Research on Fault Diagnosis of the Fuel System Based on Improved Multi-Output SVM

In order to realize accurate diagnosis of five common faults of PT fuel system, fifteen time domain feature parameters of the pressure signal which influenced the fuel system directly were selected. A multi-factor intelligent diagnosis model of PT fuel system was built based on KPCA-LSSVM. Firstly, the reduction processing of time domain feature parameters was carried out according to KPCA theory, and the redundant information was eliminated. Secondly, first two principal components were obtained after reduction and input into LSSVM model to train and learn, Finally, the intelligent diagnosis model was established. One hundred and fifty groups of pressure signal were collected, the model was trained and tested according to the proportion of 2:1, and the test results were compared with the results of other five models, The results showed that the KPCA can reduce the data information effectively, the fault diagnosis method based on KPCA-LSSVM is faster and has higher classification accuracy.