针对化工过程高维数据的故障特征难以提取的难题，提出变分自动编码器(VAE)结合深度置信网络(DBN)的混合故障诊断方法. 在VAE的编码部分对隐变量空间Z添加约束，通过重参数化方法进行反向传播训练，可无监督地学习不同故障对应的隐变量特征，其作为DBN分类模型的输入特征训练网络，输入测试集进行故障诊断. 田纳西伊斯曼流程(TE)应用结果表明，VAE能提取原始数据更加抽象有效的特征，VAE?DBN分类准确.
To extract the fault feature from a large quantity of high-dimensional data, a variational auto-encoder (VAE) and deep belief network (DBN) combined fault diagnosis method was proposed for chemical process. In the encoding process of VAE, constraints were added to the latent variable space Z, and the backward propagation training was carried out by the re-parameterization method. The latent variables corresponding to different faults could be learned without supervision. Subsequently, the latent variable features learned by VAE were used as input features of the DBN classification model to diagnose the faults. The results showed that VAE could extract more abstract and effective features from the original data, and VAE?DBN had excellent performance in classification accuracy.
[1]Tian W D, Hu M G, Li C K.Fault Prediction Based on Dynamic Model and Grey Time Series Model in Chemical Processes[J].Chinese Journal of Chemical Engineering, 2014, 22(6):643-650
[2]Tian W D, Sun S L, Guo Q J.Fault Detection and Diagnosis for Distillation Column Using Two-tier Model[J].Canadian Journal of Chemical Engineering, 2013, 91(10):1671-1685
[3]毛海涛 田文德 梁慧婷.基于双层机器学习的动态精馏过程故障检测与分离[J].过程工程学报, 2017, 17(2):351-356
[4]J B Tenenbaum, V De Silva, J C Langford.A global geometric framework for nonlinear dimensionality reduction[J].Science, 2000, 290(5500):2319-2323
[5]Roweis S T, Saul.L KNonlinear Dimensionality Reduction by Locally Linear Embedding[J].Science, 2000, 290(5500):2323-2326
[6]M Belkin, P Niyogi.Laplacian Eigenmaps for Dimensionality Reduction and Data Representation[J].Neural computation, 2003, 15(6):1373-1396
[7] V D Silva, J B Tenenbaum.Global Versus Local Methods in Nonlinear Dimensionality Reduction [J].Advances in Neural Information Processing Systems, 2003, 15:1959-1966
[8]Hinton G E.Visualizing High-Dimensional Data Using t-SNE[J].Vigiliae Christianae, 2008, 9(2):2579-2605
[9]Hinton G, Roweis S.Stochastic Neighbor Embedding[J].Advances in Neural Information Processing Systems, 2003, 41(4):833-840
[10] Laurens V D M.Accelerating t-SNE Using Tree-based Algorithms [M]. JMLR.org, 2014.
[11] Sun W, Shao S, Zhao R, et al.A Sparse Auto-encoder-based Deep Neural Network Approach for Induction Motor Faults Classification [J].Measurement, 2016, 89:171-178
[12]Lu C, Wang Z Y, Qin W L, et al.Fault Diagnosis of Rotary Machinery Components Using a Stacked Denoising Autoencoder Based Health State Identification[J].Signal Processing, 2017, 130(C):377-388
[13] Shao H, Jiang H, Zhao H, et al.A Novel Deep Autoencoder Feature Learning Method for Rotating Machinery Fault Diagnosis[J].Mechanical Systems & Signal Processing, 2017, 95:187-204
[14] A M. Cheriyadat,.Unsupervised Feature Learning for Aerial Scene Classification 439–451.[J].IEEE Transactions on Geoscience & Remote Sensing, 2014, 52(1):439-451
[15] Kingma D P, Welling M.Auto-Encoding Variational Bayes[C]// Conference proceedings: papers accepted to the International Conference on Learning Representations (ICLR) 2014. arXiv.org, 2014.
[16]Hinton G E, Osindero S, Teh Y W.A Fast Learning Algorithm for Deep Belief Nets[J].Neural Computation, 2006, 18(7):1527-1554
[17]Hinton G E, Salakhutdinov R R.Reducing the Dimensionality of Data with Neural Networks[J].Science, 2006, 313(5786):504
[18] Zhang Z, Zhao J S.A Deep Belief Network Based Fault Diagnosis Model for Complex Chemical Processes[J].Computers & Chemical Engineering, 2017
