Temperature Distribution Control in Blast Furnace by RNN

In manufacturing industries, reactor furnaces such as blast furnaces have been playing vital roles. Based on the improvements of furnace facilities, the size of a furnace has been enlarged remarkably. Recently, various studies have been undertaken to examine the inner furnace phenomena of a blast furnace. These studies are limited, however, to the evaluation of furnace performance while a design method of furnace control is left be developed. In this paper, a method is proposed to estimate and control the temperature distribution in a reactor furnace by recurrent neural network (RNN) model for shorten the transient time. For the purpose, furnace data only near a furnace wall are used to control inner furnace temperature distribution to its targeted one. Initially, a simplified furnace simulation program for the calculation of inner furnace gas flow, pressure and temperature distribution is developed based on the precise furnace model studied before. Then, the simulator is used to estimate and control the inner furnace temperature distribution. In the estimation, boundary data such as temperatures and pressures, measured near a furnace wall, are used in the furnace simulation. Further, for the control of inner furnace temperature distribution, necessary values for gas blowing at the bottom of the furnace and the burden supply at the top of the furnace are determined. Both for the estimation and the control of the inner furnace temperature distribution, a RNN model for nonlinear dynamic temperature distribution control is proposed.