Multi-domain physical system modeling and control based on meta-modeling and graph rewriting

Abstract

A methodology is presented which enables the specification and synthesis of software tools to aid in plant and controller modeling for multi-domain (electrical, mechanical, hydraulic, and thermal) physical systems. The methodology is based on meta-modeling and graph rewriting. The plant is modeled in a domain-specific formalism called the Real World Visual Model (RWVM). Such a model is successively transformed to an Idealized Physical Model (IPM), to an Acausal Bond Graph (ABG), and finally to a Causal Bond Graph (CBG). A Modelica (www.modelica.org) model, consisting of a Causal (algebraic and differential equation) Block Diagram (CBD), is generated from the CBG. All transformations are explicitly modeled using Graph Grammars. A PID controller model, specified in Modelica as a CBD is subsequently integrated with the plant model. AToM3 (atom3.cs.mcgill.ca), A Tool for Multi-formalism and Meta Modeling is used to meta-model and synthesize visual modeling environments for the RWVM, IPM, ABG, and CBG formalisms as well as for transformations between them. The entire process of modeling, transformation, and simulation is demonstrated by means of a hoisting device example. Our methodology drastically reduces development time (of the modeling tool an indirectly of the domain-specific models), integrates model checking via Bond Graph causal analysis, and facilitates management and reuse of meta-knowledge by explicitly modeling formalisms and transformations.