Author:
Soria Gomez Maria, ,Koschlik Ann-Kathrin,Arts Emy,Raddatz Florian,Wende Gerko, , , ,
Abstract
Unmanned Aerial System (UAS) activities have increased steeply in the last years, market research forecasts a continuous increase in the near future. The rapid growth of this industry, however, has outpaced the development of rules and systems to govern their use, as well as those to ensure a safe operation before, during and after flight. Maintenance, Repair and Overhaul (MRO) aspects will gain relevance as more and more UAS take to the sky. Rotary-wing UAS have 2 or more propellers, which are easily damaged during normal operation of the vehicle. The reduced thrust and increased vibration imply losing performance and setting the UAS structure under stress. With the propellers being the main source of sound of the propulsion system, we propose the use of acoustics to identify damaged propellers. Microphones placed off-board do neither disrupt UAS operation nor reduce the payload capacity. Furthermore, this method does not depend on a particular manufacturer or software. In this paper, we present a concept for the non-destructive testing of multi-copter propellers. The fault diagnosis aims at recognizing the difference in sound between damaged and undamaged propellers. This evaluation takes place before the UAS takes off the ground and after it lands, thus allowing to interrupt a possible dangerous mission or identifying damage occurred during operation. The vehicle is on the ground in an “idle state” where the propellers already spin, but not fast enough to lift it. This state is used for a first analysis of the sound of a single propeller and several propellers, as well as for the generation of data. Next, two approaches for the detection of damage are developed and their performance is evaluated: an analytical approach and a machine learning algorithm based on an autoencoder neural network....
Cited by
2 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献