資源の追加と削減に基づく並列分散最適化法の収束性の改善

概要

論文の詳細を見る
The distributed optimization method by resource addition and reduction, which is called the DORAR method, is a new optimization method for the minimization of the total resource of systems, and it has been found to be suitable for parallel processing and very robust for initial solutions. The only parameter in the method is the amount of the small additional resource used for increasing the resources of the elements of a system. It has been found that the amount of the additional resource can be determined from the required accuracy of final results. However, the problem on slow convergence still remains. This paper introduces three improvements for adjusting the additional resource aiming its fast convergence and the high accuracy of the converged solutions. One is the gradual reduction of the additional resource, one is the utilization of the sensitivity information, and one is the adaptation of the additional resource using the history of the resource reduction. These improvements are tested for optimizing truss structures and their effectiveness is examined.
2002-02-25