摘要
针对RV减速器实际监测中振动传感器的安装空间和信号采集容易受到限制和干扰等问题,提出了一种基于电机电流信号分析,稀疏自编码和Fisher准则相结合的RV减速器故障特征提取方法。首先,将采集的驱动电机电流数据转换到频域,研究了不同超参数对稀疏自编码的特征提取能力的影响,利用优化参数后的稀疏自编码对频域信号自动提取故障特征;然后,利用Fisher准则对提取的特征的判别能力进行了降序排名,取排名前n个特征,得到了最优故障特征集;最后,结合SoftMax分类层,实现了对RV减速器的故障诊断;搭建了RV减速器故障实验台,采集了电机电流数据,对基于Fisher-SAE的方法进行了验证,并将其与其他典型机器学习故障诊断方法进行了对比。研究结果表明:该方法能够从RV减速器电机电流信号中提取出故障特征,并选择最有效的故障特征集,解决了振动信号的局限性以及运用电流信号进行故障诊断难以提取有效特征的问题;相比于其他典型机器学习故障诊断方法,该方法的诊断准确率提高了10%~20%,具有更好的诊断效率和准确性。
Aiming at the problems of installation space and signal acquisition of vibration sensors in the actual monitoring of RV reducers,which were prone to limitations and interference,a sparse autoencoder(SAE)and Fisher criterion combination method was proposed for RV gearbox fault feature extraction based on motor current signal analysis(MCSA).Firstly,the collected drive motor current data were converted to the frequency domain.The effect of different hyperparameters on the feature extraction ability of sparse autoencoder was investigated,and the sparse autoencoder with optimized parameters was used to automatically extract fault features from frequency domain signals.Then the Fisher criterion was used to rank the discriminative ability of the extracted features in descending order,and the top n features in the ranking were taken to obtain the optimal fault feature set.Finally,the SoftMax classification layer was combined to achieve fault diagnosis of RV reducers.The RV reducer fault test bench was built,the motor current data was collected,the method based on Fisher-SAE was verified,and it was compared with other typical machine learning fault diagnosis methods.The research results show that the method can extract fault features from the motor current signal of RV reducer,and select the most effective fault feature set,which solves the limitation of vibration signal and the problem that it is difficult to extract effective features by using current signal for fault diagnosis.Comparing with other typical machine learning fault diagnosis methods,the diagnostic accuracy of this method is increased by 10%~20%,and it has better diagnostic efficiency and accuracy.
作者
张兹勤
王贵勇
刘韬
ZHANG Zi-qin;WANG Gui-yong;LIU Tao(Faculty of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,China;Inner Mongolia First Machinery Group Co.,Ltd.,Baotou 014000,China)
出处
《机电工程》
CAS
北大核心
2022年第7期903-910,共8页
Journal of Mechanical & Electrical Engineering
基金
国家自然科学基金资助项目(52065030)
云南省科技厅重大科技专项资助项目(202102AC080002)。
