Fault Diagnosis Of Aircraft Power System Based On Deep Learning

The engine serves as a source of power for the aircraft and plays a vital role.The normal operation of aircraft engine and drone motor system is the guarantee of safe operation of the aircraft,and it is of great significance for timely fault diagnosis.Due to the complex structure,nonlinear system model,and external noise interference of UAVs and aero engines,it is difficult to construct high-precision models.In this paper,the theory of deep learning is applied.Based on data-based methods and the advantages of dealing with complex nonlinear problems by deep learning,the method of fault diagnosis of aero engine gas path components and drone actuators is proposed.The research of this paper mainly includes the following points:(1)Build an experimental simulation and data acquisition system for six-rotor UAV and aero engine.According to the data characteristics of UAV and aero engine,preprocessing and feature extraction are adopted.The faults of UAV and aero engine are classified and coded,which lays a good foundation for the training of deep neural network.(2)Based on the theory of autoencoder network in deep learning,a fault diagnosis network model for drone actuators based on deep learning is designed.By using denoising autoencoder network,the robustness of the drone fault diagnosis network model is improved.By comparative experiments,the influence of the number of AE layers and sample length on the performance of the model are analyzed,and the network structure is optimized to improve the accuracy.(3)In this paper,the gas path characteristics and fault types of aero engine are analyzed,and the network model for fault diagnosis of aero engine based on deep learning is designed,and the whole process of fault diagnosis is introduced in detail.The influence of the number of network layers on the performance of the model is analyzed by contrastive experiments,which makes fault diagnosis model achieves a higher accuracy.(4)By the fault diagnosis experiments of UAV and aero engine,it shows the feasibility and practicability of deep learning in fault diagnosis of aero engine and UAV.And by the effective methods of preprocessing and feature extraction,the fault diagnosis performance can reach a high level,which provides an important reference value for the fault diagnosis of actual UAVs and aero engines.