Multi-section classification improving integrated fault diagnosis method based on independent component analysis and support-vector-machines

The integrated diagnosis method of independent component analysis (ICA) and support-vector-machines (SVM) is improved by multi-section classification. Fault classification model of SVM is designed for each section in the high dimensional characteristic space. By diagnosing the fault type in different section, we improve the ICA-SVM fault diagnosis performance. This method has been applied to diagnose 19 types of valve failures on the dynamic actuator reference platform (DAMADICS). Simulation results show that the ICA-MSVM fault diagnosis method based on multisection classification effectively improves the accuracy of fault diagnosis.