Neural Networks for Intelligent Aircraft Fault Tolerant Controllers

This paper presents novel schemes of using neural networks for fault-tolerant controller designs, thereby adding intelligence to existing controllers. The specific application considered is for an aircraft auto landing control under severe wind conditions and when control surface actuators get stuck. First, a neural aided control scheme using feedback-error learning is discussed in detail and then a second scheme using an adaptive back-stepping neural control scheme for handling actuator failures is briefly highlighted. Simulation studies using the full six-degree-of-freedom nonlinear aircraft model show that the above controllers are able to successfully stabilize and land the aircraft within tight touch down dispersions.