate material model for PC is the PolyUMod TNV model. In this example I first calibrated the TNV model with the following structure:
The results from this material model are shown in the figure below. The average error of the model predictions is 4.68%. That is pretty good.
I also calibrated a second alternative TNV model structure, see image below. In this alternative TNV model I used the PSC flow element.
The results from calibrating this TNV model are even more accurate, as shown in the figure below.
In this case the average error in the model predictions becomes 4.1%. Excellent.
