ICA-SVM based fault diagnosis method for complex chemical process

An integrated fault diagnosis method based on independent component analysis (ICA) and support vector machines (SVM) is proposed to resolve the problems of the difficulty in fault diagnosis for complex operation and multi-loop controls of chemical industry process. The basic idea of the proposed diagnosis method is to use ICA arithmetic to extract the essential independent components. And, I², I_e² and SPE charts are proposed as on-line fault detecting strategy. The contribution chart of every monitoring variable to I², I_e² and SPE are calculated separately using the gradient algorithm, and used to extract the preliminary possible fault resource by monitoring the change of contributions. Finally, faults are diagnosed further from possible fault resource using binary tree SVM. The proposed fault diagnosis method is proved to be effective by simulation with the data from a real fault in an industrial butadiene distillation column.