Research on an integrated ICA-SVM based framework for fault diagnosis

A chemical process has a large number of measured variables, but it is usually driven by fewer essential variables, which may not be measured. Extracting these essential variables and monitoring them will improve the process monitoring performance. In this paper, an integrated framework for process monitoring and fault diagnosis is presented, which combines independent component analysis (ICA) for feature extraction and a support vector machine (SVM) for identification of different fault source. ICA is used to determine the projection coefficient matrix which represents the features characterizing the current operating condition. Well-trained multiple support vector machines use the projection coefficient matrix as their input to identify the faults. The method is proved to be effective by an application to monitoring of the Tennessee Eastman process.