The paper concerns the problem of development of the technique for rapid fault detection for electromechanical actuator (EMA) of the aircraft related to dissipative torque change. Increased dissipative torque is a common cause of EMA failure. To address this problem, a data processing scheme has been suggested for EMA’s operation using neural networks. The work describes the feature filtering methods used for reduction of neural network input vector dimension. Trained neural networks are designed to develop algorithms for determining the EMA’s technical state. The algorithms are proposed to be used for flight safety in the health usage monitoring systems (HUMS) of the aircraft. The paper shows the computing experiment results based on the data obtained by using computational model of unmanned aircraft’s EMA while processing the take-off cyclogram. The data includes test and training sets for neural network. The computing experiment results shows efficiency of fault detection related to dissipative torque change.