Development of Empirical Relationships Between Input and Output Parameters of Sponge Iron Process Using Neural Network

Abstract

In the present study, estimation of optimum input parameters corresponding to desired output parameters is carried out for a typical sponge iron process. For this purpose data of rotary kiln of the process are collected. It includes temperatures profiles inside the rotary kiln and flow rates of air, iron ore, feed coal, slinger coal and sponge iron. These data are analysed using ANN and group method of data handling (GMDH) approaches. Sixteen ANN topologies are proposed where TOP-3 is found best, which has mean absolute error (MAE) as 8.42. However, GMDH analysed the data with MAE of 8.72. Optimum values of operating parameters are found through ANN and GMDH approach and compared with that are used in the existing process. It shows that values of input parameters found through ANN can be selected as optimum based on minimum operating cost.