This tutorial will demonstrate how to calibrate a temperature-dependent Three Network Viscoplastic (TNV) material model using MCalibration. The TNV model is an advanced material model in the PolyUMod library. Like most other material models in the PolyUMod library, the TNV model does not come with equation-based temperature dependence. Instead you need to combine TNV models that have been calibrated at different temperatures into a single Multi-Temperature model. The Multi-Temperature model is a PolyUMod framework that linearly interpolates the material parameters based on the current temperature. The example shown here will use experimental data for a PEEK material that was tested at 4 different temperatures. At each temperature, the material was tested in tension and compression at 2 different strain rates. That is a useful dataset!
Step 1. Calibrate a Material Model to the Data at Each Temperature
In this example I will use the TNV model, which is an advanced viscoplastic material model that works well for virtually all polymers. The following figure shows the calibration results at room temperature. The average error of the model predictions in this case is less than 6%.
Step 2. Export the Calibrated Material Model Parameters
Step 3. Combine the Calibrated Material Models into a Temperature-Dependent Model
The exported material model parameter files can be combined into a single temperature-dependent material model by:
- Opening the Material Model dialog box
- Click on “PolyUMod-Template”
- Then clock on the “Combine Multi-Temp” button.
This will bring up a file selection dialog box in which you should select the parameter files that were exported in Step 2. Click “Open” and then “Save” to select the new temperature-dependent material model.
More details of the PolyUMod Multi-Temperature Model framework is available in the PolyUMod User’s Manual.
The last step is to specify the actual temperatures that were used for the experimental tests. In this case the temperatures were: 296 K, 373 K, 423 K, and 474 K.
The figure shows that the temperature-dependent TNV model can predict the complete dataset with an average error of 7.9%. Very nice!
This is more accurate than the temperature dependent Three Network (TN) model which had a predicted error of 16.6% for the same experimental dataset.