This study proposes an automatic diagnosis system for engines based on sound signal analysis. First, the wavelet transform and the FFT frequency analysis are employed to extract the features of the signal of tappet clicks and misfire. Comparing results of the wavelet transform and the FFT frequency analysis, it is found that the wavelet transform method is more effective. Second, through experimental results, it is revealed that even with the high level noise in engine assembly site, the tappet clicks can be detected by means of the wavelet transform analysis, so long as the level of the noise is the same as that of the tappet clicks. The expermental results pvove the effectiveness of the proposed algorithm.