Data-driven Diagnosis of Quadcopter Thrust Fault Using Supervised Learning with Airframe Vibration Signals

This paper presents a data-driven fault diagnosis for a quadcopter that utilizes the principal direction vector of airframe vibration signals computed by principal component analysis (PCA). The fault diagnosis is based on classification using a supervised learning machine via support vector machines (SVM) and artificial neural networks (ANNs) with the gyroscope and accelerometer datasets. It is used to classify faulty and nominal flight conditions. The effectiveness of the suggested approach is validated using airframe vibration data obtained from a quadrotor hovering experiment.