Abstract: Research on typical faults of piston engine,a multi-signal fusion diagnosis method based on wavelet packet analysis and Kernel Principal Component Analysis (KPCA)is proposed.First of all,wavelet packet is used to decompose the original vibration signal,and the fault feature in frequency domain is extracted.Then combined with the vibration signal in time domain and thermal parameters such as unit exhaust temperature,using KPCA to obtain the sensitive characteristics,which will reduce the dimension of parameters.Finally,the Support Vector Machine (SVM)is used to complete the automatic classification of faults.The method is applied to the actual faults of piston engine,and it turns out to have high diagnostic accuracy,which confirmed the validity of this method for feature extraction and fault diagnosis

Key words: piston engine, wavelet packet analysis, Kernel Principal Component Analysis(KPCA), Support Vector Machine(SVM), automatic classification