A fault diagnosis method based on the auto regressive(AR) model and the neural n etworks was put forward. Through the simulation of the faults such as the gas le akage and the abnormal lash of a diesel engine, a diagnosis system for the cyli nder head where the faults occured frequently was developed by using the N1447 2 s ignal acquistion modｕle on the LabVIEW platform. An AR model was established us in g the vibration signal samples from the cylinder head, and the AR spectra were e stimated and five feature parameters were extracted. Then, a radialbasedfunc t ion(RBF) network was used to distinguish the fault type. Results show that the m ehtod is effective for developing the online monitoring and diagnosis system o f the diesel engine.