摘要
针对旋转机械智能诊断方法计算量大和抗噪能力差的问题,在经典模型LeNet-5的基础上提出基于一维卷积神经网络的故障诊断算法.采用全局平均池化层代替传统卷积神经网络中的全连接层,在降低模型计算量的同时,降低模型参数数量和过拟合的风险;利用随机破坏后的时域信号进行训练以提高其抗噪能力;采用改进后的一维卷积核和池化核直接作用于原始时域信号,将特征提取和故障分类合二为一,通过交替的卷积层和池化层实现原始信号自适应特征提取,结合全局平均池化层完成故障分类.利用轴承数据和齿轮数据进行实验验证并对比经典模型LeNet-5、BP神经网络和SVM.结果表明:采用全局平均池化层可有效降低模型计算量,提高模型在低信噪比条件下的诊断精度,采用随机破坏输入训练策略可显著提升模型的抗噪诊断能力;改进后的模型可以实现噪声环境下准确、快速和稳定的故障诊断.通过t-SNE可视化分析说明了模型在特征学习上的有效性.
A novel one-dimensional(1-D) convolutional neural network(CNN) was proposed based on the classic model LeNet-5, aiming at problems of high computational complexity and low anti-noise ability toward rotating machinery intelligent diagnosis:(1) It adopts global average pooling layer instead of fully connected layers in the conventional CNNs, which reduces the computational complexity, model parameters and risk of overfitting,(2) It is trained with randomly dropout raw signals for anti-noise purpose and(3) It uses modified 1-D convolutional and pooling filters, which works directly on raw time-domain signals, fusing two stages of fault diagnosis into a single learning body, feature learning by the alternating convolutional and pooling layers while classification by the global average pooling layer. The bearing data and gearbox data are used in experimental verification and the classic models of LeNet-5, BP neural network and SVM are used as comparison. The results show that the adoption of global average pooling layers can reduce the model computation and improve the diagnostic accuracy under low signal-to-noise(SNR) conditions, and the train strategy of randomly dropout input can significantly improve the anti-noise ability of the model. As a result, the proposed model can realize accurate, fast and robust fault diagnosis under noisy environment. At last, the t-SNE visualization analysis is used to validate the feature learning ability of the proposed model.
引文
[1] JARDINE A K S,LIN D,BANJEVIC D.A review on machinery diagnostics and prognostics implementing condition-based maintenance [J].Mechanical Systems & Signal Processing,2006,20(7):1483.DOI:10.1016/j.ymssp.2005.09.01 2
[2] GRAVES A,MOHAMED A R,HINTON G.Speech recognition with deep recurrent neural networks[C]// IEEE International Conference on Acoustics,Speech and Signal Processing.Vancouver:IEEE,2013:6645
[3] JIA F,LEI Y,LIN J,et al.Deep neural networks:A promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data[J].Mechanical Systems & Signal Processing,2016,73(7):303.DOI:10.1016/j.ymssp.2015.10.025
[4] TAMILSELVAN P,WANG P.Failure diagnosis using deep belief learning based health state classification[J].Reliability Engineering & System Safety,2013,115(7):124.DOI:10.1016/j.ress.2013.02.022
[5] LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278.DOI:10.1109/5.726791
[6] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]// International Conference on Neural Information Processing Systems.Lake Tahoe:Curran Associates Inc,2012:1097.DOI:10.1145/3065386
[7] SZEGEDY C,LIU W,JIA Y,et al.Going deeper with convolutions[C]// IEEE Conference on Computer Vision and Pattern Recognition.Boston:IEEE,2015:1
[8] HE K,ZHANG X,REN S,et al.Deep residual learning for image recognition[C]// Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:770
[9] WEN L,LI X,GAO L,et al.A new convolutional neural network based data-driven fault diagnosis method[J].IEEE Transactions on Industrial Electronics,2017,65(7):5990.DOI:10.1109/Tie.2017.2774777
[10]JANSSENS O,SLAVKOVIKJ V,VERVISCH B,et al.Convolutional neural network based fault detection for rotating machinery[J].Journal of Sound & Vibration,2016,377(7):331.DOI:10.1016/j.jsv.2016.05.027
[11]INCE T,KIRANYAZ S,EREN L,et al.Real-time motor fault detection by 1-D convolutional neural networks[J].IEEE Transactions on Industrial Electronics,2016,63(11):7067.DOI:10.1109/Tie.2016.2582729
[12]CHEN Z Q,LI C,SANCHEZ R V.Gearbox fault identification and classification with convolutional neural networks[J].Shock & Vibration,2015,2015(2):1.DOI:Artn 390134 10.1155/2015/390134
[13]ZHANG W,PENG G,LI C,et al.A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals[J].Sensors,2017,17(2):425.DOI:ARTN 425 10.3390/s17020425
[14]VINCENT P,LAROCHELLE H,BENGIO Y,et al.Extracting and composing robust features with denoising autoencoders[C]// International Conference on Machine Learning.Helsinki:ACM,2008:1096
[15]GLOROT X,BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[J].Journal of Machine Learning Research,2010,9(7):249
[16]TRA V,KIM J,KHAN S A,et al.Bearing fault diagnosis under variable speed using convolutional neural networks and the stochastic diagonal levenberg-marquardt algorithm.[J].Sensors,2017,17(12):2834.DOI:ARTN 2834 10.3390/s17122834
[17]LOU X,LOPARO K A.Bearing fault diagnosis based on wavelet transform and fuzzy inference [J].Mechanical Systems & Signal Processing,2004,18(5):1077.DOI:10.1016/S0888-3270(03)00077-3