基于Internet网络的主从式并行遗传算法及其在叶栅优化中的应用

为了解决遗传算法在优化中由于适应度评价很费时而导致计算时间过长的问题 ,本文发展了一种基于In ternet网络实现的主从式并行遗传算法。在函数优化的测试实验中 ,通过控制待优化函数适应度评价的时间 ,验证了主从式模型在适应度评价很费时且远远超过通讯时间时将获得接近于线性的加速比 ,讨论了主从式并行遗传算法应用于气动性能优化中的可行性。通过二维叶栅的优化算例 ,证明了本文提出的算法适合于需要大计算资源的叶栅气动优化设计

A Master-Server Parallel Genetic Algorithm Based on Internet and Its Application to Optimization Design of Blades

Long computation time due to fitness evaluation is known as a big problem in genetic algorithm (GA) application. To find a solution to this problem, this paper developed a master-server parallel genetic algorithm (MSPGA) based on Internet. In the numerical test of a function optimization, the time-cost of fitness evaluation of function is controlled to be optimized, since time-cost of fitness evaluation is far greater than that of data communication. It is demonstrated that the MSPGA get an acceleration rate which shows a nearly linearity. The feasibility of the MSPGA is also discussed, and it was found that the MSPAG can be used in the aerodynamic shape optimization of blade design. Finally presented are the examples of two-dimensional calculations on blade optimization, and the results demonstrate that the method presented is suitable for the aerodynamic shape optimization of blade design which need both huge computation resource and time.