In order to handle the invalidation problem of fault detection when the modeling samples got from the real
industrial process is impure, a new kernel principal component analysis(KPCA) based fault detection algorithm integrating.Fisher discriminant analysis-possibilistic C-means clustering(FDA-PCMC) is proposed. FDA based feature extraction, pre-classification and PCMC based clustering are proposed to classify and purify the modeling samples effectively, and KPCA is used for real-time fault detection. Simulation results of Tennessee Eastman(TE) process show the feasibility and effectiveness of the algorithm.