Modeling, identification and simulation of mechatronic systems

Abstract

The integration of mechanical and microelectronic components during the design and operation requires systematic methods and software tools for theoretical modeling, experimental identification and simulation. Therefore, the contribution summarizes the state-of-the-art in these areas based on the practical experience with several mechatronic systems and tries to show open problems. It is typical for mechatronic systems that they are composed of components from different physical domains, such as electrical mechanical, fluidic, thermal and thermodynamic processes. One of the key issues, therefore, is to find a unified way especially for theoretical modeling, but also for identification and simulation of these heterogeneous systems. A further task is to develop software tools for modeling, identification and simulation and to make them interconnectable. For theoretical modeling a unified procedure can be given for lumped parameter processes with energy and material flows. A multi port representation can be stated for processes with energy flows for electrical, mechanical and fluidic processes. However, problems with a unified representation arise for thermal and thermodynamic processes. The development from multi ports to block diagrams and bond graphs is shortly described. Present developments are characterized by object-oriented modeling with introduction of causalities at a late state. Identification methods for mechatronic systems are required frequently e.g. for theoretical model verification or on-line estimation of unknown and time varying parameters or identification of nonlinear static characteristics and dynamics. The properties of suitable methods are described, including parameter estimation and dynamical neural networks for nonlinear processes. Then, various kinds of simulation methods for mechanical systems are described, especially with regard to the time behaviour. For several applications real-time simulation is required. Examples are process simulation for hardware-in-the-loop simulation together with real control components, and controller simulation for control prototyping. Application examples are shown throughout the paper, e.g. identification of an turbocharger with dynamical neural networks and real-time simulation of a combustion engine.