To effectively recognize the operating condition of a diesel engine, this paper presented a new method by com bining kernel principle component analysis (KPCA) and support vector machine (SVM) for the pattern recognition. Firstly, feature extraction was used to the diesel engin e vibration signal, and a multipleparameter vector was obtained. Secondly, a new fe ature vector resulted from the feature vector by using KPCA whi ch can demonstrate the operating condition. Redundancy information was removed from the vector during KPCA. Lastly, the condition of the diesel engine was classified by inputting the new feature vector to SVM for train ing and recognizing. The combustion system pattern recognition of a diesel engine in a lab was used as an example to testify the method. The results show that the method is effective for pattern recogniti on of a diesel engine.