Experiment and simulation for a PUR-cylinder
Verification and validation is necessary in order to check whether a physical law and the material parameters used can provide realistic results. Here, the results of an experiment are compared with those of an FE simulation. The first step is to check whether an experiment used to determine material parameters can be simulated exactly. This step is called verification. A validation examines the extent to which the simulation and the material parameters determined are capable of describing the mechanical behaviour of other, more complex parts.
The parameters defined during parameter identification normally contain several sets of data. It is sensible to put more weight on those experimental data which fit the material better. This is achieved with weighting matrices shown in the following diagram.
- Weighted objective functions for several tests, Weightings for several tests, Time/loading steps and shear directions
Verification
Verification serves to check whether an experiment used to derive a set of material parameters can be simulated exactly. In this phase, the model and the parameters can still be calibrated and adjusted, in order to improve results.
A verification for PUR-cylinders is shown in the following animation. The parameters used in the simulations were derived from all three test runs.
- Verification of material parameters
Validation
As has been mentioned, a verification is not sufficient to ensure that a model and the parameters are suitable for the calculation of real parts. The main reason for this is most often the more complex multi-axial stress patterns involved. The quality of the model and its parameters can only be finally checked by comparison of experiment and simulation with a more complex part.
It is important to note that, after succesful verification, no further change in the model may be made with regard to the mathematical formulation or the material parameters. Should the validation show good agreement between between test and simulation results then one can assume that the calculation of different geometries can be performed reliably.