3128 MOGAおよびKriging法を用いたCF/GF風車翼構造の複数目標最適化(J03-1 解析・設計の高度化・最適化(1),J03 解析・設計の高度化・最適化)

抄録

Large wind tuebine blade is prospected to be CF/GF hybrid blade made with Carbon Fiber Reinforsced Plastic and Glass Fiber Reinforced Plastic, and structural optimization of the blade is multi-objective optimization problem (MOOP) about weight and material cost of blades. The optimum solutions for MOOP are Pareto optimum solutions. In this paper, the MOOP is solved using Multi-Objectibe Genetic Algorithm (MOGA). Meanwhile, blade structure is necessary to satisfy some constraints not to fracture under various load conditions. To evaluate the constraints, FEM analysis is used. This causes enormous calculation cost. In the present paper, to reduce this, some response surfaces approximating the degree of constraint satisfaction are created. Pareto solutions obtained using response surfaces don't necessarily satisfy the constraints because of approximation error. It is necessary to confirm whether they actually satisfy the constraints or not using FEM analysis. If some of the solutions don't satisfy the constraints, whose response data is added in, and the response surfaces are restructured to improve their accuracy. Repeating restructuring, obtained Pareto solutions will converge to exact Pareto Optima. As a result, the the number of FEM analysis is reduced to about 1300 times.