Towards computer-automated mechatronic design and optimization using linear graphs and evolutionary computing

Abstract

A comparison of literature between the linear graph (LG) and bond graph (BG) approaches shows that the surpassing of BGs to LGs in applications related to the modeling of mechatronic systems is driven primarily by a lack of available LG-based software. As a result, a robust software toolbox called LGtheory has been developed for automating the evaluation of LG models in the MATLAB programming environment. This thesis details the development of LGtheory, and the algorithms and procedures employed for the evaluation of LG models. In addition, demonstrations of this toolbox to a process for automating the design of electronic filter circuits, as well as, the modeling and simulation of the dynamics of a mobile robotic system are presented. The results of these demonstrations validate the accuracy of the LGtheory toolbox for modeling complex multi-domain systems, and provides the methodological basis for the automated design of mechatronic systems using LGs.