Lyapunov Theory Based Adaptive Neural Observers Design for Aircraft Sensors Fault Detection and Isolation

In this research, two novel online fault detection algorithms are proposed for sensor faults in aircraft. Radial basis function neural network (RBFNN) is used as a fault detection technique, for weight updating parameters adaptive learning rates are used instead of fixed learning rates, two different adaptive learning rate strategies are proposed based on Lyapunov functions which are compared to Extended Kalman Filter (EKF) algorithm. Boeing 747-100/200 aircraft is used for testing and validation of these algorithms. All algorithms have the ability to detect various types of faults such as simultaneous, intermittent, abrupt and incipient with high preciseness and accuracy occur in aircraft sensors. The capability of sensors fault detection of all algorithms are compared, it is proved that all algorithms have the ability to detect faults but Lyapunov function theory II based algorithm is more efficient and having a fast response in faults detection as compared to Lyapunov function theory I and EKF based algorithms. It is also proved that the Lyapunov function theory II based algorithm is more effective in reducing the computational time and computation load.