Study On Decentralized Fault Diagnosis Based On Multiblock Kernel Methods

Since industrial processes become more and more complex, the requirements of the stability, efficiency and safety of production are increasing. In order to get more accurate and timely diagnosis of complex industrial process fault, it is necessary to study more comprehensive theory and method of fault diagnosis.The paper first briefly introduced the development of fault diagnosis. Then, this paper discussed the detail of data-based fault diagnosis method. Highlighted principal component analysis (PCA), and analyzed the deficiency in dealing with the nonlinear data. In order to get better fault diagnosis performance in the monitoring of large scale nonlinear process, this paper combines the kernel method and multi-block method, proposed multi-block Kernel PCA (MBKPCA) based and multi-block kernel partial least squares (MBKPLS) based of the fault diagnosis method.When the industrial processes data have nonlinear characteristic, multiblock kernel methods can afford better fault diagnosis ability. In particular, definitions of nonlinear block contributions to SPE and T2 statistics are proposed in order to diagnose nonlinear faults. The proposed MBKPCA based method is applied to fault detection and diagnosis in the Tennessee Eastman process, and MBKPLS based method is applied to the process monitoring in a continuous annealing process. The results indicate that the proposed methods are effective.