Data-driven parameter and model identification

The participants carry out their own experiments for parameter and model identification.

• Parameter identification of material models
• Experimental design for robust parameter and model identification
• Optimization methods
• Model identification of material models
• Uncertainty quantification
• Model identification of mechanical systems
• Machine learning as a model-free method
• Physics-informed machine learning

The ability to obtain reliable models of mechanical systems from experimental data is an elementary competence for analyzing, predicting and optimizing real phenomena. The lecture presents methods for determining material parameters and analytical models from experimental data. The focus of the lecture are modern data-driven methods and machine learning methods. A practical hands-on exercise is offered. In the exercise, the students themselves will generate experimental data themselves and apply the thought methods. Planned are - tensile experiments on a material testing machine for the determination of material parameters - the use of contactless deformation measurement to determine material models - the determination of system models of vibrating systems from video files. After successfully completing the module, students will be able to - design and carry out experiments for parameter and model identification - apply data-driven methods such as sparse regression and machine learning and critically evaluate the results - assess when and how model assumptions can be replaced by data-driven methods