| The flight control system is one of the most important systems for civil airliners.Its high reliability is an important guarantee for the safe flight of the aircraft.Therefore,it is of great significance for the flight control system to carry out condition monitoring and intelligent fault diagnosis.This paper uses flight control data and intelligent learning technology to focus on the indepth study of the flight control system’s operating parameter characterization faults.The flight control elevator system is used as an example to establish a flight control elevator system fault diagnosis model.First,the composition and principle of the flight control system of the B777 aircraft were analyzed,the maintenance report of the flight control system was studied,and typical common faults were selected as the research objects.In the AMESim simulation environment,a simplified physical model of the flight control elevator system was constructed,the typical failures were simulated and analyzed,and the system’s failure data was collected.After a comparison and analysis of the parameters of the flight control system before and after the failure,the 51 parameter indicators that best characterize the system failure were extracted.Secondly,in order to study the problem of noises in the fault data collected by the aircraft flying in the real environment,the noisy fault data was studied.A generalized singular value decomposition algorithm(K-SVD)and orthogonal matching tracking algorithm were proposed.(OMP)combined fault data denoising method.The method first trains the initialization dictionary into an over-complete dictionary that can effectively characterize the original signal,and then uses the OMP algorithm to find the dictionary atoms that best represent the original signal from the over-complete dictionary,and combines them linearly to obtain the denoised signal of.In view of the high dimension of the collected fault sample data,it is not easy to establish a fault diagnosis model,etc.,and the principal component analysis method(PCA)is introduced to extract the features of the fault data.Considering the shortcomings of the PCA algorithm in processing non-linear signals,a kernel method was proposed to improve it.The effectiveness of PCA and the improved PCA method with different kernel functions in data compression and feature extraction was studied.Finally,a fault diagnosis model is established.The applicability of the extreme learning machine(ELM)algorithm in fault diagnosis is analyzed.Aiming at the disadvantages that the input weight matrix and hidden layer bias of the extreme learning machine are randomly generated,it is easy to cause network instability and large deviations.The optimization algorithm(DE)obtains the optimal input weight matrix and hidden layer bias of the extreme learning machine when the training error is minimum,and improves the extreme learning machine.At the same time,for the shortcomings of the differential evolution optimization algorithm that is easy to premature and insufficient accuracy,the adaptive evolution method is used to improve the differential evolution algorithm.The improved differential evolution algorithm not only has better global optimization capabilities,but also the optimization accuracy is improved..The improved differential evolution algorithm is applied to the optimization of the optimal input weight matrix and hidden layer bias of the extreme learning machine,and the simulation verification is performed.Simulation results show that the improved limit learning machine is applied to the fault diagnosis of the elevator control subsystem of the flight control system,and the fault diagnosis accuracy rate is 89.6%,which is 11.2% higher than that before the improvement,and the stability of the improved model is also obtained.Improve,technical methods are feasible and effective. |
