扩展T-S模糊模型的PSO神经网络优化算法

Optimization algorithm of PSO-BPNN based on extended T-S fuzzy model

针对机械设备具有模糊性和非线性的特点,提出了一种利用扩展T-S模糊模型的,自适应PSO算法和BP神经网络相结合的新型智能结构优化算法。通过自适应的高斯函数来更改基本T-S模糊模型中的隶属度函数,进而使用扩展的T-S模糊模型来调整PSO算法的参数。以BP神经网络隐含层神经元数目为设计变量,提取训练后的均方误差作为评价函数,用改进后的粒子群算法进行寻优。把优化后的网络模型应用于轮盘结构优化中,实验表明,该方法在保证轮盘性能的同时,对其结构进行了重新优化,是一种可行的结构优化方法。

Mechanical equipment with fuzzy and non-linear characteristics,a new structured intelligence method is proposed based on extended T-S（Takagi-Sugeno） fuzzy model of self-adaptive Particle Swarm Optimization（PSO） algorithm and BP Neural Network（BPNN） algorithm.The basic T-S fuzzy model is modified,and the parameter in PSO algorithm is adjusted by extended T-S fuzzy model.The BPNN with hidden layer neurons number of design variables,after extraction of the mean square error as the evaluation function,with the improved particle swarm optimization algorithm.The optimized structure of the network model is applied to optimize the wheelt,he test results show that the method to ensure the performance of wheel at the same timei,ts structure is clear that re-optimization is a feasible method of structural optimization.