基于粒子群-蛙跳算法优化BP神经网络的滚动轴承故障诊断方法

针对滚动轴承故障诊断方法存在的局限性及缺陷，在利用小波分析提取滚动轴承故障信号特征向量基础上，提出基于粒子群 蛙跳算法优化的BP神经网络滚动轴承故障诊断方法。该方法采用粒子群 蛙跳算法优化BP神经网络结构参数，利用改进BP算法和样本数据训练BP神经网络，实现滚动轴承运行正常和4种不同故障状态的诊断。实验验证结果表明，基于粒子群 蛙跳算法的BP神经网络方法诊断误差最大值仅为005，为未优化的神经网络诊断误差的1/16；与其他算法相比,基于粒子群 蛙跳算法优化的BP神经网络方法的训练时间、训练误差和诊断精度各项指标均为最优，可实现滚动轴承故障的快速、准确、有效诊断。

Rolling Bearing Fault Diagnosis Using Optimized BP Neural Network by Particle Swarm Optimization Leapfrog Algorithm

To avoid the limitations and defects of the fault diagnosis for rolling bearings,a BP neural network method for rolling bearing fault diagnosis is proposed using particle swarm optimization leapfrog algorithm based on extraction of the fault signal feature vector of rolling bearings by wavelet analysis.In this method,particle swarm optimization leapfrog algorithm is used to optimize the structural parameters of BP neural network,and the improved BP algorithm and sample data are then used to train BP neural network to realize the normal operation of rolling bearing and diagnosis of four types of its fault.The experimental results show that BP neural network optimized by particle swarm optimization leapfrog algorithm has a maximum diagnostic error of 0.05, only 1/16 of that of unoptimized neural network,and its training time,training error and diagnostic accuracy are all optimal compared to other algorithms,which ensures the fast,accurates and effective diagnosis of rolling bearing faults.