基于粒子群神经网络的轮盘优化

Optimization of turbine disk based on particle swarm optimization and neural network

将粒子群算法(PSO)和BP神经网络相结合,构建了一种新型智能结构优化算法.PSO方法除用于结构优化外,还被用于BP神经网络的构造及网络训练,使之可自适应调整优化.结构优化中,以BP神经网络取代有限元方法,通过设计变量来映射目标函数和约束,从而大大提高了计算速度.将此方法用于轮盘结构优化,使得轮盘体积减少了17.5%,结果通过检验.该方法便捷、高效,为解决工程结构优化问题提供了一个新途径.

A new computational intelligence method was established by combining particle swarm optimization （PSO） and back propagating （BP） neural network. In addition to structural optimization,PSO is also used in BP neural network＇s construction and training, thus enabling self-adaptive optimization. In the structural optimization, the neural networks were used to map object function and constraints instead of finite element method （FEM）, helping to accelerate greatly the computation speed. A disk model was optimized to decrease its volume by 17.5%. The results show that, this convenient and efficient method will provide a new approach to structural optimization.