Automated Design of Mechatronic Systems Using Bond Graphs and Genetic Programming

Michigan State University, East Lansing MI

Investigators

Abstract

This research grant supports exploration of potential methods of using genetic programming to allow design of mechatronic systems, based on joining multiport models of devices. The research objectives are: (1) to find a problem representation and set of operators which will, with high probability, allow the genetic programming system to generate multiport models of mixed-domain dynamic systems that satisfy causality constraints, (2) to create efficient ways of evaluating model performance relative to the problem's functional design specifications, (3) to conduct experiments to improve the globality of the search of the design space and the speed with which good designs are found, and (4) to determine the capabilities and limitations of the automated design system created. The desired output of the system will be sets of candidate solutions among which the design engineer can choose according to criteria specific to the implementation domain. The results of this research will ultimately lead to improved design capabilities for new mechatronic devices and systems, and will, in the shorter term, provide guidance for potential implementors of commercial mechatronic design automation systems. These results will be usable across the spectrum of applications of mechatronic systems, in such industries as automotive, computer peripherals, and durable goods (appliances such as washing machines, refrigerators, etc.). An excellent example is the automotive industry, which is greatly expanding its use of mathematical modeling in the design of subsystems and vehicles. Automotive applications for the multiport-modeling/genetic-programming approach will include such areas as vibration reduction, increased efficiency in power transmission, and perhaps also entirely new and patentable applications of mechatronics.