长短时记忆网络在电机故障诊断中的应用研究
|
摘要
针对电机故障诊断采用传统神经网络存在的梯度消失等问题,提出了一种长短时记忆(LSTM)神经网络与Softmax多分类器结合的诊断方法。首先,利用LSTM神经网络在提取时间序列特征方面的良好特性,通过LSTM神经网络与Softmax多分类器构建故障诊断模型。然后,通过Tensorflow学习框架有效提取故障数据特征,并将具有强泛化能力和鲁棒性的Softmax多分类器对其分类,从而诊断出电机内圈、外圈和滚珠三种常见故障,提高诊断结果的准确率,改善传统方法存在的不足。最后,仿真验证所提方法的有效性与可行性。与传统神经网络和堆栈稀疏自编码器分类结果相比,采用LSTM神经网络诊断方法其准确率达到98. 3%,在电机故障诊断中具有更好的诊断效果,且对提高故障诊断的准确率有一定的作用。
Aimed at the problem of vanishing gradientin motor fault diagnosis using traditional neural network,a diagnostic method combining long short-term memory( LSTM) neural network and Softmax multi-classifier is proposed. Firstly,the fault diagnosis model is constructed by using LSTM neural network and Softmax multi-classifier,in which the good features of LSTM neural network in extracting the characteristics of time series are adopted. Then,through the Tensorflow learning framework,the fault data features are effectively extracted,and the Softmax multi-classifier with strong generalization ability and robustness is classified to diagnose the common faults of inner ring,outer ring and balls of motor,to improve the accuracy of diagnostic results and overcome the shortcomings of traditional methods. Finally,the effectiveness and feasibility of the proposed method are verified by experimental simulation. Compared with results of traditional neural network and stacked sparse autoencoder classification,the accuracy of the LSTM neural network diagnosis method is up to 98. 3%,which has better diagnostic effect in motor fault diagnosis and has a certain effect on improving the accuracy of fault diagnosis.
Aimed at the problem of vanishing gradientin motor fault diagnosis using traditional neural network,a diagnostic method combining long short-term memory( LSTM) neural network and Softmax multi-classifier is proposed. Firstly,the fault diagnosis model is constructed by using LSTM neural network and Softmax multi-classifier,in which the good features of LSTM neural network in extracting the characteristics of time series are adopted. Then,through the Tensorflow learning framework,the fault data features are effectively extracted,and the Softmax multi-classifier with strong generalization ability and robustness is classified to diagnose the common faults of inner ring,outer ring and balls of motor,to improve the accuracy of diagnostic results and overcome the shortcomings of traditional methods. Finally,the effectiveness and feasibility of the proposed method are verified by experimental simulation. Compared with results of traditional neural network and stacked sparse autoencoder classification,the accuracy of the LSTM neural network diagnosis method is up to 98. 3%,which has better diagnostic effect in motor fault diagnosis and has a certain effect on improving the accuracy of fault diagnosis.
引文
[1]沈标正.电机故障诊断技术[M].北京:机械工业出版社,2001.
[2]汤蕴谬,史乃.电机学[M].北京:机械工业出版社,2006.
[3] GUO J,QI L,LI Y. Fault detection of batch process using dynamic multi-way orthogonal locality preserving projections[J]. Journal of Computational Information Systems,2015,11(2):577-586.
[4] CHIANG L H,KOTANCHEK M E,KORDON A K. Fault diagnosis based on Fisher discriminant analysis and support vector machines[J].Computers&Chemical Engineering,2004,28(8):1389-1401.
[5] ESLAMLOUEYAN R. Designing a hierarchical neural network based on fuzzy clustering for fault diagnosis of the Tennessee-Eastman process[J]. Applied Soft Computing Journal,2011,11(1):1407-1415.
[6] BENGIO Y. Learning deep architectures for AI[J]. Foundations and Trends in Machine Learning,2009,2(1):1-127.
[7]王春梅.基于深度置信网络的风电机组主轴承故障诊断方法研究[J].自动化仪表,2018,39(5):33-37.
[8]陈如清.两种基于神经网络的故障诊断方法[J].中国电机工程学报,2005(16):112-115.
[9] XIAO S,YAN J,CHU S M,et al. Modeling The Intensity Function of Point Process Via Recurrent Neural Networks[C]//AAAI,2017:1597-1603.
[10]HOCHREITER S,SCHMIDHUBER J. Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[11]SUTSKEVER I,VINYALS O,LE Q V,et al. Sequence to sequence learning with neural networks[C]//Advances in Neural Information Processing Systems,2014:3104-3112.
[12]GERS F A,SCHMIDHUBER J,CUMMINS F. Learning to forget:continual prediction with LSTM[J]. Neural Computation,2000,12(10):2451-71.
[13]Bearing data center. Case Western Reserve University[EB/OL].[2016-12-31]. Available:http://csegroups. case. edu/bearingdatacenter/pages/download-data-file.
[2]汤蕴谬,史乃.电机学[M].北京:机械工业出版社,2006.
[3] GUO J,QI L,LI Y. Fault detection of batch process using dynamic multi-way orthogonal locality preserving projections[J]. Journal of Computational Information Systems,2015,11(2):577-586.
[4] CHIANG L H,KOTANCHEK M E,KORDON A K. Fault diagnosis based on Fisher discriminant analysis and support vector machines[J].Computers&Chemical Engineering,2004,28(8):1389-1401.
[5] ESLAMLOUEYAN R. Designing a hierarchical neural network based on fuzzy clustering for fault diagnosis of the Tennessee-Eastman process[J]. Applied Soft Computing Journal,2011,11(1):1407-1415.
[6] BENGIO Y. Learning deep architectures for AI[J]. Foundations and Trends in Machine Learning,2009,2(1):1-127.
[7]王春梅.基于深度置信网络的风电机组主轴承故障诊断方法研究[J].自动化仪表,2018,39(5):33-37.
[8]陈如清.两种基于神经网络的故障诊断方法[J].中国电机工程学报,2005(16):112-115.
[9] XIAO S,YAN J,CHU S M,et al. Modeling The Intensity Function of Point Process Via Recurrent Neural Networks[C]//AAAI,2017:1597-1603.
[10]HOCHREITER S,SCHMIDHUBER J. Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[11]SUTSKEVER I,VINYALS O,LE Q V,et al. Sequence to sequence learning with neural networks[C]//Advances in Neural Information Processing Systems,2014:3104-3112.
[12]GERS F A,SCHMIDHUBER J,CUMMINS F. Learning to forget:continual prediction with LSTM[J]. Neural Computation,2000,12(10):2451-71.
[13]Bearing data center. Case Western Reserve University[EB/OL].[2016-12-31]. Available:http://csegroups. case. edu/bearingdatacenter/pages/download-data-file.