Bond graph aided design of controlled systems

Abstract

An active or controlled system is generally composed of two parts: a passive basis and a control architecture containing actuators and sensors. When dealing with such a system, the first point usually considered is the study of the system without control. To do this, we need a model in order to get simulation-based results on the frequency domain and dynamical behaviour for dimensioning purpose. The second step is then to design a control architecture, with its actuators and sensors, specified in a way allowing the objectives to be reached as accurately and cheaply as possible. Since many years, the bond graph methodology has shown its qualities for modelling and generation of physical insight, specially when applied to multidisciplinary systems. The aim of this paper is to show how a bond graph model may be used for analysis of structural properties, i.e., properties depending only on the model structure and on the type of elements composing it, but not on the numerical values of the parameters. The properties pointed out in this way are generic, and can be used for “integrated design”, i.e., the simultaneous design of the passive system model, its control architecture and control laws for specific aims. The proposed methodology depends on causal manipulations on the bond graph model (assignment of integral and derivative causality, causal path and loops); its application may necessitate a return to the model in order to check and sometimes modify the modelling hypotheses. The proposed procedure is implemented on an example, which will be the guideline of the presentation.