Machine learning approaches for predicting ultimate tensile strength in 9% Cr steels

Abstract. This paper applies machine learning techniques to analyze a dataset featuring the initial elemental composition and mechanical properties of various 9% chromium steels. Mechanical properties in the dataset are yield stress and tensile strength, elongation, and reduction in area. The primary objective is to predict the ultimate tensile strength of these steel alloys using different machine learning algorithms. The study involves preparing the dataset to make it suitable for machine learning algorithms, selecting relevant features, and implementing three different machine learning models to evaluate their performance. The performance of these models, namely linear regression, random forest, and support vector machines model, were compared to determine the most accurate and reliable method for predicting the ultimate tensile strength. Spearman’s rank correlation coefficient was performed to understand the influence of elemental composition on the ultimate tensile strength. This study was conducted using Python version 3.12.2. The objective is to demonstrate the potential of machine learning in decision-making in material science.