Research on Fault Diagnosis of Data Dimension Reduction Based on Improved Structure Preserving Algorithm
-
摘要: 传统基于核主成分分析(Kernel principal component analysis, KPCA)的数据降维方法在提取有效特征信息时只考虑全局结构保持而未考虑样本间的局部近邻结构保持问题, 本文提出一种改进全局结构保持算法的特征提取与降维方法.改进的特征提取与降维方法将流形学习中核局部保持投影(Kernel locality preserving projection, KLPP)的思想融入核主成分分析的目标函数中, 使样本投影后的特征空间不仅保持原始样本空间的整体结构, 还保持样本空间相似的局部近邻结构, 包含更丰富的特征信息.上述方法通过同时进行的正交化处理可避免局部子空间结构发生失真, 并能够直观显示出低维结果, 将低维数据输入最近邻分类器, 以识别率和聚类分析结果作为衡量指标, 同时将所提方法应用于故障诊断中.使用AVL Boost软件模拟的柴油机故障数据和田纳西(Tennessee Eastman, TE)化工数据仿真, 验证了所提方法的有效性.Abstract: The traditional data reduction method based on kernel principal component analysis (KPCA) only considers the global structure preservation when extracting effective feature information, but does not take the problem of local neighbor structure retention between samples into consider. An improved feature extraction and dimension reduction of global structure preservation algorithm is proposed which integrates the idea of kernel locality preserving projection (KLPP) of manifold learning into the objective function of kernel principal component analysis, so that the feature space after the sample projection not only remains the whole original sample space. But also maintains a local neighbor structure with similar sample space which contains more feature information. Distortion of the local subspace structure can be avoided by simultaneous orthogonalization, and the low-dimensional results can be visually displayed. The low-dimensional data is inputed into the nearest neighbor classifier, using the recognition rate and cluster analysis results as a measurement. At the same time, the proposed method is applied to fault diagnosis. The diesel engine fault data simulation simulated by AVL Boost software and Tennessee Eastman (TE) chemical data simulation verify the effectiveness of the proposed algorithm.
-
Key words:
- Feature extraction /
- data dimension reduction /
- kernel principal component analysis /
- locality preserving projection /
- fault diagnosis
1) 本文责任编委 曾志刚 -
表 1 正常工况与故障工况模拟
Table 1 The simulation of normal and fault conditions
No. 工况类型 样本个数 数据维数 1 正常工况 960 15 2 故障1_空冷器冷却不足 960 15 3 故障2_排气口堵塞 960 15 4 故障3_涡轮增压效率降低 960 15 
下载: 导出CSV
表 2 数据与台架实验数据多工况对比
Table 2 The data contrast between AVL Boost and bench test under multiple working conditions
负荷 排气温度(℃) 相对误差(%) 功率(kW) 相对误差(%) 模型数据 台架实验数据 模型数据 台架实验数据 90%负荷 329.89 328.50 0.42 3 281.40 3 277.00 0.13 75%负荷 304.39 307.30 0.95 2 839.20 2 844.00 0.17 75%推进 319.23 320.90 0.37 2 866.85 2 864.00 0.10
下载: 导出CSV
表 3 故障1识别准确率($ \% $)
Table 3 The accuracy of fault1 diagnosis ($ \% $)
方法 Fault1 KPCA KLPP KFDA LGPCA TGLSA GLSP ELM 55.32 61.38 60.58 54.21 58.69 62.97 SVM 58.69 70.61 71.68 65.34 68.49 69.27 RVM 72.77 69.59 74.21 68.98 63.40 76.35 KNN 72.26 66.86 70.38 75.49 77.36 78.53
下载: 导出CSV
表 4 故障2识别准确率($ \% $)
Table 4 The accuracy of fault2 diagnosis ($ \% $)
方法 Fault2 KPCA KLPP KFDA LGPCA TGLSA GLSP ELM 80.95 76.85 79.65 77.49 70.28 82.62 SVM 78.36 77.32 77.05 74.39 72.15 80.09 RVM 79.74 74.16 78.66 85.68 81.29 83.62 KNN 82.35 82.63 75.39 78.91 86.54 88.84
下载: 导出CSV
表 5 故障3识别准确率($ \% $)
Table 5 The accuracy of fault3 diagnosis ($ \% $)
方法 Fault3 KPCA KLPP KFDA LGPCA TGLSA GLSP ELM 70.65 72.39 77.16 74.29 70.53 79.26 SVM 66.34 68.29 68.49 65.39 60.87 66.58 RVM 59.38 62.58 55.21 59.86 60.13 66.34 KNN 58.62 62.38 65.98 63.24 61.09 65.08
下载: 导出CSV
表 6 特征提取所需时间(s)
Table 6 Feature extraction time (s)
维度 特征提取方法 KPCA KLPP KFDA LGPCA TGLSA GLSP 3 0.651 1.155 1.039 2.598 2.134 1.596 5 0.795 1.159 1.118 2.019 1.495 1.632 8 0.815 1.209 0.975 1.069 1.396 1.885 10 0.867 1.344 1.185 1.563 2.098 1.962
下载: 导出CSV
-
[1] 赵孝礼, 赵荣珍.全局与局部判别信息融合的转子故障数据集降维方法研究.自动化学报, 2017, 43(4): 560-567 doi: 10.16383/j.aas.2017.c160317Zhao Xiao-Li, Zhao Rong-Zhen. A method of dimension reduction of rotor faults data set based on fusion of global and local discriminant information. Acta Automatica Sinica, 2017, 43(4): 560-567 doi: 10.16383/j.aas.2017.c160317 [2] Garcia-Alvarez D, Fuente M J, Sainz G I. Fault detection and isolation in transient states using principal component analysis. Journal of Process Control, 2012, 22(3): 551-563 doi: 10.1016/j.jprocont.2012.01.007 [3] Han M, Jiang L W. Endpoint prediction model of basic oxygen furnace steelmaking based on PSO-ICA and RBF neural network. In: Proceedings of the 2010 IEEE International Conference on Intelligent Control and Information Processing. Dalian, China: IEEE, 2010. 388-393 [4] Han M, Zhong K, Qiu T, Han B. Interval type-2 fuzzy neural networks for chaotic time series prediction: a concise overview. IEEE Transactions on Cybernetics, DOI: 10.1109/TCYB.2018.2834356 [5] Zhang Y W, Zhou H, Qin S J. Decentralized fault diagnosis of large-scale processes using multiblock kernel principal component analysis. Acta Automatica Sinica, 2010, 36(4): 593-597 [6] 张晓涛, 唐力伟, 王平, 邓士杰.基于多尺度正交PCA-LPP流形学习算法的故障特征增强方法.振动与冲击, 2015, 34(13): 66-70 https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ201513013.htmZhang Xiao-Tao, Tang Li-Wei, Wang Ping, Deng Shi-Jie. Fault feature enhancement method based on multiscale orthogonal PCA-LPP manifold learning algorithm. Journal of Vibration and Shock, 2015, 34(13): 66-70 https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ201513013.htm [7] 李娟, 王宇平.考虑局部均值和类全局信息的快速近邻原型选择算法.自动化学报, 2014, 40(6): 1116-1125 doi: 10.3724/SP.J.1004.2014.01116Li Juan, Wang Yu-Ping. A fast neighbor prototype selection algorithm based on local mean and class global information. Acta Automatica Sinica, 2014, 40(6): 1116-1125 doi: 10.3724/SP.J.1004.2014.01116 [8] 王健, 冯健, 韩志艳.基于流形学习的局部保持PCA算法在故障检测中的应用.控制与决策, 2013, 28(5): 683-687 https://www.cnki.com.cn/Article/CJFDTOTAL-KZYC201305008.htmWang Jian, Feng Jian, Han Zhi-Yan. Locally preserving PCA method based on manifold learning and its application in fault detection. Control and Decision, 2013, 28(5): 683-687 https://www.cnki.com.cn/Article/CJFDTOTAL-KZYC201305008.htm [9] Tenenbaum J B, De Silva V, Langford J C. A global geometric framework for nonlinear dimensionality reduction. Science, 2000, 290(5500): 2319-2323 doi: 10.1126/science.290.5500.2319 [10] Sprekeler H. On the relation of slow feature analysis and Laplacian eigenmaps. Neural Computation, 2011, 23(12): 3287-3302 doi: 10.1162/NECO_a_00214 [11] Roweis S T, Saul L K. Nonlinear dimensionality reduction by locally linear embedding. Science, 2000, 290(5500): 2323-2326 doi: 10.1126/science.290.5500.2323 [12] He X F, Yan S C, Hu Y X, Niyogi P, Zhang H J. Face recognition using laplacianfaces. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005, 27(3): 328-340 doi: 10.1109/TPAMI.2005.55 [13] Deng X G, Tian X M. Sparse kernel locality preserving projection and its application in nonlinear process fault detection. Chinese Journal of Chemical Engineering, 2013, 21(2): 163-170 doi: 10.1016/S1004-9541(13)60454-1 [14] 袁暋, 程雷, 朱然刚, 雷迎科.一种新的基于MMC和LSE的监督流形学习算法.自动化学报, 2013, 39(12): 2077-2089 doi: 10.3724/SP.J.1004.2013.02077Yuan Min, Cheng Lei, Zhu Ran-Gang, Lei Ying-Ke. A new supervised manifold learning algorithm based on MMC and LSE. Acta Automatica Sinica, 2013, 39(12): 2077-2089 doi: 10.3724/SP.J.1004.2013.02077 [15] 赵旭, 阎威武, 邵惠鹤.基于核Fisher判别分析方法的非线性统计过程监控与故障诊断.化工学报, 2007, 58(4): 951-956 doi: 10.3321/j.issn:0438-1157.2007.04.026Zhao Xu, Yan Wei-Wu, Shao Hui-He. Nonlinear statistical process monitoring and fault diagnosis based on kernel Fisher discriminant analysis. Journal of Chemical Industry and Engineering (China), 2007, 58(4): 951-956 doi: 10.3321/j.issn:0438-1157.2007.04.026 [16] Kahveci N E, Impram S T, Genc A U. Boost pressure control for a large diesel engine with turbocharger. In: Proceedings of the 2014 American Control Conference. Portland, OR, USA: IEEE, 2014. 2108-2113 [17] 韩敏, 张占奎.基于改进核主成分分析的故障检测与诊断方法.化工学报, 2015, 66(6): 2139-2149 https://www.cnki.com.cn/Article/CJFDTOTAL-HGSZ201506021.htmHan Min, Zhang Zhan-Kui. Fault detection and diagnosis method based on modified kernel principal component analysis. CIESC Journal, 2015, 66(6): 2139-2149 https://www.cnki.com.cn/Article/CJFDTOTAL-HGSZ201506021.htm [18] Zhang M G, Ge Z Q, Song Z H, Fu R W. Global-local structure analysis model and its application for fault detection and identification. Industrial & Engineering Chemistry Research, 2011, 50(11): 6837-6848 [19] Deng X G, Tian X M, Chen S. Modified kernel principal component analysis based on local structure analysis and its application to nonlinear process fault diagnosis. Chemometrics and Intelligent Laboratory Systems, 2013, 127(16): 195-209 [20] Luo L J, Bao S Y, Gao Z L, Yuan J Q. Batch process monitoring with tensor global-local structure analysis. Industrial & Engineering Chemistry Research, 2013, 52(50): 18031-18042 [21] Liu Q S, Tang X O, Lu H Q, Ma S D. Face recognition using kernel scatter-difference-based discriminant analysis. IEEE Transactions on Neural Networks, 2006, 17(4): 1081-1085 doi: 10.1109/TNN.2006.875970 [22] Dufrenois F. A one-class kernel fisher criterion for outlier detection. IEEE Transactions on Neural Networks and Learning Systems, 2015, 26(5): 982-994 doi: 10.1109/TNNLS.2014.2329534 [23] Mika S, Ratsch G, Weston J, Scholkopf B, Mullers K R. Fisher discriminant analysis with kernels. In: Proceedings of the 1999 IEEE Signal Processing Society Workshop. Madison, WI, USA: IEEE, 1999. 41-48 [24] Bo L F, Wang L, Jiao L C. Feature scaling for kernel fisher discriminant analysis using leave-one-out cross validation. Neural Computation, 2006, 18(4): 961-978 doi: 10.1162/neco.2006.18.4.961 [25] 陈法法, 汤宝平, 苏祖强.基于等距映射与加权KNN的旋转机械故障诊断.仪器仪表学报, 2013, 34(1): 215-220 https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB201301030.htmChen Fa-Fa, Tang Bao-Ping, Su Zu-Qiang. Rotating machinery fault diagnosis based on isometric mapping and weighted KNN. Chinese Journal of Scientific Instrument, 2013, 34(1): 215-220 https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB201301030.htm [26] 王泽杰, 胡浩民.流形学习算法中的参数选择问题研究.计算机应用与软件, 2010, 27(6): 84-85, 102 https://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ201006027.htmWang Ze-Jie, Hu Hao-Min. On parameter selection in manifold learning algorithm. Computer Applications and Software, 2010, 27(6): 84-85, 102 https://www.cnki.com.cn/Article/CJFDTOTAL-JYRJ201006027.htm [27] Yang Y S, Ming A B, Zhang Y Y, Zhu Y S. Discriminative non-negative matrix factorization (DNMF) and its application to the fault diagnosis of diesel engine. Mechanical Systems and Signal Processing, 2017, 95: 158-171 doi: 10.1016/j.ymssp.2017.03.026 [28] 苏祖强, 汤宝平, 刘自然, 秦毅.基于正交半监督局部Fisher判别分析的故障诊断.机械工程学报, 2014, 50(18): 7-13 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201418002.htmSu Zu-Qiang, Tang Bao-Ping, Liu Zi-Ran, Qin Yi. Fault diagnosis method based on orthogonal semi-supervised local Fisher discriminant analysis. Journal of Mechanical Engineering, 2014, 50(18): 7-13 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201418002.htm [29] Cho H W. Nonlinear feature extraction and classification of multivariate data in kernel feature space. Expert Systems with Applications, 2007, 32(2): 534-542 doi: 10.1016/j.eswa.2005.12.007 [30] Yu J B. Local and global principal component analysis for process monitoring. Journal of Process Control, 2012, 22(7): 1358-1373 doi: 10.1016/j.jprocont.2012.06.008 [31] Li G, Qin S J, Ji Y D, Zhou D H. Reconstruction based fault prognosis for continuous processes. Control Engineering Practice, 2010, 18(10): 1211-1219 doi: 10.1016/j.conengprac.2010.05.012 -
图(16) / 表(6)
计量
- 文章访问数: 1275
- HTML全文浏览量: 286
- PDF下载量: 319
- 被引次数: 0
