Predicting the Product Classification of Hot Rolled Steel Sheets Using Machine Learning Algorithms

Abstract

The mechanical properties of the SAPH440 hot rolled steel sheet are mainly controlled to satisfy product specifications. Three mechanical properties including the yield strength, ultimate tensile strength, and elongation are measured and utilized in product classification. Based on these properties, the steel is classified into 3 grades: Class 1 (meets specification), Class 2 (moderate quality), and Class 3 (low). However, various factors can affect the mechanical properties, leading to a long setup time for initial production runs. Therefore, this paper aims to improve the accuracy of these predictions by using machine learning algorithms. The results of experiments showed that the random forest algorithm had the best performance, with an accuracy of 70.0% and a macro average F-1 score of 70.0%. This more accurate prediction can reduce the initial setup time and save 37,000 USD per grade in trial run costs.