350 rub
Journal Nonlinear World №8 for 2009 г.
Article in number:
The Experience of the Artificial Neural Network Application to Estimate the Aircraft Motion Math Model Actual Parameters
Authors:
A. Ya. Bondarets, O. D. Kreerenko
Abstract:
The application of artificial neural networks to solve an engineering problem of aircraft motion math model parameters identification is discussed in the paper. The experience of neural network application for determination of airplane-s wheels drag (braking, rolling resistance and contamination drag) actual coefficients on the runway covered with precipitations is introduced. The DCSL (Dynamic Cell Structure) neural network from "Adaptive Neural Network Library" is selected as the identification tool. The problem is solved using Matlab Simulink tool. The data accumulated from flight tests in real conditions were used to form samples for training of neural networks. The obtained by identification dependencies describing wheels resistance and braking performances on precipitation-covered runway have been compared with calculation methods given for estimation of such parameters in European certification requirements (NPA No. 14/2004), developed by JAA. The math modeling results have been tested for convergence with experimental data. Calculations and mathematical modeling are essential for the aircraft development and determination of its operating limitations, including estimation of aircraft behavior safety limits. However, application of computational methods requires compliance between the computation (math modeling) results and the experimental data, i.e. it is necessary to identify the math model parameters from the experimental data of real object behavior. The experience in development and application of the procedure for the flight dynamics math model parameters estimation according to flight tests data has shown that the most complicated element of practical identification tasks is the adjustment of identification results obtained from different samples of initial data. Artificial neural networks (ANN) allow determination of the required relations between input and output parameters of the object. Moreover, unlike the traditional identification methods, neural networks have a memory: it means that the results could be verified and accumulated during repeated "training" cycles (during the processing of new samples of initial data). Thus, neural networks allow getting the required relations for wide range of conditions at once and, as a result, to align random factors, which are unavoidable for experimental data. Neural networks were applied as a tool for identification of airplane-s wheels drag (braking, rolling resistance and contamination drag) actual coefficients on the runway covered with precipitations. Dependencies of these coefficients from the aircraft ground speed and the wheels inflation have been obtained during ANN "training" process. To solve this problem the simplified math model of aircraft motion along runway was developed. For identification purposes, math model is implemented in the form of longitudinal accelerations computation in the process of real aircraft run conditions reproduction. The ANN "training" depends on minimization of mismatch between empirical and calculated accelerations. The accomplished research shows the possibility of neural networks application for the aircraft math model parameters identification. It is shown that ability to deal with dependencies instead of separate values and to align random factors in experimental data sets are provided using ANN-based approach. Test results have shown the acceptable convergence of experimental and computational data.
Pages: 593-604
References
  1. Бондарец А.Я. Процедура уточнения математической модели динамики полета по результатам летных испытаний // Сб. докл. 3-й научн. конф. «Гидроавиасалон-2000». М. 2000.
  2. Бондарец А.Я., Оголев Ю.А. Результаты уточнения математической модели динамики полета самолета-амфибии Бе-200 по результатам летных испытаний // Сб. докл. 4-й научн. конф. «Гидроавиасалон-2002». М. 2002.
  3. Живов Ю.Г., Каргопольцев А.В., Фадеев А.С. Использование искусственной нейронной сети для определения аэродинамических характеристик самолета из летных испытаний // Техника воздушного флота. 2007.№ 3/4.
  4. Williams-Hayes P.S. Selected flight tests results for online learning neural network-based flight control system. NASA/ TM-2004-212857. 2004.
  5. Бондарец А.Я., Крееренко О.Д. Использование нейронной сети для идентификации параметров математической модели самолета по данным из летных экспериментов // Сб. докл. Междунар. научн.-технич. конф. «Новыерубежиавиационнойнауки» (ASTEC-07). М. 2007.
  6. JAA NPA. No 14/2004 on certification specifications for large aeroplanes (CS-25) Operation on Contaminated Runways
  7. Giampero Campa & Mario Luca Fravolini, «Adaptive Neural Network Library. Version 3.1 (Matlab R11.1 through R13). West Virginia University. Jul 2003.
  8. ТереховВ.А. Нейросетевыесистемыуправления. М.: Высшаяшкола 2002.
  9. Потапов А.А., Гуляев Ю.В., Никитов С.А., Пахомов А.А., Герман В.А. Новейшие методы обработки изображений / Под ред. А.А. Потапова. М.: ФИЗМАТЛИТ. 2008.