Judging the States of Blast Furnace by ART2 Neural Network

摘要

An improved unsupervised neural network of ART2 is proposed to judge the pattern of blast furnace states. In this method six variables viz. charging speed, air flow, air temperature, air pressure, permeability indices and Si composition of liquid iron are determined to express the blast furnace states in a smelting process. The values of these variables are gained from the slide windows in order to overcome their time-varying difficulty. The pattern of blast furnace states is classified by ART2's competition learning and self steady mechanics. The simulation shows this method is effective.