Neural Network Modeling of Coefficient of Burden Resistance to the Gas Movement in the Lower Part of the Blast Furnace in Conditions of Operation with Coke Nut

Abstract

A mathematical model based on the use of artificial neural networks for forecast of resistance coefficient of burden to the gas at the bottom of the blast furnace with using of coke nut by processing of data array for the OJSC "MMK" blast furnaces (capacity of 1370 m3), equipped with a chute-type bell-less charging device has been created. This test has shown the adequacy of the model to real data. Influence of such factors as characteristics of blast (oxygen content, temperature, natural gas and water steam consumption), iron ore (raw material consumption per time unit, FeO, MgO, Al2O3 content, fraction, basicity), coke (wearability (M10), impact strength (M25), coke strength reactivity (CSR), coke reactivity index (CRI)) on gas dynamics variation at the lower part of the black furnace have been determined. Average relative prediction error does not exceed 0.28 %, the maximum of the sample is 2.82 %. The oxygen content in the blast has the biggest effect on the burden resistance coefficient. When oxygen concentration is more than 25.2 %, the increase of natural gas consumption improves gas-dynamic conditions in the lower part of blast furnace. With the decrease of oxygen content in the blast, the influence of natural gas consumption on coefficient of burden resistance varies in the opposite direction. The reduction of coke wearability (M10) by 0.05 % abs. or the increase of coke strength reactivity (CSR) by 0.14 % abs. has compensated negative effect of coke nut (consumption 4 kg/t of iron) on blast furnace operation.