How to Calibrate a Material Model When You Have Results From Multiple Repeated Experiments


I often recommend using multiple repeated experiments for each test condition, just to make sure you understand how repeatable the experimental data is. There are no hard rules for how many repeated experiments one should use, but it is common to use 2 to 4 repetitions for each test condition. The figure below shows the results from repeated experiments of a thermoplastic material that was cyclically loaded in compression.


Method 1: Use All Data

The easiest way to handle data from multiple repeated experiments is to simply use all data! This is easy to do in MCalibration. Just create one load case for each experimental test. MCalibration will then calibrate a material model that fits all data in an average sense. The downside of this approach is that it can make the material model calibration run a bit slower (but only if you don’t allocate one CPU for each load case in MCalibration, which can be set in the Options dialog box).

Method 2: Select the Median Stress-Strain Curve

Another approach is to look at the experimental data and select the data set that seems to be in the middle (the median). The image below shows the stress-strain curve that looks to be the medial curve. This approach works well if  one curve is close to the medial response for all strain levels (which is not always true). The material model calibration will then be performed using this median curve only, and the experimental data from the other tests will be neglected (which can feel odd to simply ignore a bunch of experimental data).

Method 3: Use the Average Stress-Strain Curve

A third approach is to use MCalibration to calculate the average stress-strain curve from the all the available tests. The figure below shows the average curve in thicker black dashed lines. The average curve seems to be a good representation of the individual tests. The main issue with this method is that it tends to round off corners at stress-strain reversals. Note that the average curve can be quickly and automatically calculated in MCalibration by simply selecting the load cases that you want to combine, and then select “Combine Load Cases” item in the “Load Case” menu.



My personal approach to how to handle data from multiple repeated tests is to use all experimental data files. If that slows down the calibration more than I feel is worth it, then I switch over to only using the median stress strain curve. I tend to not use the Average curve since it is less accurate at stress-strain reversals. I encourage you try all three methods, and find your own preferred approach.


More to explore

Leave a Comment