A Novel Method of Fault Detection Method for TEP based MIPCR

Abstract

Principal component regression (PCR) is not only a kind of multivariate statistical method, but also a type of data-driven method. The improved PCR (IPCR) optimizes the performance of fault detection for Tennessee Eastman process (TEP). IPCR could solve the unsatisfactory detection performance generated by the incomplete sample decomposition. Multiple IPCR (MIPCR) is a novel improved method relative to IPCR. It uses multiple quality variables to detect product quality at the same time. And the results, obtained via MIPCR, are fused. Then screening the variables via the fault performance is done. Simulations for Tennessee Eastman process (TEP) are presented with PCR, IPCR and MIPCR. Via the simulations, the validity and superiority of MIPCR are all verified.