Abstract: In order to realize the minimization of fault diagnosis system of diesel engine and generalization of system-level diagnosis model, a data structure and a model algorithm of fault diagnosis system of diesel engine were presented, which were mainly embodied in machinery fault diagnosis data processing based on extensive research on the algorithm of characteristic value extraction, the algorithm of main characteristic value extraction and the neural network expert system. The software package of machinery fault diagnosis system (MFDS) was developed in combination with MATLAB toolkit. 6135 diesel engine was used for the test. The measured values were studied and diagnosed by MFDS. The results showed that the fault diagnosis system could be used to give the diagnosis results quickly and accurately. The study has practical significance to improve the fault diagnosis theory and level of complicated machinery such as diesel engine.