Design window search using continuous evolutionary algorithm and clustering––its application to shape design of microelectrostatic actuator

Abstract

This paper describes a search method for multi-dimensional design window (DW), which is defined as an existing area of satisfactory solutions in a design parameter space. The method consists of the following two steps: (1) direct search for satisfactory design solutions using a continuous evolutionary algorithm (CEA) from a wide area of the design parameter space in a robust manner, and (2) identification of a precise structure of DW by clustering the detected satisfactory solutions with a modified K-means algorithm. The CEA is a kind of genetic algorithms modified to deal with continuous variables. The modified K-means clustering algorithm contains an explicit procedure to determine the optimum number of clusters. The proposed DW search method was implemented to an integrated computational aided engineering system for multi-disciplinary structural design developed by the authors, and then the method was applied to shape design of a microelectrostatic actuator for next generation high-density optical memory. A DW consisting of four clusters, i.e. four sub-DWs was obtained, and the features of the representative design solutions of the four sub-DWs were compared with each other in detail, and a final design solution was determined.