Affiliations: [a] School of Automation, Northwestern Polytechnical University, Xi’an, China
| [b] Department of Aeronautics and Astronautics, Institute of Space Technology, Islamabad, Pakistan
Correspondence:
[*]
Corresponding author. Muhammad Taimoor, E-mail: [email protected].
Abstract: An online fault detection, isolation, and reconstruction strategy is proposed for actuators and sensors fault detection of an aircraft. For increasing the fault detection capabilities, the Extended Kalman Filter (EKF) is used for the weight updating parameters of multi-layer perceptron (MLP) neural network. The main purpose of using the EKF is to make the weight updating parameters of MLP adaptive in order to increase the fault detection, isolation and reconstruction preciseness, efficiency and rapidness compared to the conventional MLP where the fixed learning rate due to which it has slow response to faults occurrence. Because of the online adaptation of weighting parameters of MLP, the preciseness of the faults detection is increased. For testing and validation of the proposed strategy, the nonlinear dynamics of Boeing 747 100/200 are used. Results demonstrate that the proposed strategy has better accuracy and rapid response to fault detection compared to convention multi-layer perceptron neural network based faults detection schemes.
Keywords: Actuators, sensors, fault detection and isolation, aircraft, neural networks, nonlinear systems
