An intelligent neuro-controller based on system parameter estimation

To enhance intelligent control system performance, an on-line procedure for sensing and adapting to changes in the dynamics of the controlled plant is essential. A methodology that uses a neural network for plant failure accommodation is presented. The main idea is to constantly monitor system input-output signals for off-nominal behavior (failures) and to use this information to generate appropriate control actions. Based on this idea, an integrated, intelligent neuro-control system consisting of a baseline controller, a system parameter estimator, a failure detector, and a failure accommodator is constructed. Using a continuous stream of plant input-output data, the parameter estimator computes plant parameter estimates and sends them to the failure detector. The failure detector analyzes the system's condition using these parameter estimates and reports its results to the failure accommodator. Using these results, the failure accommodator computes baseline controller coefficient values that are appropriate for the current plant dynamics. New coefficient values are continuously downloaded to the baseline controller to insure stable system operation and to accommodate failures in the plant as they occur. Results from a preliminary simulation, conducted on the control of a tilt-rotor airplane, showed that the intelligent controller is able to maintain system stability even in cases of severe changes in the airplane dynamics.