University of Twente Student Theses


Realtime and Onboard fault diagnosis of UAV motors using RNN prediction model

Boe, M. (2022) Realtime and Onboard fault diagnosis of UAV motors using RNN prediction model.

[img] PDF
Abstract:In recent years public interest in UAVs (unmanned aerial vehicles) has increased. Existing solutions are used to reduce risks and detect critical failures on system level or single output sensors. This research presents a solution for fault diagnosis of UAV motors using RNN prediction models. The work presented distinguishes itself from standard and existing solutions by not only performing fault diagnosis on a component level rather than using system models or analytical redundancies but also presenting on board, in real-time, physical experiments. The fault diagnosis accuracy is evaluated for several types of input functions including hand-flown flights and the fault diagnosis timing aspects are evaluated using the onboard processor. Results show on a static test setup, input data with high dynamic nature impose problems for the developed prediction model. The physical hand-flown experiments show that when the input data is generated by the PX4 flight controller, 100% detection rate is achieved. The developed prediction model runs at 3322Hz on the selected RPI4B and theoretical approaches are presented to calculate the response time of the system and calculate a theoretically maximum network size based on measurements.
Item Type:Essay (Master)
Saxion, Enschede, Netherlands
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:31 mathematics, 50 technical science in general, 54 computer science
Programme:Embedded Systems MSc (60331)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page