改进PSO算法的模糊神经网络设备故障诊断  

Fuzzy Neural Network Equipment Fault Diagnosis Based on Improved PSO Algorithm

作　　者：侯国安[1] 孔政敏[2] 肖东彩[1] HOU Guo-an;KONG Zheng-min;XIAO;Dong-cai(Yinchuan Energy College,Yinchuan Ningxia 750015,China;Wuhan University,Wuhan Hubei 430072,China)

机构地区：[1]银川能源学院,宁夏银川750015 [2]武汉大学,湖北武汉430072

出　　处：《计算机仿真》2024年第12期531-535,共5页Computer Simulation

基　　金：国家自然科学基金(62173256);宁夏回族自治区自然科学基金(2020A0363)。

摘　　要：对于一些多优化问题,T-S模糊神经网络具有精确性不高、容易进入局部最优值、收敛速度不快等不足。为了弥补以上不足,提出了将模糊神经网络和粒子群算法相融合,为了优化T-S模糊神经网络的模型参数,进一步改进粒子群算法的学习因子和惯性权重,创新了算法模型。通过改进粒子群算法,可以使迭代过程中自适应地对收缩和扩张系数进行调节,让新模型更具动态自适应性。将上述模型应用到旋转机械设备故障诊断中,测试创新算法的优劣。通过与BP神经网络、PSO-BP网络仿真结果相对比,改进PSO-BP网络具有更高的精度,在旋转机械故障诊断中具有良好的应用效果。For some multi-optimization problems,T-S fuzzy neural network has some shortcomings,such as low accuracy,easy to enter local optimal value and slow convergence speed.In order to make up for these shortcomings,a fusion of fuzzy neural networks and particle swarm optimization(PSO)is proposed.By improving PSO,the shrinkage and expansion coefficients can be adjusted adaptively in the iterative process,making the new model more dynamically adaptive.Compared with the simulation results of the BP neural network and PSO-BP network,the improved PSOBP network has higher accuracy and good application effect in rotating machinery fault diagnosis.

关 键 词：模糊神经网络 旋转机械 故障诊断 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]