Abstract: To improve the optimization design quality of tandem blade，an automatic optimization system of tandem blade was developed based on Modified Particle Swarm Optimization Algorithm (MPSO). The mechanism of original PSO was deeply researched and a modified method of PSO was presented. The performance of MPSO was compared with genetic Algorithm and original PSO. The comparison indicates MPSO can obtain better convergence speed and precision. An optimization system is developed based on BP neural network and MPSO. The optimization system is validated by optimizing a tandem blade of 50° blade turning. Both parameters of blade and parameters of relative position of tandem blade are treated as optimization variables. Results indicate that the total pressure loss coefficient of optimized blade decreases by 22% and the static pressure ratio increases by 0.6% at design incidence. The flow performance of optimized blade is improved at negative incidence. Decreasing the radius of leading and trailing edge properly can reduce the loss of tandem blade. Rational structure of the slot can effectively reduce separation loss of boundary layer on pressure surface of front blade and suction surface of rear blade.

Key words: Tandem blade；Blade optimization；BP neural network；Particle swarm algorithm

摘要： 为了提高串列叶型优化设计的质量，设计了一套基于改进粒子群算法的串列叶型自动优化系统。研究了原始粒子群算法，提出了一种粒子群算法的改进方法。结果发现，改进粒子群算法的收敛速度和收敛精度明显优于原始粒子群算法和遗传算法。以50°大弯角串列叶型为研究对象，使用程序对串列叶型参数化。以叶型参数和串列叶型相对位置的参数作为优化变量，结合BP神经网络和改进粒子群算法对串列叶型进行优化设计。优化结果表明，优化后的叶型在设计攻角时的总压损失系数降低了22%，静压比升高了0.6%，在负攻角时，优化后的叶型的流动性能得到了明显改善。适当减小串列叶型前后缘的半径可以减小叶型的损失，合理的缝隙结构可以有效减小前排叶型压力面和后排叶型吸力面附面层的分离损失。

关键词: 串列叶型；叶型优化；BP神经网络；粒子群算法