Cyber-physical modelling in Modelica with model-reduction techniques

Abstract

Object-oriented modelling of cyber-physical systems with Modelica and similar environments has brought many advantages, especially the efficient re-use of models and thus the possibility of creating powerful multi-domain libraries. Unfortunately, the models have become highly complex, which causes serious problems during processing and execution. Consequently, verification and debugging is becoming an increasingly challenging task. The continuous investigation of simplifications and reductions in all phases of model developments is thus urgent.The present paper deals with reduction methods based on metric ranking and preserve realisation, which means that the structure and the parameters of the model remain physically interpretable. Two model-reduction methods are described and implemented in Open Modelica. The first operates on a set of differential-algebraic equations, and the second is based on modified bond-graphs-reduction techniques. The latter approach is suitable for component-based models in Modelica that are usually represented graphically with object diagrams. The paper briefly describes the research area, the problems of the adoption of the developed model reduction techniques to the Modelica environments, and the final implementation. Both proposed approaches are tested on the model of a car suspension system and briefly discussed.