WCCM V -- Authors' Guidelines for Full Paper Submission

In the design of many mechanical systems, the human designer knows its topology and shape. The goal for computational design thus results in finding a parameter set that optimizes multiple design goals. This paper presents a novel optimization method for such parametric design, the process of which is based on the human design process, but the search algorithm of which is gradient-based for efficient computation. In sanction with the human design process, the proposed method first searches for multiple solutions or Pareto-optimal solutions, which correspond to the solution space of the multi-objective optimization problem, directly by minimizing multiple objective functions. The proposed method then finds the final design by the center-of-gravity method proposed by the authors, where the closest to the center-of-gravity of the solutions in parameter space becomes the final design. There exist several evolutionary algorithms that can find Pareto-optimal solutions. The performance of the proposed optimization was investigated with numerical examples and compared to a multi-objective evolutionary algorithm. The numerical results first demonstrate that the proposed method can search Pareto-optimal solutions only after 20 iterations, which could not be obtained by the evolutionary algorithm even after 500 iterations. The final solution obtained also correlated well with the exact robust solution of the problem. Tomonari Furukawa, Shinobu Yoshimura and Hiroshi Kawai