A Multicriteria System-Based Method for Simulation-Driven Design Synthesis

Abstract

This paper presents an algorithmic, physics-based, design synthesis method aimed at facilitating synthesis through automated generation of a range of feasible and optimally directed design alternatives. The method assists designers in the exploration of performance limits and trade-offs for synthesis tasks as well as reducing design time through rapid, computational generation. The method introduced combines a multicriteria generate-and-test search algorithm, called Burst, with a Connected Node System (CNS) design representation and provides automatic links to multiphysics simulation for quantitative evaluation of design performance throughout the synthesis process. The CNS-Burst method is applied to two benchmark synthesis tasks in the domain of MEMS to validate the method. The solutions generated meet the modeled design requirements and the variety of designs generated offers designers the possibility of selecting devices according to their preferences among performance trade-offs. The potential for extension to larger, more complex MEMS design synthesis and optimization tasks is discussed.