Mechanical Performance and Microstructure Prediction of Hypereutectoid Rail Steels Based on BP Neural Networks
Yong Deng, Ling Qiao, Jingchuan Zhu, Bin Yang
Abstract
Rapid development of railway has heightened the need for the researches on hypereutectoid heavy rail steels. Artificial intelligence method has become an effective tool to realize materials composition design. In this paper, BP neural network models are constructed to determine the relationship among (Cr, P, S, V) alloying elements, mechanical performance and microstructure of hypereutectoid rail steels. Analysis based on this model reveals that Cr is the most prominent element for mechanical properties. The tensile strength, yield strength and hardness can be improved with the increasing content of Cr and V. The addition of P and S seems to decrease the strength and hardness of rail steels. Furthermore, the addition of (Cr, P, S, V) has a slight impact on the content of pearlite dual phases. The increase of (Cr, V) and decrease of (P, S) can contribute to an increase in ferrite content with the associated decrease in cementite. Experimental results agree well with the prediction based on the BP neural network model. This work provides an excellent basis for assessing the mechanical performance and microstructure of hypereutectoid heavy rail steels.