Machine Learning‐Guided Design of Pearlitic Steel with Promising Mechanical and Tribological Properties

Abstract

Herein, attempts have been made to design and develop pearlitic steels for application in heavy‐haul rails. The hardness plays a vital role in studying the mechanical and tribological properties, which is theoretically related to the alloying composition of steel. With aid of machine learning (ML) method, the particle swarm optimization (PSO) improved generalized regression neural network (GRNN) is utilized to model the relationship between composition and hardness of pearlitic steel. The results show that the designed steel exhibits superior hardness and mechanical properties with fine pearlite lamellar microstructure. In addition, the wear behavior of the steel and its wear mechanism are systematically studied by tribological testing and electron probe microanalysis (EPMA) observations of worn surface and wear particles. With composition optimization, the wear resistance has further improved as evidenced by the lower friction coefficient and reduction of wear volume. The pearlitic steels exhibit a combined wear mechanism including adhesive wear, abrasion, delamination, and plastic deformation. As a result, the designed steels offer high hardness with very good mechanical and tribological properties which are far superior to previously reported pearlitic steels. This work may assist in developing the appropriate composition to create the desired hardness, mechanical, and tribological properties.