Fault Detection of Complex Multi-stage Process Based on Improved PCA
To solve the problem that it is difficult to monitor complex multi-stage processes effectively,the time-space nearest neighborhood standardization and principal component analysis(TSNS-PCA)method is proposed for fault detection.TSNS uses the means and standard deviations of the nested neighbor set from time to space to standardize the samples of each stage.TSNS can Gaussianize the multi-stage data into single-stage data,separate the fault samples,and remove the temporal correlation between sam-ples at previous and subsequent moments in the process data.TSNS helps PCA to overcome the effects of dynamics,non-linearities,and multi-phase features effectively,significantly improving the fault detection rate of PCA.Through the fault detection experiment of the penicillin fermentation process,TSNS-PCA is compared with some classical methods,which proves that it has a higher fault detection rate in complex multi-stage processes.
万方数据
维普
