Comparative analysis between data-driven and visual-based approach in drone propeller imbalance detection

Abstract

Early monitoring of the drone’s critical components, such as the motor, propeller, and electronic speed controller (ESC), can prevent unwanted accidents from happening to the drone. A failure of one of the drone’s components may lead the drone to crash. The fault detection approach can be made by analyzing certain parameters such as vibration, acoustic, current, and speed. In this paper, a comparative analysis between data-driven and visual-based approaches in detecting an imbalanced propeller in the drone is presented. The analysis will focus on vibration and acoustic parameters, with the use of CZN-15 and ADXL335 MEMS sensors to capture vibration and acoustic data, respectively. The Silky EvCam vibration camera will be used for the visual-based approach. Experimental results demonstrated that both data-driven and visual-based approaches could detect the imbalanced UAV propeller. There is a shift of vibration frequency when imbalances occur in the UAV propeller. However, in the ground and thrust modes, the acoustic-based method has an accuracy of 54% and 93%, respectively, which is lower compared to the vibration-based (100% accuracy in both modes) and visual-based techniques.