<p>In rail transit systems, rail wear shortens service life, making it crucial to pre-strengthen critical weak regions. Friction stir processing (FSP), as a solid-state strengthening method with low heat input, can effectively refine grains and improve the hardness and strength. However, as a high-carbon steel, the microstructure evolution and performance of the rail steel are highly sensitive to thermal-mechanical behavior. Therefore, establishing a numerical model that can quantitatively analyze the heat and mass transfer characteristics is of great significance. In this study, a constitutive model optimized by a differential evolution algorithm is proposed based on the true stress-true strain values obtained from hot compression. Furthermore, the numerical model of U75V rail steel FSP is established and validated. The differences between conical and round pins in heat generation and material flow are quantitatively analyzed. When the conical pin is used, the interfacial heat generation and viscous dissipation heat generation in the shear layer are more intense, resulting in higher temperature and material flow velocity. Specifically, on a circle located 0.5 mm below the shoulder with a radius of 5 mm, the processing temperature is approximately 30 K higher, and the material flow velocity is about twice that achieved with the round pin. The conical pin exhibits a stronger ability to promote vertical material migration, while the round pin facilitates horizontal migration. Based on this work, theoretical basis is provided for the process control in U75V rail steel FSP.</p>