From sound to vision : applying beamforming to reduce parasitic noise for fault detection

Abstract

The number of applications of drones or unmanned aircraft systems (UAS) has rapidly increased over the last years. The widespread commercial use of UAS and their reliable and safe operation requires novel maintenance, repair and overhaul (MRO) schemes.Most commercial UAS consist of multiple rotary propellers, which are prone to damage, wear and tear. Propeller blade damage can cause increased mechanical stress on UAS components, performance degradation and decreased stability. Acoustic camera measurements of partially damaged blades show higher sound pressure levels (SPL) at the blade tips and an individual frequency response.With the rotating propellers being the main source of emitted sound, an acoustic detection system is proposed to identify damaged propeller blades without the need to intervene in the UAS’s hardware or software and seamlessly integrate the inspection with the UAS operation. To isolate the acoustic signature of the individual propeller blades and reduce parasitic environmental noise, an acoustic camera (CAE Systems Bionics M112) is used. The output data of the microphone array is processed using beamforming algorithms to isolate the individual propeller sound. In the following step, the data is processed by a neural network, which is trained to diagnose the propeller’s health state.