Automated Multi-domain Engineering Design through Linear Graph and Genetic Programming

AUMATED MULTI-DOMAIN ENGINEERING DESIGN THOUGH LINEAR GRAPS AND GENETIC PROGRAMMING Eric McCormick*, Haoxiang Lang* and Clarence W. de Silva** *GRASP Laboratory, Department of Automotive and Mechatronics, Ontario Tech University, Oshawa, ON, Canada. Emails: eric.mccormick@ontariotechu.net, haoxiang.lang@ontariotechu.ca **Industrial Automation Laboratory, Department of Mechanical Engineering, The University of British Columbia, Vancouver, BC, Canada. Email: desilva@mech.ubc.ca Abstract This paper proposes a methodology of integrating the Linear Graph (LG) approach with Genetic Programming (GP) for generating an automated multi-domain engineering design approach by using the in-house developed LG MATLAB toolbox and the GP toolbox in MATLAB. The necessary background for the development are presented, and the methodology used in this work to facilitate the construction and evaluation of filter circuits, using LG models, is described. Designing electronic filter circuits through an evolution from electronic components to the completed circuits is demonstrated. The topology and component values of three types of filter circuits: low pass, high pass, and band pass, are designed through this evolutionary approach, for various cut-off frequencies. Furthermore, the paper demonstrates through examples of these evolved filter circuits, the combined GP and LG approach is successful in constructing high order filter circuits that are capable of attenuating undesired frequencies while maintaining desirable ones. The work presented in the paper is a key step towards the integration of LG modeling, through the use of the LGtheory MATLAB Toolbox, with machine learning techniques for the automated design of dynamic multi-domain mechatronic systems.