The invention relates to the field of FPGA hardware accelerator and
aviation equipment fault diagnosis cross technology, in particular to an aero-engine fault
sound recognition system based on an FPGA. The
system comprises a hardware platform and a function module, the hardware platform is composed of a high-disturbance-rejection sound sensor, a low-power-consumption single-
chip microcomputer, a high-performance
FPGA chip, a high-speed FLASH
chip and an
airborne early warning display terminal, and
data transmission and interaction are achieved through cooperation of multiple buses; the function module comprises a sound acquisition module, a
feature extraction module, a fusion convolutional
spiking neural network module and a fault early warning module, and the ICSNN module adopts a CNN and SNN parallel architecture and realizes complementary extraction of time-frequency features and time-sequence features in combination with a sparse
convolution technology and an LIF
neuron model. The method meets the requirements of low
power consumption and
low delay of airborne equipment, is high in
fault recognition rate and low in omission ratio, and is suitable for real-time monitoring and early warning of hidden faults of aero-engines in high-temperature and high-vibration environments.