On-board sensor data monitoring system for unmanned aerial vehicle PHM

Due to the excellent performance and cost-effective, unmanned aerial vehicle (UAV) has been widely used in civil and military fields. But the accident rate of UAV is much higher than that of manned aircraft. Therefore, the sensor data monitoring of UAV has become a research hotspot, which can further support UAV Prognostics and Health Management (PHM). However, the on-board computing resources and power are limited, and most state-of-the-art sensor data monitoring methods can only be operated on ground. A huge challenge is presented to UAV real-time condition monitoring. In this paper, an on-board system is developed for real-time fixed-wing UAV sensor monitoring. Firstly, an LSTM network is designed to fulfill accurate estimation of UAV sensor data. Secondly, the sensor data estimation model with high computational complexity is accelerated by utilizing High Level Synthesis (HLS). Finally, the calculation optimized model is deployed in an on-board embedded hardware platform. The simulated fixed-wing UAV flight data are used to verify the performance of the proposed system. The experimental results show that the proposed system is effective for fixed-wing UAV real-time sensor data estimation.