Optimization Algorithms for Nonlinear System Identification and Parameter Estimation Problems in Simulation of Mechanical Systems

We are investigating the possibilities of evolutionary algorithm application for modeling of (nonlinear) systems based on real-world measurements and analytical expressions. Particularly the Genetic Programming (GP) approach and other methods from the area of data mining are analyzed. In a second focus we concentrate on methods for optimization of model parameters in order to achieve an optimal fit between a simulation model and its real-world counterpart. In addition to conventional algorithms we focus on population based methods and various versions of genetic programming. An example and proof of concept will be the system calibration of a full-flight simulator.