作为一种新兴的机器学习方法，深度学习在故障诊断领域逐渐得到了应用。其中，堆叠降噪自编码（Stacked de-noising auto-encoders，SDAE）算法先对原始数据添加"损伤噪声"，然后通过自编码网络进行数据重构，从而得到更鲁棒性的特征表示，易于进行故障分类。然而针对具体的故障诊断问题，网络隐含层节点数、稀疏参数以及输入数据置零比例将直接影响诊断的结果。因此，提出一种改进的SDAE诊断方法，利用粒子群算法（Particle swarm optimization，PSO）对DAE网络超参数进行自适应的选取来确定SDAE网络结构，据此得到故障状态的特征表示，输入到Soft-max分类器中进行故障分类识别。通过变转速工况下的滚动轴承故障仿真和模拟试验对算法进行验证，试验结果表明，基于PSO-SDAE网络的诊断方法在泛化性、故障识别率方面均优于支持向量机（Support vector machine，SVM）、反向传播神经网络（Back propagation，BP）以及深度置信网络（Deep belief network，DBN）。
As a new machine learning method, deep learning has been used gradually in the field of fault diagnosis. Stacked denoising auto-encoders (SDAE), as one of the deep learning algorithms, could acquire more robust feature representation for effective fault classification by adding "corrupted noise" to the original data, and then reconstructing the input data with the auto-encoder network. However, for a specific diagnosis problem, the number of network hidden nodes, sparse parameters and random zero proportion of input data directly affects the diagnosis results. Based on particle swarm optimization (PSO), an improved SDAE algorithm is proposed for SDAE network hyper-parameters adaptive selection. Then, the determined SDAE networks are used to obtain the feature representations of fault conditions, which could be an input to a soft-max classifier for fault classification. Bearing fault simulation and experiments were conducted under varying running conditions to verify the effectiveness of the proposed method. Experimental results demonstrate that, considering the generalization capability and classification performance, the proposed PSO-SDAE algorithm is superior to support vector machine (SVM), artificial neural network (BP), and deep belief network (DBN).
[1] 伍奎,李润方,刘景浩. 智能化系统的知识表达与推理机制[J].机械工程学报, 2005, 41(5):98-103. WU Kui, LI Runfang, LIU Jinghao. Knowledge express and integrated reasoning mechanism in intelligent system[J]. Chinese Journal of Mechanical Engineering, 2005, 41(5):98-103.
[2] ZHANG X L, WANG B J, CHEN X F. Intelligent fault diagnosis of roller bearings with multivariable ensemble-based incremental support vector machine[J]. Knowledge-Based Systems, 2015, 89:56-85.
[3] 袁胜发,褚福磊,何永勇. 基于网格支持矢量机的涡轮泵多故障诊断[J]. 机械工程学报, 2007, 43(04):152-158. YUAN Shengfa, CHU Fulei, HE Yongyong. Multi-fault diagnosis for turbo-pump based on mesh support vector machines[J]. Chinese Journal of Mechanical Engineering, 2007, 43(4):152-158.
[4] 陈果. 滚动轴承早期故障的特征提取与智能诊断[J]. 航空学报, 2009, 30(2):2-367. CHEN Guo. Early bearing failure in feature extraction and intelligent diagnosis[J]. Journal of Aeronautics, 2009, 30(2):2-367.
[5] 刘建伟,刘媛,罗雄麟. 深度学习研究进展[J]. 计算机应用研究, 2014, 31(7):1921-1930. LIU Jianwei, LIU Yuan, LUO Xionglin. Deep learning research progress[J]. Computer application research, 2014, 31(7):1921-1930.
[6] CHERIYADAT A M. Unsupervised feature learning for aerial scene classification[J]. IEEE Transactions on Geoscience & Remote Sensing, 2014, 52(1):439-451.
[7] RANZATO M, BOUREAU Y L, LECUN Y. Sparse feature learning for deep belief networks[J]. Advances in Neural Information Processing Systems, 2007, 20:1185-1192.
[8] SHIN H C, ORTON M R, COLLINS D J, et al. Stacked autoencoders for unsupervised feature learning and multiple organ detection in a pilot study using 4D patient data[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8):1930-1943.
[9] 雷亚国,贾峰,周昕,等. 基于深度学习理论的机械装备大数据健康监测方法[J]. 机械工程学报, 2015, 51(21):49-56. LEI Yaguo, JIA Feng, ZHOU Xin et al. A deep learning based method for machinery health monitoring with big data[J].Journal of Mechanical Engineering, 2015, 51(21):49-56.
[10] 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. Recognition of bearing state research based on deep learning theory[J]. Journal of Virbration and Shock, 2016, 35(12):166-170.
[11] VINCENT P, LAROCHELLE H, LAJOIE I, et al. Stacked denoising autoencoders:Learning useful representations in a deep network with a local denoising criterion[J]. Journal of Machine Learning Research, 2010, 11(12):3371-3408.
[12] 孙文珺,邵思羽,严如强. 基于稀疏自动编码深度神经网络的感应电动机故障诊断[J]. 机械工程学报, 2016, 52(9):65-71. SUN Wenjun, SHAO Siyu, YAN Ruqiang. Induction motor fault diagnosis based on sparse auto-encoder deep neural network[J]. Journal of Mechanical Engineering, 2016, 52(9):65-71.
[13] EBERHART R C, KENNEDY J. A new optimizer using particle swarm theory[C]//Proceedings of the sixth international symposium on micro machine and human science. 1995, 1:39-43.
[14] BANKS A, VINCENT J, ANYAKOHA C. A review of particle swarm optimization. Part I:background and development[J]. Natural Computing, 2007, 6(4):467-484.
[15] OLSHAUSEN B A, FIELD D J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J]. Nature, 1996, 381(6583):607-609.
[16] KULLBACK S, LEIBLER R A. On information and sufficiency[J]. BPals of Mathematical Statistics, 1951, 22(22):79-86.
[17] BENGIO Y, LAMBLIN P, POPOVICI D, et al. Greedy layer-wise training of deep networks[J]. Advances in Neural Information Processing Systems, 2007, 19:153-176.
[18] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324.
[19] KENNEDY J. Particle swarm optimization[M]. Encyclopedia of Machine Learning. Springer US, 2011:760-766.
[20] LAROCHELLE H, BENGIO Y, LOURADOUR J, et al. Exploring strategies for training deep neural networks[J]. Journal of Machine Learning Research, 2009, 10(1):1-40.
[21] RANAEE V, EBRAHIMZADEH A, GHADERI R. Application of the PSO-SVM model for recognition of control chart patterns[J]. ISA transactions, 2010, 49(4):577-586.
[22] TRELEA I C. The particle swarm optimization algorithm:Convergence analysis and parameter selection[J]. Information Processing Letters, 2003, 85(6):317-325.
