Nonlinear Process Fault Detection Based on Crayfish Optimization Algorithm with MKPCA

In order to solve the problem of nonlinear feature extraction in industrial processes, the traditional kernel principal component analysis method only considers the global nonlinear features of the data, and does not consider the noise and local features when performing fault detection. In this paper, a fault detection method combining Crayfish Optimization Algorithm (COA) and Multiple Kernel Principal Component Analysis (MKPCA) is proposed. Firstly, MKPCA is used to extract nonlinear features from the normalized raw data, and COA is used to optimize the kernel function weights of MKPCA to find the optimal weight ratio. Then, the T2 and SPE statistics are constructed for the principal metaspace and the residual space, and the control limits are calculated. Finally, the effectiveness of the COA-MKPCA algorithm is verified by numerical cases and Tennessee Eastman (TE) process simulation experiments.