The fault detection and fault-tolerant control ability of electromechanical actuator is very important especially for safety-oriented aircraft industry. In this paper, a novel active fault-tolerant control strategy based on double neural networks combined with improved fast integral terminal sliding mode control is proposed. The faults only related to the system states are considered, which is very common in electromechanical system. Considering their strong approximation ability, two neural networks with different inputs are added in a high-gain observer to estimate system model uncertainties and fault respectively. Thus, the fault detection observer can be sensitive to faults without false alarm. An integral term of position error is introduced into a fast terminal sliding surface to design an improved controller. Model uncertainties and fault are compensated for in the controller with feedforward cancellation technique based on neural networks’ estimation. The Lyapunov stability theory can guarantee the bounded stability of the proposed control strategy. Extensive comparative simulations and experimental results are obtained to verify the better performance of the proposed control strategy compared with some other traditional fault tolerant strategy.