针对内燃机缸盖振动信号的非平稳时变特点，通过给出小波包变换的一种改进算法，提出从振动信号的小波包分解系数中实现整循环征兆提取和故障识别的方法．实验结果表明了该方法的工程有效性． Aiming at the non-stationary and time-varying characteristics of cylinder head vibration signal of an internal combustion engine, an improved algorithm of wavelet packet transform is proposed and a method to extract the whole cycle symptom and identify the fault from the wavelet packet decomposition coefficients of the vibration signal is proposed. Experimental results show the effectiveness of this method.