The electromechanical actuation system is the last link for completing all the task operation in the modern airplane, and its working condition directly influences the flight safety. According to the characteristic that the fault information of different components of the aircraft electromechanical actuation system is in different bands, a new pattern recognition method of fault diagnosis based on "energy－fault state" is proposed. First the original signal is orthogonally decomposed into different bands by wavelet packet transformation, and the target fault features are extracted by the chosen eigenvector. Then the mapping relation between energy and fault state is obtained, which can be used as the judgment of fault degree through the comparison with the standard sample of the corresponding eigenvector. Finally the paper verifies the correctness and effectiveness of the method by the simulation example of the fault diagnosis in inter-turn short circuit