Prediction of end point %C of CONARC® furnace using machine learning methods.

Autor: PARUL, KOTA, SAMIRAJ, ALBIN ROZARIO, HAZRA, SUJOY S
Předmět:
Zdroj: Sādhanā: Academy Proceedings in Engineering Sciences; Sep2023, Vol. 48 Issue 3, p1-13, 13p
Abstrakt: CONARC® steel making process is a combination of convertor steel making and electric arc steelmaking to get the benefit of both the process and make it flexible in terms of using raw material feed mix. Raw material feed mix in this furnace on an average is 60% hot metal (HM), 38% cold direct reduced iron (CDRI) and 2% steel scrap. In this furnace operation there are two phases, namely, the oxygen blowing phase and arcing phase followed by tapping of steel into the ladle. During the oxygen blowing phase, the HM carbon content is reduced from 4.5% to 0.3%–0.5%, and further reduced to 0.025–0.03% in the arcing phase depending upon the grade of steel produced. During the arcing stage, CoJet™ lances are used for the oxidation of the bath and reduction of the carbon content to the desired values. The end point %C parameter is very important in CONARC® steel making as it determines the productivity and quality of the steel produced. Based on the analysis, mathematical and machine learning approach was adopted to predict the end point %C during the arcing stage of the furnace. The algorithms which are used and compared are the tree-based models and support vector machines. After comparing the results, the tree based model seems best fit after further optimization to get an accuracy of 83%. The model was validated with plant trials and the accuracy was found to be within ±0.013 %C. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz:
Nepřihlášeným uživatelům se plný text nezobrazuje