PENG Hong-xing, PAN Li-li, ZHAO Hong-tu. Study on abnormal diagnosis of sensor in monitoring system based on KPCA[J]. 2016, 35(12): 33-36. DOI: 10.13873/J.1000-9787(2016)12-0033-04.
Aiming at problem of low accuracy of fault detection when abnormal data occurs in monitoring system
put forward a detecting method of kernel principal component analysis( KPCA)
which is based on sensor abnormal of monitoring system. The method KPCA detects faults and locates the fault source via the statistics of square prediction error( SPE) and the mean square contribution value from KPCA method
improve problem of low accuracy of PCA applied to fault detecting of nonlinear system. Fault detecting models based on KPCA and PCA are used for simulation and comparison. Experimental results show that KPCA improve accuracy of sensor abnormal diagnosis of nonlinear monitoring system.
Nonlinear process monitoring using kernel principal component analysis [J] . Jong-Min Lee,ChangKyoo Yoo,Sang Wook Choi,Peter A. Vanrolleghem,In-Beum Lee. Chemical Engineering Science . 2003 (1)
Nonlinear Component Analysis as a Kernel Eigenvalue Problem [J] . Bernhard Schö,lkopf,Alexander Smola,Klaus-Robert Mü,ller. Neural Computation . 1998 (5)