Generalized Hebbian Algorithm for fault detection of CSTR model

By applying Generalized Hebbian Algorithm (GHA), this work deals with the problem of on-line process monitoring for a continuously stirred tank reactor (CSTR) model using Principal Component Analysis (PCA) method. Diverse studies have shown the efficiency of PCA for fault detection. However, this method for a large number of samples, becomes difficult to directly solve the eigenvalue problem especially with a large number of samples, which make it not suitable in the cases of on-line process monitoring. In this paper, Iterated PCA have been proposed to alleviate the impact of this problem. This method uses GHA for optimizing the memory efficiency. The simulation results show the effectiveness of the IPCA method in terms of fault detection accuracies, false alarm rates for detection of single as well as multiple sensor faults through its two charts Q and Hotelling T
²
statistics.