Aerospace Instrument-Making Annotation << Back PREDICTION OF FUEL CONSUMPTION BY MI-8 HELICOPTER WITH USE OF MODULAR NEURAL NETWORK A.V. Grabik, A.P. Karpenko, A.V. Nekrasov, A.G. Trofimov, A.S. Chernyshev The main result of the project is means evaluation of prediction of fuel consumption by Mi-8 helicopter with the use of modular neural networks. Its classifier is built on Kohonen network's basis, and predictive artifitial neural networks is built on NARX networks. The result of this work extend and specify the results of previous projects done by the authors. We set the aim of neural network prediction of fuel consump-tion, proposed a solution based on modular ANN. We introduced architectures of neural networks, corresponding software and the results of numerical experiment on efficiency of methods, algorithms and software decisions for short-term prediction. Based on the results it is possible, firstly, to deduce the prospective of future research of modular ANN applications for predicting the fuel consumption by the aircraft. Secondly, the results of this project helps to form recommendations on the choice of the quantity of neurons used by ANN and methods of their learning. In the future the authors plan to investigate efficiency of long-term prediction by various methods of adaptive approximation of the data. Keywords: neural network, prediction of fuel consumption, Kohonen network, modular neural network. Contacts: E-mail: gralexey@mail.ru, E-mail: apkarpenko@mail.ru, E-mail: al_nekrasov@mail.ru, E-mail: atrofimov@list.ru, E-mail: alexey.chernushev@gmail.com Pp. 28-38.