Performance in different failure conditions of the predictor-corrector neural controller within the NASA IFCS F-15 WVU simulator

The objective of this paper is to present from the evaluation of a neural control scheme implemented within the NASA IFCS F-15 project. In particular, the neural network controller is based on an adaptive predictor-connector control strategy. The study is conducted within the WVU IFCS F-15 simulation environment, which enables to evaluate a wide range of failure scenarios. Analysis of the controller performance is performed in situations where the state variables vary within a wide portion of the flight envelope and in different failure conditions. Post failure performance evaluation concerns maneuvers accomplished both with the pilot-in-the-loop and with pre-recorded reference signals. Failure test-cases are grouped in three categories: actuator lockage at an imposed deflection, actuator lockage at the current deflection, missing of a control surface portion. Results demonstrate the capability of the neural controller in dealing with slight and hazardous damages.