Classification and Assessment of Propeller Faults in Electric Unmanned Aerial Vehicle Drive Trains

Abstract

Propellers are critical to the safe operation of multicopter unmanned aerial vehicles (UAVs), as faults can decrease the efficiency of the propulsion system and affect flight performance. Depending on the type and extent of the fault, the effects can range from a slight reduction in performance to a significant loss of thrust that could compromise safety.  Because of the limited amount of sensor data available on board a UAV, propeller damage cannot be measured directly. Therefore, a data-based prediction using available sensors is required. This paper focuses on establishing and predicting a health index for damaged propellers.A test bench is used to investigate the effects of two different types of damage: broken propeller tips and notches at the leading edge. Each type of damage is examined at three levels of severity.  Based on sensor data collected from the test bench, a health index is defined to characterize the remaining performance of the damaged propellers. A two-stage approach for the data-based health prediction is implemented by first classifying the type of the propeller faults, and then employing a random forest regressor to estimate the remaining health.