Due to the limited telemetry data for the turbojet engine during the flight test, it is difficult to locate accurately the turbojet engine’s fault. After the de-noising extraction with wavelet transform of the data, such as speed of the turbojet engine, former pressure of nozzle etc, the singularity is found and the data file is formed through the system noise filtering. Using the BP network pattern classification function, the moment of failure occurs and default parts are detected. With the ground test database in the development process, it can locate accurately the turbojet engine defaults parts and types. The results show that the method is effective and can be applied in other subsystem fault analysis.