Combination method of support vector machine and fisher discriminant analysis for chemical process fault diagnosis

For chemical process, a new fault diagnosis method based on multi-phases is presented to overcome its difficulty in nonlinear and non-uniform sample data. Support vector machine is first used for phase identification, and for each phase, fisher discriminant analysis is developed to analyze and recognize fault patterns. Variable weighted discriminant matrix and similarity measurement based on manifold distance are proposed to enhance the incremental clustering capability of FDA. The proposed method is applied to citric acid fermentation process, and the comparison results indicate that the proposed algorithm has better capability to classify fault samples as well as high diagnosis precision.