PreSound: UAV Diagnostic System Enabled by Acoustics and Vibration-Based Machine Learning

PreSound identifies specific defects on aircraft through intelligent analysis of vibration and acoustic measurements. It operates in two modes: a preflight inspection routine that uses an aircraft#39;s own propellers as a vibration stimulus, and passive in-flight monitoring, identifying defects and alerting personnel or triggering automated contingency measures. Although there are commercial systems that use vibration and acoustic data to guide maintenance intervals, none are able to diagnose defects before takeoff, and most of their focus is on guiding additional human inspection and maintenance activities rather than providing input into automated systems. Future high assurance aviation systems will require the higher fidelity and more effective automated health diagnostics that PreSound provides.During Phase I, GreenSight successfully proved the feasibility of PreSound in its small UAS implementation. We were able to demonstrate using a preflight ground testing routine not only that PreSound can reliably detect the presence of an off-nominal condition, but that it can identify the specific nature of the defect. nbsp;With future extension to in-flight use in mind, we formulated a novel preprocessing method that is especially immune to the uncontrolled airframe stimulus that will be encountered in actual flight conditions. In Phase 2, we intend to develop and commercialize two variants of PreSound. PreSound-S is optimized for sUAS and will run entirely on the native flight control and computing hardware that is already on the aircraft, gathering accelerometer data from the flight controller#39;s IMU. PreSound-L will be optimized for larger, more complex, passenger-carrying aircraft. It will use multiple dedicated yet inexpensive and low-SWAP hardware monitoring and computing nodes to run the PreSound detection algorithms. nbsp;