基于IBA优化BP神经网络的滚动轴承故障诊断方法
|
摘要
以反向传播神经网络为基础,引入改进的蝙蝠算法对其初始阈值和权值进行优化处理,并针对滚动轴承信号的特征针对性构建了故障诊断系统。针对轴承数据的振动信号选取时、频、多尺度排列熵等提取方式进行多特征参量提取,构造了滚动轴承正常及故障状态下的特征样本并对优化后的神经网络进行训练。然后,使用训练完成的网络对各状态下的随机样本进行诊断测试,诊断结果表明,本文构建的神经网络系统与未优化的BP神经网络相比,可以更为准确地识别出滚动轴承的故障类型,误差降低约一个量级,与未改进的优化算法相比,所介绍的改进算法在保证精度的同时可以有效增加算法的优化效率,同时对强噪音环境下的缺陷具有更高的鉴别率,更高的实用价值。
Based on back-propagation neural network(BPNN), this paper introduced an improved bat algorithm(IBA) to optimize its initial threshold and weight, and constructed a fault diagnosis system for rolling bearing signals. Finally, the training matrix of normal and fault states of bearing was built up with entropy, time domain, time-frequency characteristics of data. After that, the fault state of bearing has been diagnosed using the above network which has been trained well. The results shown that the optimized neural network system can accurately identify the fault type of rolling bearing and the diagnosis result of IBA-BPNN is better than neural network which is unoptimized. The optimized neural network also has higher recognition rate for defects under strong noise environment and has higher practical value than other networks.
Based on back-propagation neural network(BPNN), this paper introduced an improved bat algorithm(IBA) to optimize its initial threshold and weight, and constructed a fault diagnosis system for rolling bearing signals. Finally, the training matrix of normal and fault states of bearing was built up with entropy, time domain, time-frequency characteristics of data. After that, the fault state of bearing has been diagnosed using the above network which has been trained well. The results shown that the optimized neural network system can accurately identify the fault type of rolling bearing and the diagnosis result of IBA-BPNN is better than neural network which is unoptimized. The optimized neural network also has higher recognition rate for defects under strong noise environment and has higher practical value than other networks.
引文
[1] 吴春晓,行鸿彦,张漪俊.基于BP神经网络的地温推演模型[J].电子测量与仪器学报,2017,31(10):1561-1567.
[2] 褚继花.遗传算法优化BP神经网络水文预报过程模型研究[J].水利规划与设计,2018,1:65-66,118.
[3] 丁硕,巫庆辉,常晓恒,等.基于灰色BP神经网络的实验材料供应预测[J].国外电子测量技术,2016,35(12):78-82.
[4] CAI M.Global optimization algorithm for BP neural networks[J].Engineering Journal of Wuhan University,2013,46(6):794-810.
[5] 邵婷婷,张博超,周美丽,等.基于RBF神经网络的测斜仪方位角校正研究[J].国外电子测量技术,2016,35(2):77-79.
[6] 朱晓青,马定寰,李圣清,等.基于BP神经网络的微电网蓄电池荷电状态估计[J].电子测量与仪器学报,2017,31(12):2042-2048.
[7] 马峻,赵飞乐,徐潇,等.MRA-PCA-PSO组合优化BP神经网络模拟电路故障诊断研究[J].电子测量与仪器学报,2018,32(3):73-79.
[8] 李小珉,尹明.基于遗传算法的BP神经网络电子系统状态预测方法研究[J].电子测量技术,2016,39(9):182-186.
[9] YANG X S.Bat Algorithms[J].Nature-Inspired Optimization Algorithms,2014:141-154.
[10] 李巍华,戴炳雄,张绍辉.基于小波包熵和高斯混合模型的轴承性能退化评估[J].振动与冲击,2013,32(21):35-40,91.
[11] ZHENG J D.Generalized composite multiscale permutation entropy and Laplacian score based rolling bearing fault diagnosis[J].Mechanical Systems and Signal Processing,2018(15):229-243.
[12] 章浙涛,朱建军,匡翠林,等.小波包多阈值去噪法及其在形变分析中的应用[J].测绘学报,2014,43(1):13-20.
[13] 梁含笑,高晓蓉,邱春蓉.基于脉冲相关性的时域增强EMD轴承故障诊断[J].信息技术,2017(11):17-21.
[14] 陈远鸣,常建华,沈婉,等.基于改进型BP神经网络的SF6气体传感器[J].电子测量与仪器学报,2017,31(10):1582-1588.
[15] 张达敏,林辉品,林智勇,等.基于神经网络预测控制的节能电梯能量管理[J].仪器仪表学报,2017,(12):3137-3142.
[2] 褚继花.遗传算法优化BP神经网络水文预报过程模型研究[J].水利规划与设计,2018,1:65-66,118.
[3] 丁硕,巫庆辉,常晓恒,等.基于灰色BP神经网络的实验材料供应预测[J].国外电子测量技术,2016,35(12):78-82.
[4] CAI M.Global optimization algorithm for BP neural networks[J].Engineering Journal of Wuhan University,2013,46(6):794-810.
[5] 邵婷婷,张博超,周美丽,等.基于RBF神经网络的测斜仪方位角校正研究[J].国外电子测量技术,2016,35(2):77-79.
[6] 朱晓青,马定寰,李圣清,等.基于BP神经网络的微电网蓄电池荷电状态估计[J].电子测量与仪器学报,2017,31(12):2042-2048.
[7] 马峻,赵飞乐,徐潇,等.MRA-PCA-PSO组合优化BP神经网络模拟电路故障诊断研究[J].电子测量与仪器学报,2018,32(3):73-79.
[8] 李小珉,尹明.基于遗传算法的BP神经网络电子系统状态预测方法研究[J].电子测量技术,2016,39(9):182-186.
[9] YANG X S.Bat Algorithms[J].Nature-Inspired Optimization Algorithms,2014:141-154.
[10] 李巍华,戴炳雄,张绍辉.基于小波包熵和高斯混合模型的轴承性能退化评估[J].振动与冲击,2013,32(21):35-40,91.
[11] ZHENG J D.Generalized composite multiscale permutation entropy and Laplacian score based rolling bearing fault diagnosis[J].Mechanical Systems and Signal Processing,2018(15):229-243.
[12] 章浙涛,朱建军,匡翠林,等.小波包多阈值去噪法及其在形变分析中的应用[J].测绘学报,2014,43(1):13-20.
[13] 梁含笑,高晓蓉,邱春蓉.基于脉冲相关性的时域增强EMD轴承故障诊断[J].信息技术,2017(11):17-21.
[14] 陈远鸣,常建华,沈婉,等.基于改进型BP神经网络的SF6气体传感器[J].电子测量与仪器学报,2017,31(10):1582-1588.
[15] 张达敏,林辉品,林智勇,等.基于神经网络预测控制的节能电梯能量管理[J].仪器仪表学报,2017,(12):3137-3142.