Multi-level and multidisciplinary optimisation of microelectromechanical systems

Abstract

A comparative investigation into the role multi-level and multidisciplinary design optimisation can play in the automated design synthesis of microelectromechanical systems (MEMS) is presented. Microelectromechanical systems are a field grown out of the integrated circuit industry, with the goal of developing smart micro devices which can interact with the environment in some form. They promise to revolutionise our present day lifestyles as much as the integrated circuit has done in recent decades. The complexity in fabrication, the delicacy in size that each device encompasses and the multidisciplinary nature means design synthesis is a highly complicated process. Current challenges stemming from their design include the high levels of computational cost required in their modeling and analysis, and the often increasing complexity of design through the coupling of multiple components and devices into a functioning system. The development of automated design synthesis tools and methodologies to aid MEMS design is therefore important to overcome these challenges in order to accommodate the growing field of MEMS as it expands into more and more areas and continues opening up to more and more applications. An update of the current state of the art in automated MEMS design synthesis and optimisation is first presented, utilizing state of the art multi-objective evolutionary algorithms over five separate MEMS design optimisation case studies. The field of multilevel and multidisciplinary optimisation is critically reviewed and discussed with respect to their application to MEMS design synthesis and optimisation. The outcome is twofold, with the construction of both a novel multidisciplinary optimisation algorithm tailored towards MEMS design and a set of multi-level design optimisation strategies. This thesis next outlines and develops a novel modular soft computing framework to house the multi-objective, multi-level and multidisciplinary design optimisation strategies. In order to evaluate both the current state of the art in automated MEMS design synthesis and the multi-level and multidisciplinary optimisation strategies outlined a hierarchical MEMS bandpass filter case study has been constructed. Incorporating a novel state of the art electrical equivalent modelling and design synthesis approach, six novel design problems structured around the MEMS bandpass filter were developed and formed the basis for the comparative study to follow. Finally both the current state of the art in automated MEMS design synthesis, multiobjective evolutionary algorithms, and the outlined and developed multi-level and multidisciplinary optimisation strategies are applied to the six design problems developed. Comparative analysis and discussion is then given, showing a marked improvement in MEMS design synthesis for the multi-level and multidisciplinary optimisation strategies over the current state of the art methodology.