Webinar: Parameter Identification with optiSLang
This webinar gives information about the mathematical basis of model calibration. Furthermore, the issue of relevance and quality of the identified parameters will be discussed. These methods can be applied easily for any RDO task with the help of optiSLang. A key role is the definition of signals and signal functions as well as the sensitivity analysis using the Metamodel of Optimal Prognosis (MOP).
Model calibration means to adapt the results of simulation models to actual measurement data. Here, a measured response curve, e.g. a load displacement curve, is taken as a reference and parameters of the simulation model will be modified until the best correlation between reference and simulation is obtained. This method is also known as "reverse engineering". Using this methodology, parameters that cannot be measured directly, such as material parameters, are identified. Therefore, this method is called parameter identification.
1. Basics of model calibration
- Maximum likelihood method
- Weighted least squares method
- Definition of signals and signal functions
2. Application of sensitivity analysis for calibration problems
- Definition of parameter spaces and design of experiments
- Metamodel of Optimal Prognosis
- Exclusion of non-identifiable variables
3. Solving of calibration problems as an optimization task
- Definition of proper objectives
- Globale vs. local search
The webinar aims at engineers and developers regularly dealing with the task of model calibration in the field of CAE.