Component-Based Design of Simulation Models Utilizing Bond-Graph Theory

Abstract

Simulation models are becoming efficient tools for real plant testing and control system fine-tuning. They are necessary for advanced process control. Since a design phase of simulation models is time-consuming and error-prone, this paper proposes a method for the systematic and semi-automated design of simulation models. It is intended mainly for signal-oriented simulators such as MATLAB-Simulink. The method presumes that large-scale industrial systems consist of subsystems, whose behavior is known. The task for the proposed approach is to create a simulation model from atomic simulation blocks approximating system components. The proposed solution utilizes the bond graph theory in a non-traditional way. Compared to the classical use of bond graphs, the proposed method does not create simulations from scratch, but it re-uses existing simulation components. It extends bond graphs with a concept of component implementations and annotations of their interfaces. Thus, models are assembled from atomic component implementations, whose selection is done by the proposed method for each node of the system topology.