Methods for dynamic modeling of nonlinear mechatronic systems

M.Sc. Matthias Kahl

April 2013 - March 2017

Industry

For the development, analysis and optimization of modern technical systems model-based methods are used increasingly. Modelling a technical system requires a lot of effort. Especially when used for simulation, high demands are made on the models regarding approximation accuracy and computational complexity. Simultaneously, the system boundaries become broader due to the increasing demands on modern technical systems. Building a model for such a system based on physical laws is therefore becoming more complex and requires a high degree of application-specific knowledge. An alternative to this so-called physical modeling is the data-driven modeling, which enables to generate models with high approximation quality, even without deep knowledge about the system to be modeled, and to automate the modeling process in many parts.

Hence, in this project methods for data-driven dynamic modeling will be developed which allow a higher degree of automation of the identification process. The target category of systems to be modeled are mechatronic multivariable systems with complex dynamic behavior. For this purpose the following sub-problems of the identification of nonlinear dynamic models will be treated:

• extensively automated selection of informative data segments for system identification from normal system operation,
• systematic and semi-automated structure selection,
• deployment of functions to support data acquisition.