Fault Detection for the Chemical Process Based on LGRSAE Algorithm
Aiming at the conventional linear dimension-reducing method's incapability in extracting local and global structural features of the real complex nonlinear industrial data effectively,a local and global stack-retaining auto-encoder(LGRSAE)and a process fault detection method based on LGRSAE were devel-oped.In addition,the objective constraints of the locality preserving projection(LPP)algorithm and the principal component analysis(PCA)algorithm were introduced into the objective function of the auto-en-coder(AE)to extract the features related to the local and global structure of the data,including having LGRAE stacks constructed to form LGRSAE to extract features related to deep local and global structure of the process data.The fault detection results on the TE(Tennessee-Eastman)process data set show that,the average fault detection rate of the feature extraction method of LGRSAE is higher than those of the MRSAE and KPCA algorithms and the false alarm rate is lower.
fault detectionstacked auto-encoderfeature extractionlocal and global structure preserv-ingdata dimension reduction
万方数据
维普
