Abstract: In order to achieve the purpose of diagnosing the engine fault without disassembly, and thus improve the diagnostic efficiency, this article realized engine condition monitoring and fault diagnosis by collecting vibration signal of an engine cylinder head. With the help of acceleration sensor BZ1185, data acquisition card PCI8210 and computer hardware and so on, the acquisition system of the engine vibration signal was constructed. Based on LabVIEW software platform, virtual instrument analysis software was developed. The vibration signals were used for time-domain analysis, correlation-domain analysis (autocorrelation analysis, cross-correlation analysis), frequency-domain analysis (Fourier transformation, auto-power spectrum analysis, cross-power spectrum analysis, cepstrum analysis) JTFA-Joint Time-frequency Analysis (DSTFT-discrete short-time Fourier transformation, wavelet analysis). Therefore the offline fault diagnosis was realized. Taking the 4G65 engine as the research object, the fault of engine cylinder broken down was studied at different speeds by means of power spectrum and wavelet analysis. The results show that it is feasible to apply the real-time state monitoring and fault diagnosis by the system for the 4G65 engine. The research can also provide a reference for the development and application of other engine fault diagnosis.