Air Data Sensor Fault Detection and Diagnosis Using Robust Unscented Kalman Filter

The air data system measures and provides vital information like an aircraft's speed, altitude, and other safety parameters essential for flight. Errors in such critical system components result in loss of control of the aircraft, leading to aircraft crashes. Thus, redundancy for this critical component is highly important. The proposed work introduces a new method for identifying problems with the air data sensor in an airplane. Current techniques involve a complicated voting system that requires multiple sensors to function properly. However, this new approach uses analytic redundancy to monitor the sensor's performance continuously, which is more desirable than relying on air data probe measurements. The method involves using a mathematical model that considers only the inertial measurement unit and navigation system data to estimate air data and detect faults in the sensors. The robust three-stage unscented Kalman filter is used to provide an accurate estimation. By comparing the estimated data with the physical measurements, the algorithm can track the performance of each sensor and identify any faults. The algorithm is tested in X-Plane software and validated using real-time data from a Cessna Citation II transport aircraft. A sensor stuck fault was introduced twice in the 60-second run time. The UKF outperformed the novel EKF in most of the cases and showed its robustness during various fault conditions with a 1–2 % deviation in RMSE values. The results obtained showed that the estimated data can be used as a replacement for a redundant sensor, reducing the complexity and cost of these sensors.