Multiobjective optimization of charging programs for optimal gas flow conditions in the blast furnace

O2 Other


Internal Authors/Editors


Publication Details

List of Authors: Tamoghna Mitra, Frank Pettersson, Henrik Saxén, Nirupam Chakraborti
Publication year: 2015
Book title: KomPlasTech, XXII Conference Computer Methods in Materials Technology, January 4-11. 2015, Krynica-Zdrój, Poland


Abstract

The burden distribution is an essential factor which helps the operator to achieve favorable operating conditions in a blast furnace. For manipulating the burden distribution the operator has to influence the radial distribution of porous coke and comparatively less porous iron-ore agglomerate (pellets). This in turn influences the gas flow conditions in the furnace. Furthermore, the distribution of the layers determines the shape and position of the cohesive zone in a blast furnace. The operator controls the burden distribution by selecting an appropriate charging program, which lists the order, position and amount of raw material (ore or coke) that is charged in the furnace. The selection of an appropriate charging program for a particular operating situation involves billions of degrees of freedom, only few of which are feasible. Therefore a method is necessary to model the furnace conditions and evaluate different charging programs.

Large numbers of mathematical models have been developed to simulate the burden distribution and the gas flow in the blast furnace, ranging from simple zero-dimensional models to very detailed CFD-DEM models. In this study, the simplified burden distribution model has been chosen which is based on falling trajectories, simple distribution logics and volume balances for the charged materials. For the gas flow simulation another simplified model has been used which makes some convenient assumptions to avoid detailed CFD analysis.

In this paper various charging programs are evaluated using the model to achieve a required temperature and gas distribution in blast furnace. As exhaustive search is impossible, therefore, genetic algorithm is utilized to achieve the goal. A multi-objective 1-k-dominance predator prey method is used as the number of objectives is large. This allows relaxation of the strict conditions of Pareto optimality at users’ discretion and allows computation of problems of large dimensions in the objective space, which would be otherwise impossible in an evolutionary way. Various objectives include the temperature at different positions on the in-burden probe, pressure drop across the packed bed, optimum overall ore-to-coke- ratio, etc. and the results are critically analyzed.


Last updated on 2019-15-11 at 03:03

Share link