Real-time quadrotor actuator fault detection and isolation using multivariate statistical analysis techniques with sensor measurements

Quadrotor, as an unmanned aerial vehicle, has significant potential among military and commercial applications and is utilized in various fields. As the quadrotor usage is more popularized, fault diagnosis becomes important for safe quadrotor flight. In this work, we use data-driven approaches which generally do not require a complex model of quadrotor to detect and to isolate an actuator fault in a quadrotor. We made a circuit that artificially blocks a motor signal to stop the propeller motor and collected real-time sensor measurements in a normal condition and each actuator's fault condition. Then, we applied various statistical analysis techniques on the collected data to train the diagnosis model and used this model on the new data to test and to compare the performance of the techniques. Those techniques are linear discriminant analysis, principal component analysis, multi-principal component analysis, fisher discriminant analysis, partial least squares regression, and canonical variate analysis. Among the techniques, partial least squares regression shows the best performance for detecting and isolating an actuator fault of a quadrotor.