Yasunari Mimura

Development of Multi-Stage and Hybrid Real-Coded Genetic Algorithms With Active Constraints

In this study, we propose multi-stage and hybrid real-coded genetic algorithm. In the proposed algorithm, there are two stages. In the first stage, Real-coded Genetic Algorithm with Active Constraints (RGAAC) is applied to find a solution that is close to the global optimum. In RGAAC, indviduals who are out of the feasible region are pulled back into feasible region. Therefore, the effective search can be carried out even in the constraints problems. In the second stage, Feasible Region Limiting Method (FRLM) is applied to obtain an optimum solution. FRLM uses the solution that is derived in the first stager as an initial point. In this study, RGAAC is applied to solve the truss structure problems. Through these problems, the effectiveness and the searching mechanism of RGAAC is discussed. The, the proposed algorithm is also applied to 2D problem. In this problem, there are about 1000 design variables. The proposed method can derive the reasonable solution. From these results, it is concluded that the proposed method is effective to solve optimzation problems of large scale structures.© 2002 ASME

A System for Optimization of Structural Design of Artifacts on the Grid

This paper describes the development of a robust and efficient system for optimization of structural design of artifacts on the Grid, considering the following requirements : (1) Computation subtasks with less dependency are processed asynchronously, and (2) An optimal solution with specified accuracy is obtained in a robust and prompt manner. To satisfy Req. (1), evaluation calculations such as finite element (FE) structural analyses are asynchronously performed on the Grid. Each FE analysis is performed using an open source parallel FE system named ADVENTURE. To satisfy Req. (2), a robust and efficient optimization algorithm based on real-coded genetic algorithms is developed. In addition, robust processing mechanisms are implemented to perform a whole of computation on the Grid without interruption. In this study, we test the following two kinds of Grid environments : a group of PC clusters distributed in remote sites and one of Japanese national Grid projects named ITBL (IT-based Laboratory). Practical performances of the developed system are demonstrated through some examples.

A Human-Like Optimization Method for Constrained Parametric Design

This paper describes a numerical technique that enables the optimal parametric design of an engineering system in a human-like manner. Thanks to the human-like process in design, the proposed method can handle design problems that have multiple objectives. Moreover, the proposed method can find solutions subject to constraints. The effectiveness of the proposed method was confirmed by solving constrained multi-objective optimization problems where the exact solution is known. The proposed method was further applied to the design of a micro-accelerometer, and optimal solutions, which well satisfy constraints, were successfully found.