Design of surface micro-machined inertial MEMS sensor with meta-heuristic topology optimization

Abstract

Micro-Electro-Mechanical Systems (MEMS) are getting popular for a variety of applications due to their small size, low cost, and good performance. MEMS devices are designed by combining simple shapes based on the experience of previous cases. Using mathematical design methods such as Topology Optimization (TO) helps to find the best possible design for a specific application. In a landslide monitoring application, high sensitive MEMS accelerometers are required to accurately predict the occurrence of landslides. But available TO methods are unable to find a design with maximum sensitivity, variable loading condition, and subjected to several constrains. Constraints are because of manufacturing limits in MEMS fabrication and performance limits from the physical problem. The aim of this thesis is to develop a meta-heuristic TO method using Simulated Annealing (SA) to solve non-convex and multi constraint TO problems without gradient information. This TO utilizes crystallization factors to improve the convergence and reduces computational costs in TO. The proposed method is validated with benchmark problems in the literature and it is successfully used for the TO MEMS accelerometers. Analysis of optimization parameters in this design gives some useful information about the convergence and uniqueness of the optimum solution. The optimized designs are then compared to available designs for several performance parameters. Additionally, some filtering and post-processing methods are developed to apply manufacturing limits in the lithography process.