Aircraft sensor fault detection and system reconstruction based on artificial neural networks

The purpose of this study is to detect and reconstruct a fault that may occur in the angle of attack, airspeed and altitude sensors, which are important components of the air data system for aircrafts, in a timely intervention without the need for any additional sensor measurement and without a false alarm. Fault detection and reconstruction systems used for modern aircrafts in the literature are methods such as majority voting and weighted mean based on the majority accuracy of the primary/redundant sensors on the aircrafts. Instead of using these methods based on hardware redundancy, by the designed analytical redundant system, while saving weight and fuel burned, the aircraft can travel longer distances and thereby reducing CO2 emissions and noise, the aircraft can perform more environmentally friendly flights. In addition, with the removal of the redundant sensors, the purchase cost of the aircraft to the airline company and the maintenance costs of the related parts can be reduced. The real flight data of a commercial aircraft collected from a local airline were used for the system design. Correlation analysis was used to select data related to the angle of attack, airspeed and altitude sensors. Machine learning-based methods were used to detect and reconstruct the fault. MATLAB program was used for the related operations. The fault situations that may occur in the angle of attack, airspeed and altitude sensors were modeled and different fault scenarios that can be encountered in all flight phases were examined. Fault detection was performed in the range of 0-2 s without false alarms. While reconstructing the fault, it is ensured that the system output follows the estimated sensor data.