350 rub
Journal Information-measuring and Control Systems №6 for 2010 г.
Article in number:
Neuronet estimation algorithm of store separation interference-s characteristics
Keywords:
interference characteristic
neuronet algorithms for estimation
flight experiment results
areodynamic by flight
Authors:
S. V. Alexeev, N. A. Baranov
Abstract:
Procedure of construction of the guaranteed set of estimations of characteristics of an aerodynamic interference by results of supervision of a vector of phase coordinates of a separated pendant cargo is offered. For the decision of a problem of definition of factors of approximation of revolting influence (characteristics of an aerodynamic interference) it is offered to use neuronet al-gorithms.
One of problems of creation of models of objects, absence of authentic data on its characteristics is. At modelling dynamics of branch of pendant cargoes sharply there is a problem of an estimation aerodynamic interference characteristics of a pendant cargo. Aerodynamic interference characteristics represent functional dependences of factors of additional aerodynamic forces and the moments on a relative arrangement of the plane-carrier and a pendant cargo, and also conditions of flight of the carrier. Presence interference forces and the moments essentially changes character of aerodynamic forces and the moments.
Experience of creation of models of branch of pendant cargoes shows, that use in mathematical modelling the aerodynamic inter-ference characteristics received settlement by or on the basis of purges in wind tunnels, leads to unpredictable results which are not observed in flying experiment.
Let's consider a problem of reception of the guaranteed estimations of aerodynamic characteristics of a separated pendant cargo by results of experiment.
Dynamics of a pendant cargo is described by the vector differential equation which use demand the essential computing expenses caused by necessity of performance of iterative cycles for definition of factors of approximation on each interval of time. Perfor-mance of each the iteration is connected with necessity of integration of the equations of dynamics of system that is wide recource procedure, for example, for a problem of definition of characteristics of an aerodynamic interference of a separated rocket. The control system on a site of branch works in a mode of indemnification of starting indignations.
It is possible to provide decrease in computing expenses, if for definition of factors of approximation of revolting influence (characteristics of an aerodynamic interference) to use neuronet technologies.
It is necessary to construct a neural network which input is the set of vectors of supervision of a vector of phase coordinates of system during the moments of time, and an output - set of vectors of factors of approximation of function of revolting influence.
Pages: 53-57
References
- Алексеев С. В., Баранов Н. А., Васильев И. В.Применение нейронных сетей для оценки характеристик аэродинамической интерференции по результатам летных испытаний // Проблемы совершенствования робототехнических и интеллектуальных систем летательных аппаратов. М.: МАИ. 2005. С. 183-186.
- Алексеев С. В., Васильев И. В. Построение гарантированных оценок интерференционных аэродинамических характеристик отделяемых грузов // Передача обработка и отображение информации. Ставрополь: Ставропольское ВВАИУ. 2006. С. 5-6.
- Горбань А. Н.Обобщенная аппроксимационная теорема и вычислительные возможности нейронных сетей // Сибирский журнал вычислительной математики. 1998. Т. 1. № 7. С. 12-24.