GREENSIGHT INC. — National Aeronautics and Space Administration SBIR Phase I: A2

Description

A small, inexpensive UAV that takes off, performs a mission, lands, recharges, and safely stows itself has myriad applications.nbsp; But such systems require pre-flight inspections conducted by humans. This requirement is a major barrier to fully automated UAV operations.nbsp; To address this, we propose ldquo;PreSoundrdquo;: a system that uses vibration and acoustic data, combined with machine learning, to replace preflight visual inspections.nbsp; PreSound does not add any significant equipment to the UAV, but intelligently repurposes hardware already on most UAVs: accelerometers, computers, and small microphones. Combined with the vehiclersquo;s motors and propellers, this is all the hardware necessary to create a vibration and acoustic stimulus and measurement device to conduct a pre-flight test and determine whether the airframe is safe to fly.nbsp; Each UAV motor, running in reverse, will cycle through its entire range of speeds. An accelerometer and microphone will collect data from each of the motors. Defects such as loose screws, frame cracks, and broken propellers will change how the airframe vibrates and the resulting sounds. Analyzing these samples to determine whether they come from a defect falls to a pre-trained machine learning classification network that runs on the on-board computer after data collection is complete.