The identification method of multi-source data such as engine gas path parameters, control system parameters, lubricating oil parameters and vibration parameters is studied. Based on a large number of flight test data under the normal operation settings of the engine, the parameter identification models of the all settings and whole envelope operation of the engine are established by using the Artificial Neural Network. The model calculation results are in good agree with the flight test results, which indicates that the parameter identification method is reasonable and accurate. Models are used as engine setting reference and main diagnosis basis, assisted by ground maintenance information, fault alarm generated by FADEC and other information sources. By judging the sequence of parameters beyond the reference scope, the fault diagnosis of control, gas path, wear and vibration in flight test is realized. The research on fault prediction is carried out based on the parameter identification models. Compared with the traditional over limit alarm mode after the limit is exceeded, the method can detect the early signs of the fault, so as to realize the prediction and warning of the fault. The method of fault diagnosis and prediction is verified in the off-line and on-line settings of flight test, and the result is good, which shows the feasibility, effectiveness and practicability of the method proposed in this paper.