Chemistry-driven Predictive Modelling of Mechanical Performance in 42CrMo4 (AISI 4140) Steel with Machine Learning
Sustainable Manufacturing and Foundry Practices · 21.05.2026 · 1 (1)
DOI: 10.1177/iif.261438647 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sustainable Manufacturing and Foundry Practices, online erschienen am 21.05.2026. Verfasst von Subhasis Das Gupta und 2 weiteren Autor:innen.
Abstract (Originalsprache)
The 42CrMo4 low-alloy steel shafts, gears, and other high-strength parts used in the automotive industry offer a strong balance of toughness, strength, and surface quality. Traditionally, alloy development has relied on empirical knowledge and repeated experimental trials to adjust chemical composition and achieve targeted mechanical properties. In this study, we developed a machine learning model to predict the mechanical properties of 42CrMo4 steel from its chemical composition. Our dataset comprised 1,000 heat measurements listed as weight percentages, along with tensile strength, yield strength, proof stress at 0.2%, and surface roughness (Ra) from machining. After removing anomalous data with extreme or inconsistent compositional values, nine key alloying elements, Ti, Ni, Cr, Mo, Cu, Mn, P, Si, and C, were selected as input parameters for the machine learning model. Different machine learning models were trained separately for each identified property, ensuring tailored predictions. We tune their hyperparameters using a five-fold cross-validation grid search, helping us find the best settings. The Gradient Boosting Regressor algorithm performs well, showcasing its effectiveness and reliability, with R2 values ranging from 0.64 to 0.73 across the four targets, indicating reliable predictions. Additionally, error metrics such as root mean squared error and mean absolute error indicate that proposed predictions remain within practical engineering limits, providing confidence in their accuracy. To improve the model, we employed important grouped features and correlation heatmaps for our analyses. These show that Ti, Ni, Cr, and Mo contribute most to strength properties, while Mn, P, and Si have a greater impact on Ra. Further, the results are comparable to the metallurgical information on solid-solution strengthening, carbide formation, and surface finish sensitivity. The suggested framework demonstrates that composition-based machine learning models can support alloy design, reduce experimental trials, and provide a digital tool for predicting the mechanical properties of 42CrMo4 steel.
Zitieren
APA
Gupta, S. D., Anand, A., & Krishna, R. (2026). Chemistry-driven Predictive Modelling of Mechanical Performance in 42CrMo4 (AISI 4140) Steel with Machine Learning. Sustainable Manufacturing and Foundry Practices, 1(1). https://doi.org/10.1177/iif.261438647
Vancouver
Gupta SD, Anand A, Krishna R. Chemistry-driven Predictive Modelling of Mechanical Performance in 42CrMo4 (AISI 4140) Steel with Machine Learning. Sustainable Manufacturing and Foundry Practices. 2026;1(1). doi:10.1177/iif.261438647
BibTeX
@article{Gupta2026,
author = {Gupta, Subhasis Das and Anand, Abhinav and Krishna, Ram},
title = {Chemistry-driven Predictive Modelling of Mechanical Performance in 42CrMo4 (AISI 4140) Steel with Machine Learning},
journal = {Sustainable Manufacturing and Foundry Practices},
year = {2026},
volume = {1},
number = {1},
doi = {10.1177/iif.261438647}
}
EUVIMED