基于小波包-AR谱和GA-BP网络的轴承故障诊断研究
|
摘要
针对轴承振动信号具有非平稳、非线性特点,提出将小波包-AR谱和采用遗传算法(genetic algorithm,GA)优化的BP神经网络相结合的轴承故障诊断方法。该文对滚动轴承振动信号进行小波包分解和自回归(auto-regressive,AR)谱分析以得到不同频段的能量,然后将提取到的特征向量输入到BP神经网络进行模型训练和测试。鉴于BP神经网络的诊断效果并不是很好,因此应用遗传算法对BP神经网络的权值和阈值进行优化并再次进行诊断。对比实验结果表明,经遗传算法优化后的BP神经网络的仿真误差大大降低,相关故障诊断准确率达到了100%。
In view of the non-stationary and non-linear characteristics of bearing vibration signals, a bearing fault diagnosis method combining the wavelet packet-AR spectrum with genetic algorithm(GA) BP neural network is proposed.In this paper,wavelet packet decomposition and auto-regressive(AR) spectrum analysis were performed on the vibration signals of rolling bearings to obtain energy of different frequency bands.The extracted feature vectors were input into BP neural network for model training and testing. Considering that the diagnosis effect of BP neural network is not very good, the weight and threshold value of BP neural network are optimized by genetic algorithm and diagnosed again.The experimental results show that the simulation error of BP neural network optimized by genetic algorithm is greatly reduced, and the relevant fault diagnosis accuracy rate reaches 100%.
In view of the non-stationary and non-linear characteristics of bearing vibration signals, a bearing fault diagnosis method combining the wavelet packet-AR spectrum with genetic algorithm(GA) BP neural network is proposed.In this paper,wavelet packet decomposition and auto-regressive(AR) spectrum analysis were performed on the vibration signals of rolling bearings to obtain energy of different frequency bands.The extracted feature vectors were input into BP neural network for model training and testing. Considering that the diagnosis effect of BP neural network is not very good, the weight and threshold value of BP neural network are optimized by genetic algorithm and diagnosed again.The experimental results show that the simulation error of BP neural network optimized by genetic algorithm is greatly reduced, and the relevant fault diagnosis accuracy rate reaches 100%.
引文
[1] 孟文俊,徐光华,姜阔胜,等.基于LabVIEW的滚动轴承非平稳过程监测诊断及性能评估系统的开发[J].工业仪表与自动化装置,2015(02):18-22.
[2] Wei Z,Wang Y,He S,et al.A novel intelligent method for bearing fault diagnosis based on affinity propagation clustering and adaptive feature selection[J].Knowledge- Based Systems,2017,116(C):1-12.
[3] 常春,王国威,梅检民,等.基于小波包-AR谱和支持向量机的连杆轴承故障诊断[J].军事交通学院学报,2015,17(04):40-44.
[4] 李国勇,杨丽娟.神经·模糊·预测控制及其MATLAB实现[M].北京:电子工业出版社,2013:17-26.
[5] Li X,Xiang S,Zhu P,et al.Establishing a Dynamic Self- Adaptation Learning Algorithm of the BP Neural Network and Its Applications[J].International Journal of Bifurcation & Chaos,2015,25(14):1540030.
[6] 张玲玲,赵懿冠,肖云魁,等.基于小波包-AR谱的变速器轴承故障特征提取[J].振动.测试与诊断,2011,31(04):492-495+537.
[7] 肖云魁,李世义,王建新,等.基于小波包-AR谱技术提取柴油发动机曲轴轴承故障特征[J].北京理工大学学报,2004(06):508-511.
[8] 赵永标,张其林,康长青.基于GA-BP算法的水电机组故障诊断模型[J].洛阳理工学院学报(自然科学版),2011,21(01):41-43+52.
[9] Jin M,Li R,Xu Z,et al.Reliable fault diagnosis method using ensemble fuzzy ARTMAP based on improved Bayesian belief method[J].Neurocomputing,2014,133:309-316.
[10] The case western reserve university bearing data center [EB/OL].(2012-11-07) [2013-12-01].http://csegroups.case.edu/bearingdatacenter/pages/download-data-file.
[2] Wei Z,Wang Y,He S,et al.A novel intelligent method for bearing fault diagnosis based on affinity propagation clustering and adaptive feature selection[J].Knowledge- Based Systems,2017,116(C):1-12.
[3] 常春,王国威,梅检民,等.基于小波包-AR谱和支持向量机的连杆轴承故障诊断[J].军事交通学院学报,2015,17(04):40-44.
[4] 李国勇,杨丽娟.神经·模糊·预测控制及其MATLAB实现[M].北京:电子工业出版社,2013:17-26.
[5] Li X,Xiang S,Zhu P,et al.Establishing a Dynamic Self- Adaptation Learning Algorithm of the BP Neural Network and Its Applications[J].International Journal of Bifurcation & Chaos,2015,25(14):1540030.
[6] 张玲玲,赵懿冠,肖云魁,等.基于小波包-AR谱的变速器轴承故障特征提取[J].振动.测试与诊断,2011,31(04):492-495+537.
[7] 肖云魁,李世义,王建新,等.基于小波包-AR谱技术提取柴油发动机曲轴轴承故障特征[J].北京理工大学学报,2004(06):508-511.
[8] 赵永标,张其林,康长青.基于GA-BP算法的水电机组故障诊断模型[J].洛阳理工学院学报(自然科学版),2011,21(01):41-43+52.
[9] Jin M,Li R,Xu Z,et al.Reliable fault diagnosis method using ensemble fuzzy ARTMAP based on improved Bayesian belief method[J].Neurocomputing,2014,133:309-316.
[10] The case western reserve university bearing data center [EB/OL].(2012-11-07) [2013-12-01].http://csegroups.case.edu/bearingdatacenter/pages/download-data-file.