Electromechanical actuator is gradually applied in new more-electric aircraft. However, sensor failure in EMA system has a great impact on the normal operation of the system. Considering the problem of the correct sensor determination after the redundant sensors fault and the signal recovery after complete loss of signal, the dynamic neural network was used to recover the sensor signal, so as to solve the problem of faulty input signals after the signal loss. The differences between the computed results of the nonlinear and the NARX dynamic neural network were compared. The simulation and test results show that the NARX can be used to effectively and accurately perform signal recovery under signal loss, and its recovery result is better than that of the nonlinear dynamic neural network considering the linear regression， error self-correlation and cross-correlation coefficient.