Модель пластического состояния материала для широкого диапазона температур и скоростей деформаций на основе нейронной сети

Abstract

An approach to model development of a material based on artificial neural network describing plastic flow was considered. Model’s feature is possibility to model flow behaviorof material in wide range of strain, strain rate (0,001…105s−1) and temperature (20-1000 °С). For model implementation the neural network based on the perceptron withone hidden layer has been used. On an example of the experimental family of strain curves for AISI 45 the benchmark analysis of training for various variants of the learning data preparation was carried out.