V.P. Dobritsa - Dr.Sc. (Phys.-Math.), Professor, Department of Information Systems and Technologies, Southwest State University (Kursk) E-mail: dobritsa@mail.ru E.A. Kuleshova - Lecturer, Department of Information Systems and Technologies, Southwest State University (Kursk) E-mail: lena.kuleshova.94@mail.ru

In recent decades, widespread neural network technology in various fields - pattern recognition, time series prediction, di-agnosis of diseases, information security and others. The article describes the use of neural networks to identify the properties of materials such as alloys, composite ma-terials. Artificial neural networks are a set of connected and interacting with each other formal neurons. Despite the simplicity of individual neurons, being connected by a fairly large and complex network with a large number of connections and manages the interaction of such «locally primitive processors» together we can solve quite complex problems. Experimentally establish compliance with the selected values of input parameters corresponding to the output values of the parameters characterizing the properties of the alloy. After a predetermined number of experiments, we get a series of training. The more training series, the more reliable is the forecast of the alloy properties in selected parts of its components. Thus, with the help of a trained neural network can be without undue experimentation to select the composition of the alloy with the desired properties. The proposed approach makes it possible to predict the properties of the material using an appropriate neural network in other cases. For example, for composite materials. The input parameters in this case may be: the materials constituting the layers, the thickness of these layers, the composition of the adhesive material, etc.

 

1. KHejjkin S.Nejjronnye seti. Polnyjj kurs. Id-e 2. M. SPb. Kiev: ID «Viljams». 2006.
2. Golovko V.A.Nejjronnye seti: obuchenie, organizacija i primenenie. M.: Radiotekhnika. 2001.
3. Kolmogorov A.N. Predstavlenie nepreryvnykh funkcijj mnogikh peremennykh superpoziciejj funkcijj odnojj peremennojj i slozheniem // DAN SSSR. 1958. S. 953-956.
4. Eremin A.V., Dobrica V.P.Algoritm obuchenija mnogoslojjnykh nejjronnykh setejj s neskolkimi vykhodnymi nejjronnymi na osnove minimizacii oshibok vykhodnykh vektorov // Nejjrokompjutery: razrabotka, primenenie. 2011. №3. S. 24-28.