Modeling of the Blast Furnace Burden Distribution by Evolving Neural Networks

Abstract

A model of burden layer formation in the blast furnace is developed on the basis of layer thicknesses estimated from radar measurements of the burden (stock) level in the furnace. The dependence between the layer thickness and charging variables is modeled by neural networks. Parsimonious networks are determined by an evolutionary algorithm, which simultaneously trains weights and network connectivity. The efficiency of the training procedure is enhanced by tackling part of the numerical optimization by linear least squares. The resulting network models are utilized in a hybrid model, which considers practical constraints of the charging process in the furnace. The hybrid model is used to evaluate the impact of altered boundary conditions in novel charging programs.

A Data Mining Method Applied to a Metallurgical Process

“…3 with hidden nodes rearranged for the purpose of illustration: It has a sparse and an intuitively appealing connectivity, where three hidden nodes are devoted to the approximation of the square of X4 while a joint hidden node is used in the approximation of the product between X5 and X7. Such a sparse network lends itself perfectly to a deeper theoretical analysis of how it constructs its approximation and of the nonlinearity of the transformations [10]. For an evaluation of the pruning model on data with noise, the reader is referred to [ Schematic of the network with a lower-layer complexity of 8 (cf.…”

Evolutionary Data Driven Modeling and Multi Objective Optimization of Noisy Data Set in Blast Furnace Iron Making Process

“…have been very effectively applied till date to model and optimize the blast furnace processes. With the advent of genetic algorithms and particularly with the development of their relatively recent multiobjective adaptions, there are now several such applications in the area of ferrous and nonferrous production metallurgy . The basic algorithms are being continuously upgraded since then and are now capable of tackling problems which were unsolvable only recently.…”

“…With the advent of genetic algorithms [14] and particularly with the development of their relatively recent multiobjective adaptions, [15] there are now several such applications in the area of ferrous and nonferrous production metallurgy. [16][17][18] The basic algorithms are being continuously upgraded since then and are now capable of tackling problems which were unsolvable only recently. It was demonstrated by Vilfredo Pareto ) that if one tries to simultaneously optimize a set of mutually conflicting objectives, most often a set of solutions become optimum, instead of an unique and single optimum, that is, more common in single objective optimization.…”

Modelling Noisy Blast Furnace Data using Genetic Algorithms and Neural Networks

“…For certain objectives the data driven models can be a more appropriate choice, and their efficacy, particularly when applied through Artificial Neural Networks (ANN), in predicting various trends in the blast furnace process is already documented [4][5][6]. In recent times the biologically inspired Genetic Algorithms (GAs) have found significant applications in the field of ferrous production metallurgy [7][8][9][10][11][12][13][14], and a combination of GAs and ANN has already been elaborated on a problem related to blast furnace operations [15,16]. The complex relationships between the different variables of the blast furnace operation are ideally suited for being tackled by a data driven analysis based on a GAs-ANN model.…”

“…In this study we will demonstrate the application of a genetic algorithms based neural network to a set of noisy blast furnace data where the above problems are considered. The methodology adopted here differs from the earlier approaches [15,16], and will be elaborated as needed.…”