The mechanical faults at different parts of a diesel engine are characterized by confusion and nonstationary cycling. A fault diagnose method based on polar coordinate enhancement of time-frequency diagram is proposed. The Gabor transform is applied to achieve the time-frequency feature of the vibration signals; then, the time-frequency features are re-sampled at constant angular inter val and transformed to the polar coordinate as enhanced feature of certain areas. Thus, the periodic transient features are enhanced. Polar diagram areas energy of 6 rotation cycle of different diesel engine status is extracted as the fault feature, which is used to train support vector machine (SVM) for the pattern recognition. Experimental result shows that this method highlights the features of five typical engine faults, extracts the fault feature information, and distinguishes different types of wear fault.