Masih Hosseinzadeh - Computer Aided Process Engineering (CAPE) Center, School of Chemical, Petroleum and Gas Engineering, Iran University of Science & Technology, Narmak, Tehran ۱۶۸۴۶۱۳۱۱۴, Iran
Norollah Kasiri - Computer Aided Process Engineering (CAPE) Center, School of Chemical, Petroleum and Gas Engineering, Iran University of Science & Technology, Narmak, Tehran ۱۶۸۴۶۱۳۱۱۴, Iran

Steel production processes are divided into two categories: direct reduction and blast furnace. The direct reduction process is carried out using a moving bed reactor called a shaft furnace. In this reactor, a non-catalytic gas-solid process is carried out and its output is sponge iron. In this study, by using the machine learning algorithm of a support vector machine, a model has been developed that can predict the amount of iron produced in sponge iron by using effective parameters without using mathematical modeling. Different algorithms and different hyperparameters were compared and the best prediction model was obtained.Different kernels of the support vector machine model including linear, polynomial, and RBF were compared. The results showed that the model was able to predict the conversion rate in the shaft furnace well. optimum model is determined with a polynomial sixth-degree kernel and epsilon ۰.۰۰۱. the best model has MSE ۵.۴۲۱۰-۶ and RMSE ۷.۳۶۱۰-۳ and R۲ test ۰.۹۹۹۹ which is a little more accuracy than previous models. The outputs of this simulation can be used to control the degree of metallization of sponge iron in shaft furnaces. As much as possible, the degree of metallization can be better controlled, and higher quality steel can be obtained.

SVM, non-catalytic reaction, shaft furnace, machine learning, sponge iron