基于深度卷积神经网络的未知复合故障诊断
|
摘要
针对振动信号的非平稳性、非线性以及未知复合故障难以诊断的问题,提出了一种基于深度卷积网络的未知复合故障诊断模型。首先将采集到的时域振动信号通过小波变换生成频谱图像;然后将频谱图输入卷积神经网络(convolutional neural network,CNH),利用卷积网络自适应的特征提取能力对复合故障进行特征学习;最后将深度卷积网络输出的特征通过分类器对故障进行诊断分类。在实验室模拟采集的不同数据集上进行实验,结果表明:基于深度卷积网络的未知复合故障诊断模型与基于改进CDCGAN的复合故障诊断方法相比,对未知复合故障的诊断率提高了44%,达到85.77%;使用不同类型的单一未知复合故障和多种未知复合故障进行实验,验证了所提模型的泛化能力和鲁棒性。
A diagnostic model for unknown compound fault based on deep convolutional networks is proposed.It is used to solve the problems of non-stationary and nonlinear vibration signals,and unknown compound fault.Firstly,the model collects vibration signal in the time domain,and generates the spectrum image through wavelet transform.Then the spectrograms are input into the convolutional neural network(CNN),to learn the features of the compound faults by the adaptive feature extraction capability of the network.Finally,the output features of the deep convolutional network are classified to different fault types by the classifier.Experiments on different simulation data sets collected in the laboratory show that the diagnostic rate of unknown compound faults is 85.77%.The results have risen by 44%,compared with improved CDCGAN.Experiments were carried out by different types of single and multiple unknown compound faults,to verify the generalization and robustness of the proposed model.
A diagnostic model for unknown compound fault based on deep convolutional networks is proposed.It is used to solve the problems of non-stationary and nonlinear vibration signals,and unknown compound fault.Firstly,the model collects vibration signal in the time domain,and generates the spectrum image through wavelet transform.Then the spectrograms are input into the convolutional neural network(CNN),to learn the features of the compound faults by the adaptive feature extraction capability of the network.Finally,the output features of the deep convolutional network are classified to different fault types by the classifier.Experiments on different simulation data sets collected in the laboratory show that the diagnostic rate of unknown compound faults is 85.77%.The results have risen by 44%,compared with improved CDCGAN.Experiments were carried out by different types of single and multiple unknown compound faults,to verify the generalization and robustness of the proposed model.
引文
[1]张可,周东华,柴毅.复合故障诊断技术综述[J].控制理论与应用,2015,32(9):1143-1157.ZHANG Ke,ZHOU Donghua,CHAI Yi.Review of multiple fault diagnosis methods[J].Control Theory and Application,2015,32(9):1143-1157.(in Chinese)
[2]JIANG J,JIA F.A robust fault diagnosis scheme based on signal modal estimation[J].International Journal of Control,1995,62(2):461-475.
[3]崔玲丽,高立新,张建宇,等.基于EMD的复合故障诊断方法[J].工程科学学报,2008,30(9):1055-1060.CUI Lingli,GAO Lixin,ZHANG Jianyu,et al.Composite fault diagnosis method based on empirical mode decomposition[J].Journal of Engineering Science,2008,30(9):1055-1060.(in Chinese)
[4]马新娜,杨绍普.滚动轴承复合故障诊断的自适应方法研究[J].振动与冲击,2016,35(10):145-150.MA Xinna,YANG Shaopu.Adaptive compound fault diagnosis of rolling bearings[J].Journal of Vibration and Shock,2016,35(10):145-150.(in Chinese)
[5]ZHANG Min,CAI Zhenyu,CHENG Wenming.Multiple-fault diagnosis method based on multiscale feature extraction and MSVM_PPA[J].Shock and Vibration,2018,115(10):1-12.
[6]孙志诚,沈长青,王富东,等.基于时频分析与人工神经网络的轴承诊断研究[J].机电一体化,2017,23(4):21-27.SUN Zhicheng,SHEN Changqing,WANG Fudong,et al.Bearing fault diagnosis based on time-frequency analysis and artificial neural network[J].Mechatronics,2017,23(4):21-27.(in Chinese)
[7]赵海洋,王金东,刘树林,等.基于神经网络和支持向量机的复合故障诊断技术[J].流体机械,2008,36(1):39-42.ZHAO Haiyang,WANG Jindong,LIU Shulin,et al.Compound fault diagnosis technique based on artificial neural network and support vector machine[J].Fluid Machinery,2008,36(1):39-42.(in Chinese)
[8]SANZ J,PERERA R,HUERTA C.Fault diagnosis of rotating machinery based on auto-associative neural networks and wavelet transforms[J].Journal of Sound&Vibration,2007,302(4/5):981-999.
[9]KUO J T,HSIEH M H,LUNG W S,et al.Using artificial neural network for reservoir eutrophication prediction[J].Ecological Modelling,2007,200(1/2):171-177.
[10]YU Kai,LIN Yuanqing,LAFFERTY J.Learning image representations from the pixel level via hierarchical sparse coding[C]∥Computer Vision and Pattern Recognition.Kauai,USA:IEEE,2011:1713-1720.
[11]DAHL G E,ACERO A.Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition[J].IEEE Transactions on Audio Speech&Language Processing,2012,20(1):30-42.
[12]GLOROT X,BORDES A,BENGIO Y.Domain adaptation for large-scale sentiment classification:a deep learning approach[C]∥International Conference on International Conference on Machine Learning.Washington,USA:Omnipress,2011:513-520.
[13]郭亮,高宏力,张一文,等.基于深度学习理论的轴承状态识别研究[J].振动与冲击,2016,35(12):166-170.GUO Liang,GAO Hongli,ZHANG Yiwen,et al.Research on bearing condition monitoring based on deep learning[J].Journal of Vibration and Shock,2016,35(12):166-170.(in Chinese)
[14]陈伟.深度学习在滚动轴承故障诊断中的应用研究[D].成都:西南交通大学,2017.CHEN Wei.Application of deep learning in rolling bearing fault diagnosis[D].Chengdu:Southwest Jiaotong University,2017.(in Chinese)
[15]PARK S R,LEE J.A fully convolutional neural network for speech enhancement[J].Arxiv Preprint Arxiv,2016,16(9):7132-7138.
[16]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]∥International Conference on Neural Information Processing Systems.Doha,Qatar:Curran Associates Inc.,2012:1097-1105.
[17]CHEN Yushi,LIN Zhouhua,ZHAO Xing,et al.Deep learning-based classification of hyperspectral data[J].IEEE Journal of Selected Topics in Applied Earth Observations&Remote Sensing,2014,7(6):2094-2107.
[18]HINTON G E,SRIVASTAVA N,KRIZHEVSKY A,et al.Improving neural networks by preventing co-adaptation of feature detectors[J].Arxiv Preprint Arxiv,2012,12(7):580-588.
[19]ABADI M,AGARWAL A,BARHAM P,et al.TensorFlow:large-scale machine learning on heterogeneous distributed systems[J].Arxiv Preprint Arxiv,2016,16(3):4467-4482.
[20]李永健,刘吉华,张卫华,等.改进样本熵及其在列车轴承损伤检测中的应用[J].仪器仪表学报,2018,39(9):179-186.LI Yongjian,LIU Jihua,ZHANG Weihua,et al.Improved multiscale sample entropy and its application in train axle bearing fault detection[J].Chinese Journal of Scientific Instrument,2018,39(9):179-186.(in Chinese)
[2]JIANG J,JIA F.A robust fault diagnosis scheme based on signal modal estimation[J].International Journal of Control,1995,62(2):461-475.
[3]崔玲丽,高立新,张建宇,等.基于EMD的复合故障诊断方法[J].工程科学学报,2008,30(9):1055-1060.CUI Lingli,GAO Lixin,ZHANG Jianyu,et al.Composite fault diagnosis method based on empirical mode decomposition[J].Journal of Engineering Science,2008,30(9):1055-1060.(in Chinese)
[4]马新娜,杨绍普.滚动轴承复合故障诊断的自适应方法研究[J].振动与冲击,2016,35(10):145-150.MA Xinna,YANG Shaopu.Adaptive compound fault diagnosis of rolling bearings[J].Journal of Vibration and Shock,2016,35(10):145-150.(in Chinese)
[5]ZHANG Min,CAI Zhenyu,CHENG Wenming.Multiple-fault diagnosis method based on multiscale feature extraction and MSVM_PPA[J].Shock and Vibration,2018,115(10):1-12.
[6]孙志诚,沈长青,王富东,等.基于时频分析与人工神经网络的轴承诊断研究[J].机电一体化,2017,23(4):21-27.SUN Zhicheng,SHEN Changqing,WANG Fudong,et al.Bearing fault diagnosis based on time-frequency analysis and artificial neural network[J].Mechatronics,2017,23(4):21-27.(in Chinese)
[7]赵海洋,王金东,刘树林,等.基于神经网络和支持向量机的复合故障诊断技术[J].流体机械,2008,36(1):39-42.ZHAO Haiyang,WANG Jindong,LIU Shulin,et al.Compound fault diagnosis technique based on artificial neural network and support vector machine[J].Fluid Machinery,2008,36(1):39-42.(in Chinese)
[8]SANZ J,PERERA R,HUERTA C.Fault diagnosis of rotating machinery based on auto-associative neural networks and wavelet transforms[J].Journal of Sound&Vibration,2007,302(4/5):981-999.
[9]KUO J T,HSIEH M H,LUNG W S,et al.Using artificial neural network for reservoir eutrophication prediction[J].Ecological Modelling,2007,200(1/2):171-177.
[10]YU Kai,LIN Yuanqing,LAFFERTY J.Learning image representations from the pixel level via hierarchical sparse coding[C]∥Computer Vision and Pattern Recognition.Kauai,USA:IEEE,2011:1713-1720.
[11]DAHL G E,ACERO A.Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition[J].IEEE Transactions on Audio Speech&Language Processing,2012,20(1):30-42.
[12]GLOROT X,BORDES A,BENGIO Y.Domain adaptation for large-scale sentiment classification:a deep learning approach[C]∥International Conference on International Conference on Machine Learning.Washington,USA:Omnipress,2011:513-520.
[13]郭亮,高宏力,张一文,等.基于深度学习理论的轴承状态识别研究[J].振动与冲击,2016,35(12):166-170.GUO Liang,GAO Hongli,ZHANG Yiwen,et al.Research on bearing condition monitoring based on deep learning[J].Journal of Vibration and Shock,2016,35(12):166-170.(in Chinese)
[14]陈伟.深度学习在滚动轴承故障诊断中的应用研究[D].成都:西南交通大学,2017.CHEN Wei.Application of deep learning in rolling bearing fault diagnosis[D].Chengdu:Southwest Jiaotong University,2017.(in Chinese)
[15]PARK S R,LEE J.A fully convolutional neural network for speech enhancement[J].Arxiv Preprint Arxiv,2016,16(9):7132-7138.
[16]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]∥International Conference on Neural Information Processing Systems.Doha,Qatar:Curran Associates Inc.,2012:1097-1105.
[17]CHEN Yushi,LIN Zhouhua,ZHAO Xing,et al.Deep learning-based classification of hyperspectral data[J].IEEE Journal of Selected Topics in Applied Earth Observations&Remote Sensing,2014,7(6):2094-2107.
[18]HINTON G E,SRIVASTAVA N,KRIZHEVSKY A,et al.Improving neural networks by preventing co-adaptation of feature detectors[J].Arxiv Preprint Arxiv,2012,12(7):580-588.
[19]ABADI M,AGARWAL A,BARHAM P,et al.TensorFlow:large-scale machine learning on heterogeneous distributed systems[J].Arxiv Preprint Arxiv,2016,16(3):4467-4482.
[20]李永健,刘吉华,张卫华,等.改进样本熵及其在列车轴承损伤检测中的应用[J].仪器仪表学报,2018,39(9):179-186.LI Yongjian,LIU Jihua,ZHANG Weihua,et al.Improved multiscale sample entropy and its application in train axle bearing fault detection[J].Chinese Journal of Scientific Instrument,2018,39(9):179-186.(in Chinese)