基于组合核函数的类均值核主元故障诊断方法
|
摘要
慢漂移是指系统参数从额定值发生偏移的现象,系统正常的慢漂移影响故障诊断的结果。针对单一核函数对慢漂移系统故障误诊率高的问题,提出了基于组合核函数的类均值核主元分析故障诊断方法。首先,使用组合核函数将原始样本数据映射到高维特征空间,求出各类映射数据的类均值矢量;然后在类均值矢量张成的子空间上对类均值矢量进行主元分析;最后,利用建立的类均值核矩阵构建类均值核主元算法。将该方法应用于田纳西-伊斯曼(TE)和蒸馏塔过程,仿真结果显示,该方法提高了故障的检测率,柔性因子的引入满足故障监控模型对灵敏度、鲁棒性的动态平衡要求,具有较好的过程监控性能。
Slow drift is the phenomenon that the system parameters are shifted from the rated value, the normal slow drift of the system affects the result of fault diagnosis. For the problem that the fault misdiagnosis rate of single kernel function for the slow drift system is high, the paper proposed a fault diagnosis method with class mean kernel principal component analysis based on the combined kernel function. Firstly, the original sample data were mapped to the high-dimensional feature space by using the combined kernel function, and the class mean vector of all kinds of mapping data was obtained. Then, the principal component analysis was carried out for the class mean vector on the subspace of the class mean vector. Finally, the class mean kernel principal component algorithm was constructed based on established class mean kernel matrix. The method was applied to the process of Tennessee-Eastman(TE) and distillation column. The simulation results show that the proposed method improves the detection rate of the fault, and the introduction of the flexible factor satisfies the dynamic balance requirement of the fault monitoring model for the sensitivity and robustness, and the method has good performance of process monitoring.
Slow drift is the phenomenon that the system parameters are shifted from the rated value, the normal slow drift of the system affects the result of fault diagnosis. For the problem that the fault misdiagnosis rate of single kernel function for the slow drift system is high, the paper proposed a fault diagnosis method with class mean kernel principal component analysis based on the combined kernel function. Firstly, the original sample data were mapped to the high-dimensional feature space by using the combined kernel function, and the class mean vector of all kinds of mapping data was obtained. Then, the principal component analysis was carried out for the class mean vector on the subspace of the class mean vector. Finally, the class mean kernel principal component algorithm was constructed based on established class mean kernel matrix. The method was applied to the process of Tennessee-Eastman(TE) and distillation column. The simulation results show that the proposed method improves the detection rate of the fault, and the introduction of the flexible factor satisfies the dynamic balance requirement of the fault monitoring model for the sensitivity and robustness, and the method has good performance of process monitoring.
引文
[1] 纪洪泉,何潇,周东华. 基于多元统计分析的故障检测方法[J].上海交通大学学报, 2015,49(6):842-848,854.
[2] Jolliffe, Ian. Principal Component Analysis[M]. Springer-Verlag, 2005.
[3] B Sch?lkopf, A Smola, K R Müller. Nonlinear component analysis as a kernel eigenvalue problem[J]. Neural Computation, 2014,10(5): 1299-1319.
[4] 万家强,王越,刘羽. 改进KPCA对分类数据的特征提取[J]. 计算机工程与设计, 2010,31(18):4085-4087,4092.
[5] 余文利,余建军,方建文. 一种新的基于KPCA和改进ε-SVM的入侵检测模型[J]. 计算机工程与应用, 2015,51(11):93-98.
[6] C Lin, et al. Research on PCA and KPCA Self-Fusion Based MSTAR SAR Automatic Target Recognition Algorithm[J]. Journal of Electronic Science and Technology, 2012,10(4):352-357.
[7] Jiang Qingchao, Xuefeng. Statistical Monitoring of Chemical Processes Based on Sensitive Kernel Principal Components[J]. Chinese Journal of Chemical Engineering, 2013,21(6):633-643.
[8] 张恒,赵荣珍. 故障特征选择与特征信息融合的加权KPCA方法研究[J]. 振动与冲击, 2014,33(9):89-93.
[9] 赵小强,杨武,薛永飞. 可变窗自适应核主元分析的化工过程故障诊断算法[J]. 东南大学学报(自然科学版), 2013,43(s1):93-97.
[10] 欧阳高强. 基于相对主元分析的故障诊断方法研究[D]. 北京化工大学, 2014.
[11] 张敏龙,等. 分步动态自回归核主元分析及其在故障诊断中应用[J]. 计算机应用, 2016,36(5):1464-1468.
[12] 刘春燕,于春梅. 基于改进核主元分析的TE故障诊断[J]. 计算机测量与控制, 2016,24(10):36-38+41.
[13] J M Lee, et al. Nonlinear process monitoring using kernel principal component analysis[J]. Chemical Engineering Science, 2004,59(1): 223-234.
[14] N Zhang, et al. Fault detection of chiller based on improved KPCA[C]. Control and Decision Conference. IEEE, 2016:2951-2955.
[15] J J Downs, E F Vogel. A plant-wide industrial process control problem[J]. Computers & Chemical Engineering, 1993,17(3):245-255.
[2] Jolliffe, Ian. Principal Component Analysis[M]. Springer-Verlag, 2005.
[3] B Sch?lkopf, A Smola, K R Müller. Nonlinear component analysis as a kernel eigenvalue problem[J]. Neural Computation, 2014,10(5): 1299-1319.
[4] 万家强,王越,刘羽. 改进KPCA对分类数据的特征提取[J]. 计算机工程与设计, 2010,31(18):4085-4087,4092.
[5] 余文利,余建军,方建文. 一种新的基于KPCA和改进ε-SVM的入侵检测模型[J]. 计算机工程与应用, 2015,51(11):93-98.
[6] C Lin, et al. Research on PCA and KPCA Self-Fusion Based MSTAR SAR Automatic Target Recognition Algorithm[J]. Journal of Electronic Science and Technology, 2012,10(4):352-357.
[7] Jiang Qingchao, Xuefeng. Statistical Monitoring of Chemical Processes Based on Sensitive Kernel Principal Components[J]. Chinese Journal of Chemical Engineering, 2013,21(6):633-643.
[8] 张恒,赵荣珍. 故障特征选择与特征信息融合的加权KPCA方法研究[J]. 振动与冲击, 2014,33(9):89-93.
[9] 赵小强,杨武,薛永飞. 可变窗自适应核主元分析的化工过程故障诊断算法[J]. 东南大学学报(自然科学版), 2013,43(s1):93-97.
[10] 欧阳高强. 基于相对主元分析的故障诊断方法研究[D]. 北京化工大学, 2014.
[11] 张敏龙,等. 分步动态自回归核主元分析及其在故障诊断中应用[J]. 计算机应用, 2016,36(5):1464-1468.
[12] 刘春燕,于春梅. 基于改进核主元分析的TE故障诊断[J]. 计算机测量与控制, 2016,24(10):36-38+41.
[13] J M Lee, et al. Nonlinear process monitoring using kernel principal component analysis[J]. Chemical Engineering Science, 2004,59(1): 223-234.
[14] N Zhang, et al. Fault detection of chiller based on improved KPCA[C]. Control and Decision Conference. IEEE, 2016:2951-2955.
[15] J J Downs, E F Vogel. A plant-wide industrial process control problem[J]. Computers & Chemical Engineering, 1993,17(3):245-255.