신경회로망을 이용한 재형상 비행제어법칙 설계

When control surface failure occurs, it is conventional to correct a current control or to transform to other control. In this paper, instead of adopting a conventional way, a reconfiguration method which compensate the failure with alternative control surface deflection, depending on the level of failure, by using neural network and PCH(Pseudo-Control Hedging). The Conroller is designed of inner-loop(SCAS : Stability Command Augmentation System) with DMI(Dynamic Model Inversion) and outer-loop with Y axis acceleration feedback for a coordinate turn. Additionally, double PCH method was adopted to prevent actuator saturation and input command was generated to compensate for failure. At the end, The feasibility of the method is validated with randomly selected failure scenarios.