Sequential Approximate Optimization Procedure based on Sample-reusable Moving Least Squares Meta-model and its Application to Design Optimizations
Jin Yeon Cho; Min Hwan Oh

doi:10.3970/cmes.2010.066.187
Source CMES: Computer Modeling in Engineering & Sciences, Vol. 66, No. 3, pp. 187-214, 2010
Download Full length paper in PDF format. Size = 2,796,534 bytes
Keywords sequential approximate optimization(SAO), sample-reusable meta-model, moving least squares(MLS), trust region algorithm, response surface.
Abstract In this work, a sample-reusable sequential approximate optimization (SAO) procedure is suggested. The suggested sequential approximate optimization procedure utilizes a newly proposed sample-reusable meta-model along with the trust region algorithm. Domain of design is sequentially updated to search for the optimal solution through the trust region algorithm, and the system response in the updated design region at each sequential stage is approximated by the proposed sample-reusable meta-model. The proposed sample-reusable meta-model is based on the moving least squares(MLS) approximation scheme. Thanks to the merits of moving least squares scheme, the proposed meta-model can fully utilize the previously sampled responses as well as the currently sampled responses of the system, and consequently it makes it possible to enhance the accuracy and robustness of meta-model (often called response surface) for system response.
Through the typical optimization problems, the performance of proposed approach is investigated. After the investigations, a preliminary optimal design of compound helicopter is carried out by using the proposed sample-reusable sequential approximate optimization procedure as a practical example of design optimization.
PDF download PDF