Towards Grey Box Modeling in Modelica

Abstract

A dynamic system model of a motorcycle is described that has been implemented in the modeling language Modelica. Particular mathematical relations therein are learned by feedforward artificial neural networks which are trained using simulation data generated from the original model. These trained models are then used to replace the original relations in new models. This emulates a situation in which the original relations (e.g. those of a real system) are unknown but measurements indicate an interdependence between variables. The simulation results of those derived models are then evaluated and an approach to detect unreliable extrapolation of the artificial neural networks is proposed.