Fault Classification in Aircraft Electrical Systems Using Transients and FPGA-Based Deep Learning

The shift of aviation industry towards More Electric Aircraft (MEA) has led to increasingly complex aircraft power systems characterized by highly inter-connected electrical and electronic components. Electri-cal monitoring systems are indispensable equipment to ensure flight safety and reliability in aircraft environments. Field-Programmable Gate Arrays (FPGAs) offer compelling solutions for the implementation of these systems in aircraft due to their signal processing-based capabilities, reduced power consumption and compact size suitable for integration into avionics. Additionally, FPGAs can process large amounts of data in real-time at high speeds. This paper explores the FPGA implementation of a deep learning model to detect and classify faults in aircraft power systems using transients that are introduced by electrical faults such as short and open circuits. The proposed approach employs an LSTM neural network with a classification accuracy of 94.09% deployed on a Xilinx Zynq UltraScale+ MPSoC ZCU102 development board. Experimental results demonstrate the effectiveness of the FPGA implementation in classifying electrical faults based on high-frequency transients.