MEMS Design Synthesis and Optimization Corie L Cobb

MEMS Design Synthesis and Optimization Corie L. Cobb Ying Zhang Prof. Alice M. Agogino Prof. Carlo H. Séquin KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

Project Goals v Create design synthesis tools for MEMS devices v Develop rapid, optimal configurations for a given set of performance and constraint guidelines v Create a case library which will include useful MEMS GA (Genetic Algorithm) building blocks and serve as a design tool for the MEMS community KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

Why Automated Synthesis for MEMS? v Currently, MEMS design still relies on human knowledge and experience v Necessary for increased design performance and complexity of MEMS KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

Hierarchical MEMS Synthesis Architecture v A hierarchical MEMS synthesis and optimization architecture has been developed. The architecture integrates: Ø an object-oriented component library with the MEMS simulation tool SUGAR Ø two levels of optimization: global genetic algorithms (GA) and local gradient based refinement KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

MEMS Design Synthesis Flow Chart Design Objectives Constraints, Stop Criteria Fabricated & Validated Designs Knowledge Base and Component Library Casebased Indexed Design Library Fabrication & Simulation Constraints Evaluation, Validation & Refinement of Design Rules Design Encoding Scheme: object elements & clusters Fabrication, Testing & Characterization Finite Element Analyses (FEM) Evolutionary Synthesis Initial Designs Algorithm: genetic selection, mutation, Pareto-ranking based on performance Simulation Engine KSJP 27/JD © 2005 University of California Local Optimization, & Human Refinement Synthesized Designs Design Synthesis Module Prepublication Data Spring 2005

v Stochastic Optimization – Multi-Objective Genetic Algorithm (MOGA) ØCan handle configuration optimization as well as parametric optimization. ØEnables exploration of non-traditional designs (i. e. nonsymmetric, non-Manhattan geometry) Fig 1. Resonators from a test run of MOGA. (Left : non-symmetric polyline springs with unconstrained angle, Middle: xy symmetric polyline spring legs with unconstrained angle, Right : xy symmetric resonator with mixed springs. ) KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

v Each MEMS component in the object-oriented data structure is represented as an entity with its own geometrical parameters, connectivity, and restrictions Ø Each component has flags for mutation, crossover and symmetry Fig 2. Example of Resonator Composition KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

v. Advantages of MEMS Synthesis Architecture: Ø Component library integrated with GA can find promising conceptual designs and topologies for MEMS resonators. Ø Can easily assemble and configure new design problems Ø Gradient descent optimization performed at the end of the design synthesis process is an effective way to refine the most promising design candidates KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

Case-Based Design Library v Will utilize Case-Based Reasoning to: Ø Reduce the need and time to create a MEMS design ‘from scratch’ each time Ø Allow the reuse and modification of previous successful designs to help deal with the complexities of a current design problem v Designs have been extracted from MEMS related journals, articles, and other forms of design documentation v Suspension variants and resonator design cases have been recreated using the SUGAR netlist format v A case library of MEMS resonators is currently being developed with a text based search KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005

Case-Based Reasoning Cycle for MEMS Synthesis Input Design Specifications CASE LIBRARY • MEMS devices • GA Building Blocks (Clusters & Primitives) Add to Case Library Test and Evaluate Retrieve Relevant Cases Adapt Cases Based on Current Problem KSJP 27/JD © 2005 University of California Optimize and Simulate Layout & Fabricate New Designs Prepublication Data Spring 2005

Future Work v Create an initial working database of MEMS designs and extend the case library to include more MEMS devices v Extend the data structures for the component library to enable the modeling of more complex MEMS designs KSJP 27/JD © 2005 University of California Prepublication Data Spring 2005
- Slides: 11