Development of a Material Selection Model for Rolls Rolling Mills Using Neural Networks

The mathematical model of optimization of selection materials for rolls rolling mills based on neural network modeling was developed. The analysis of neural network models to determine the most appropriate architecture and other parameters for solving the problem was performed. Specialized software for designing neural network models based on the Keras deep learning library had been created. A neural network with a multilayer perceptron (MLP) structure was developed, the adequacy of the resulting neural network was checked. The resulting neural network is able to predict the amount of roll wear from the specified parameters. A special algorithm for the selection of the roll material which under the specified conditions would have acceptable wear at minimal cost of repair rolls was developed.