Fault diagnosis of the light-rail´s cast steel pedestal system based on KPCA and One-class SVM

The scarcity of fault samples often occurs in the fault diagnosis of Chongqing light-rail´s cast steel pedestal system. In the case of this situation, this paper first puts forward a fault diagnosis method based on One-class Support Vector Machine (One-class SVM). This method can build up one-class classifier to distinguish between normal condition and abnormal condition as long as the normal data samples are provided. In the process of test, Kernel Principal Component Analysis (KPCA) is used as data preprocessing to extract the features from vibration impulse response signal as the input of One-class SVM classifier. The test result shows that the feature extraction based on KPCA can concentrate fault information more effectively and make the the One-class SVM classifier identify the fault samples more accurately.