Online thresholding techniques for process

Abstract

Process monitoring via Principal Component Analysis (PCA) and Dynamic Principal Component Analysis (DPCA) suffer from the False Alarms (FAs), Missed Detection (MD) and to the Detection Time Delay (DTD). In this work, a Modified Moving Window PCA (MMWPCA) with Fuzzy Logic Filter (FLF) and its dynamic extension (MMWDPCA) with FLF are proposed to overcome these issues. The techniques are based on PCA, DPCA and Moving Window PCA (MWPCA) to generate adaptive thresholds with fixed statistical models. The applications of the proposed methods have been carried out on the Tennessee Eastman Process (TEP) (both old and revised models), Photovoltaic system and cement rotary kiln. The performances of the developed techniques are compared against recent Fault Detection and Diagnosis (FDD) works. The results demonstrate the superiority of the proposed monitoring schemes in detecting different types of faults with high accuracy and with shorter time delay