改进的遗传算法用于离心叶轮优化设计

Optimal Design of Centrifugal Impeller From Improved Genetic Algorithm

为了防止遗传算法中过早收敛引起的算法失效 ,提出了一种改进的遗传算法 ,来研究离心叶轮的形状优化问题 .在计算过程中 ,增加了一个自适应算子 ,来不断调节个体变异概率 ,以保证搜索的全局性和种群的多样性 ,同时为了不引入过多的边界参数 ,采用B样条曲线对边界进行描述 .最后 ,对一个数学算例和某离心叶轮进行了计算 ,给出了最终的优化结果和收敛情况 .计算结果表明 ,改进的遗传算法能够很好地适用于各种应力约束和几何约束 ,并可得到形状优化的最优解 ,验证了该方法的合理性和正确性 .

Premature convergence is a problem in genetic algorithm. in order to study shape optimization of centrifugal impeller, an improved genetic algorithm is proposed, which increases an adapting operator to adjust mutation probability. Boundaries of structural components are encoded using B-spline curves, which permit the description of more possible shapes with less parameters. The method was applied to the study of the optimal shape of a centrifugal impeller of compressor. It shows that the improved genetic algorithm can be adapted for a variety of stress constraints. It also shows that this algorithm is reasonable and reliable.