Decentralized Fault Diagnosis of Large-scale Processes Using Multiblock Kernel Principal Component Analysis

In this paper, a multiblock kernel principal component analysis (MBKPCA) algorithm is proposed. Based on MBKPCA, a new fault detection and diagnosis approach is proposed to monitor large-scale processes. In particular, definitions of nonlinear block contributions to T2 and the squared prediction error (SPE) statistics are first proposed in order to diagnose nonlinear faults. In addition, the relative contribution, which is the ratio of the contribution to the corresponding upper control limit, is considered to find process variables or blocks responsible for faults. The proposed method is applied to fault detection and diagnosis in the Tennessee Eastman process. The proposed decentralized nonlinear approach effectively captures the nonlinear relationship in the block process variables and shows superior fault diagnosis ability compared with other methods.