Performance recognition method based on multi-index and multi-layer DAE-SOM algorithm

doi:10.11949/j.issn.0438-1157.20170973

Abstract

Abstract:

As disturbances and other factors often lead to shifting of work point in industrial process, it is particularly important to timely and accurately identify process changes. Current working condition identification methods mainly focus on whether or what fault occurs, but few consider process conditions from viewpoint of safety, economy, fault and other aspects. Decision criteria at different safety and economic conditions were proposed by combination of historical process data and related operational requirements. With these criteria, data characteristics was extracted by DAE method and extracted feature data was clustered by SOM method. Then, the process state was visually projected to two-dimensional maps. In this method, the DAE method can reduce influence of industrial process disturbance on data and the SOM method can better visually monitor process performance. Experimental study showed that the DAE-SOM multi-layer mapping method can determine security level, type of failure, and current economic efficiency of the system.

Key words: DAE, SOM, performance index, visualization, multi-layer mapping, condition recognition

摘要：

由于工业过程常常受到扰动等因素影响导致工作点发生偏移，所以及时准确地掌握过程运行状况的变化显得尤为重要。目前的工况识别主要针对是否发生故障或者发生何种故障，很少有文献能够从安全、经济、故障等多方面考虑过程工况。针对这一问题，结合过程的历史数据和相关操作要求，获取不同安全状况、经济状况等的评判标准。基于这些判断准则，使用双层降噪自编码（denoising autoencoder，DAE）方法提取数据特征，并用自组织映射神经网络（self-organizing map，SOM）方法对特征提取后的数据聚类，将过程的状况可视化到二维映射图上。在该方法中，DAE方法可以降低工业过程扰动对数据的影响，而SOM方法能够更好地实现过程性能的可视化监控。通过实验可以发现，基于DAE-SOM的双层映射方法可以很好地判断系统的安全级别以及发生的故障类型、当前系统的经济效益状况等。

关键词: 降噪自编码, 自组织映射神经网络, 性能指标, 可视化, 双层映射, 工况识别

CLC Number:

TP277

LI Mengyao, DU Wenli, QIAN Feng. Performance recognition method based on multi-index and multi-layer DAE-SOM algorithm[J]. CIESC Journal, 2018, 69(2): 769-778.

李梦遥, 杜文莉, 钱锋. 基于双层DAE-SOM的多指标工况识别方法[J]. 化工学报, 2018, 69(2): 769-778.

References

[1] 马立玲, 徐发富, 王军政. 一种基于改进核Fisher的故障诊断方法[J]. 化工学报, 2017, 68(3):1041-1048. MA L L, XU F F, WANG J Z. A fault diagnosis method based on improved kernel Fisher[J]. CIESC Journal, 2017,68(3):1041-1048.
[2] 江伟, 王振雷, 王昕. 基于混合分块DMICA-PCA的全流程过程监控方法[J]. 化工学报, 2017, 68(2):759-766. JIANG W, WANG Z L, WANG X. Plant-wide process monitoring based on mixed multiblock DMICA-PCA[J]. CIESC Journal, 2017, 68(2):759-766.
[3] 王坤, 杜文莉, 钱锋. 基于小波核聚类的非高斯过程故障检测方法[J]. 化工学报, 2011, 62(2):427-432. WANG K, DU W L, QIAN F. Non-Gaussian process fault detection based on wavelet kernel clustering[J]. CIESC Journal, 2011, 62(2):427-432.
[4] 齐咏生, 王普, 高学金, 等. 一种基于改进MPCA的间歇过程监控与故障诊断方法[J]. 化工学报, 2009, 60(11):2838-2846. QI Y S, WANG P, GAO X J, et al. Batch process monitoring and fault diagnosis based on improved multi-way principal component analysis[J]. CIESC Journal, 2009, 60(11):2838-2846.
[5] BALDI P, HORNIK K. Neural networks and principal component analysis:learning from examples without local minima[J]. Neural Networks, 1989, 2(1):53-58.
[6] ROWEIS S. EM algorithms for PCA and SPCA[J]. Advances in Neural Information Processing Systems, 1998, 10:626-632.
[7] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323:533-536.
[8] CHANDAR A P S, LAULY S, LAROCHELLE H, et al. An autoencoder approach to learning bilingual word representations[J]. Advances in Neural Information Processing Systems, 2014, 3:1853-1861.
[9] LORE K G, AKINTAYO A, SARKAR S. LLNet:a deep autoencoder approach to natural low-light image enhancement[J]. Pattern Recognition, 2017, 61:650-662.
[10] MEHTA J, MAJUMDAR A. RODEO:robust DE-aliasing autoencoder for real-time medical image reconstruction[J]. Pattern Recognition, 2017, 63:499-510.
[11] DENG J, ZHANG Z, EYBEN F, et al. Autoencoder-based unsupervised domain adaptation for speech emotion recognition[J]. IEEE Signal Processing Letters, 2014, 21(9):1068-1072.
[12] DENG J, ZHANG Z, MARCHI E, et al. Sparse autoencoder-based feature transfer learning for speech emotion recognition[C]//Affective Computing and Intelligent Interaction. IEEE, 2013:511-516.
[13] BETECHUOH B L, MARWALA T, TETTEY T. Autoencoder networks for HIV classification[J]. Current Science, 2006, 91(11):1467-1473.
[14] LI J, STRUZIK Z, ZHANG L, et al. Feature learning from incomplete EEG with denoising autoencoder[J]. Neurocomputing, 2015, 165:23-31.
[15] XIA R, DENG J, SCHULLER B, et al. Modeling gender information for emotion recognition using denoising autoencoder[C]//IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2014:990-994.
[16] MARCHI E, VESPERINI F, EYBEN F, et al. A novel approach for automatic acoustic novelty detection using a denoising autoencoder with bidirectional LSTM neural networks[C]//IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2015:1996-2000.
[17] VINCENT P, LAROCHELLE H, BENGIO Y, et al. Extracting and composing robust features with denoising autoencoders[C]//International Conference on Machine Learning. ACM, 2008:1096-1103.
[18] PRINCIPE J C, WANG L. Non-linear time series modeling with self-organization feature maps[C]//Neural Networks for Signal Processing. IEEE, 1995:11-20.
[19] FORNELLS A, GOLOBARDES E, MARTORELL J M, et al. Measuring the applicability of self-organization maps in a case-based reasoning system[C]//Iberian Conference on Pattern Recognition and Image Analysis. Springer-Verlag, 2007:532-539.
[20] KOHONEN T. The self-organizing map[J]. Neurocomputing, 1990, 21(1/2/3):1-6.
[21] DOWNS J J, VOGEL E F. A plant-wide industrial process control problem[J]. Computers & Chemical Engineering, 1993, 17(3):245-255.
[22] GAO X, HOU J. An Improved SVM Integrated GS-PCA Fault Diagnosis Approach of Tennessee Eastman Process[M]. Amsterdam:Elsevier Science Publishers B. V., 2016.
[23] CHEN H, TI?O P, YAO X. Cognitive fault diagnosis in Tennessee Eastman process using learning in the model space[J]. Computers & Chemical Engineering, 2014, 67(3):33-42.
[24] WANG A, SHA M, LIU L, et al. Fault diagnosis of TE process based on ensemble improved binary-tree SVM[C]//IEEE Fifth International Conference on Bio-Inspired Computing:Theories and Applications. IEEE, 2010:908-912.
[25] GOLSHAN M, PISHVAIE M R, BOOZARJOMEHRY R B. Stochastic and global real time optimization of Tennessee Eastman challenge problem[J]. Engineering Applications of Artificial Intelligence, 2008, 21(2):215-228.
[26] GOLSHAN M, BOOZARJOMEHRY R B, PISHVAIE M R. A new approach to real time optimization of the Tennessee Eastman challenge problem[J]. Chemical Engineering Journal, 2005, 112(1):33-44.
[27] RICKER N L. Optimal steady-state operation of the Tennessee Eastman challenge process[J]. Computers & Chemical Engineering, 1995, 19(9):949-959.
[28] RICKER N L. Decentralized control of the Tennessee Eastman challenge process[J]. Journal of Process Control, 1996, 6(4):205-221.
[29] TIAN Y, DU W L, QIAN F. High dimension feature extraction based visualized SOM fault diagnosis method and its application in p-xylene oxidation process[J]. Chinese Journal of Chemical Engineering, 2015, 23(9):1509-1517.
[30] CHEN X Y, YAN X F. Using improved self-organizing map for fault diagnosis in chemical industry process[J]. Chemical Engineering Research & Design, 2012, 90(12):2262-2277.