Design of chromium-vanadium steel based on design loop combining property-oriented design criteria and machine learning prediction model

Abstract

Due to the lack of design guidance for high-performance Cr–V steel, it is challenging to identify the target steel within a vast design space. In this paper, a design loop combining property-oriented design criteria and machine learning prediction model is proposed. Based on the design criteria, the alloying values of the initial sample and the allowable variation range are given to narrow the search space. Then, the machine learning model is employed to predict properties across all design nodes within the design space. The results suggest that the samples within the design space, meeting the proposed design criteria, demonstrate higher hardness. However, wear resistance decreases rapidly as Cr increases and V/Cr decreases. The predicted optimal sample C2.1V3Cr4Mo2.5 is prepared and its properties are measured, showing that the absolute errors between the measured and predicted values are less than 5%. In addition, C2.1V3Cr4Mo2.5 is more cost-effective than Cr26 and V10. This design loop facilitates the rapid selection of the most promising steel for experimentation and provides a new avenue for the development of wear-resistant alloy steels.

Keywords

• Cr–V steel
• Design loop
• Machining learning
• Search space
• Optimal composition combination