Online Fault Detection Methods and Fault Detection Indices Based on PCA Approach (original) (raw)

— For the improvement of reliability, safety and efficiency advanced methods of supervision, fault detection and fault diagnosis become increasingly important for many technical processes. This holds especially for safety related processes like aircraft, trains, automobiles, power plants and chemical plants. The fault detection based upon multivariate statistical projection method such as Principal Component Analysis (PCA) has attracted more and more interest in academic research and engineering practice. The PCA is an appropriate method for the control of the process based on selection of an optimal number of principal components. In this paper we present the design and a comparative study of offline fault detection indices based on PCA method and adaptive fault detection techniques which used the PCA method. These indices are Squared Prediction Error (SPE), Hotelling's Statistic (T 2), Filtred Squared Prediction Error (Filtred SPE) and i D Index. These indices and the adaptive detection methods are evaluated by handling a numerical example and a Continuous Stirred Tank Reactor (CSTR) benchmark.