Application of artificial neural network for determination of the additives amount in the automated process control system of steelmaking in basic oxygen furnace

Abstract

This paper describes an algorithm of determining the amount of deoxidizing and alloying materials that are loaded into the basic oxygen furnace (BOF) and steel ladle on the base of information about burdening of melting and chemical composition of the steel using artificial neural network (ANN). The analysis of resent researches and publications regarding mathematical modeling of BOF melting and application of ANN as such models was made. This analysis show that selected topic has novelty and relevance. The schematic of interaction of different kinds of mathematical models in the system of automated control of BOF melting is offered. The research of applicability of artificial neural networks for determination of quantity of deoxidizing and alloying components is performed. The place of the obtained artificial neural network in the overall system of automated control of basic oxygen melting is described. The description of the multistep selection process of the ANN architecture is given. The correlation coefficients and mean square deviations for all parameters are found. The results of performed analysis are considered satisfactory. The recommendations for replacement of alloying and deoxidizing components in the absence of any of them in stock are given.

В статье приводится алгоритм определения количества раскисляющих и легирующих материалов, загружаемых в кислородно-конвертерную печь и стальковш, на основании информации о шихтовке плавки и химическом составе стали с использованием искусственной нейронной сети (ИНС).

У статті наводиться алгоритм визначення кількості розкислюючих та легуючих матеріалів, що загружаються до киснево-конвертерної печі та стальковша, на основі інформації про шихтовку плавки та хімічний склад сталі за допомогою штучної нейронної мережі (ШНМ).