Jialiang Wang,∗ Lidong Lu,∗ Penghao Wu,∗∗ Yiyang Chen,∗∗ Weidong Zhang,∗ and Hongtian Chen∗∗∗
Intelligent fault diagnosis, data-driven, multi-mode system, large- scale avionics systems, nonlinear model identification
In aviation engineering, it is crucial to ensure the safe operation and effective fault diagnosis of large-scale avionics systems of aircraft. Traditional model-based approaches face challenges, such as complex system modelling, inaccuracies, and oversimplified handling of nonlinearity. This paper introduces a data-driven approach to develop finite impulse response filters within TimesNet framework. By utilising TimesNet’s periodic sequence characteristics, this approach enables residual generation, offline system identification, and model learning to support online fault diagnosis during the cruise phase. The approach employs distributed segment data training to enhance transferability for multi-mode systems, achieving high modelling accuracy and robust nonlinear identification, which allows rapid and precise online fault diagnosis. Validation with real aircraft data from large-scale avionics systems demonstrated the approach’s accurate and efficient fault identification.
Important Links:
Go Back
