For the existing shortcomings of current misfire diagnosis algorithm such as complex manual feature extraction process and uncertain feature extraction criteria, a method of diagnosing partial misfire fault for diesel engine was proposed based on the cylinder head vibration signal combined with deep learning algorithm. The misfire fault experiment of diesel engine was carried out under different load and speed conditions, and the collected cylinder head vibration signal was normalized and used as the data set for training the deep learning algorithm. The convolutional neural network and deeper residual network were designed successively. Through the comparison of experiment, it was found that the residual network with more layers could better realize fault diagnosis of diesel engine partial misfire, which had high fault diagnosis accuracy and good generalization performance and hence good diagnosis performance in a wide range of working conditions.