Constitutive Relationship Study of Laves Phase NbCr2/Nb Two-Phase Alloy Using Modified J-C Model and Back Propagation Neural Network Model

Abstract

The Laves phase NbCr₂/Nb two-phase alloy has received significant research interest as a potential high-temperature structural material. Based on isothermal and constant strain rate compression experiments conducted on the alloy within a temperature range of 1000 -1200 °C and strain rate range of 0.001-0.1 s⁻¹, the flow stress constitutive relationship of the alloy was established using the J-C model and BP artificial neural network model, respectively. It was found that the conventional J-C model fails to describe the flow stress softening behavior of the alloy. In contrast, the modified J-C model provides a better prediction of the flow stress softening phenomenon and accurately characterizes the flow stress behavior of the alloy, it exhibits high prediction accuracy as indicated by the correlation coefficient (R) of 0.9902, average absolute relative error (AARE) of 8.773% and mean relative error (MRE) of 7.389%. The flow stress behavior of the alloy can be more accurately characterized using the constitutive relationship built by the BP neural network model. The model exhibits higher prediction accuracy with R of 0.9998, AARE of 2.232% and MRE of 0.870%. The results demonstrate that the BP neural network model has superior capability in predicting the flow stress behavior of the alloy. The established flow stress constitutive relationship can provide more accurate and reliable fundamental data with respect to flow stress for finite element simulations of forging deformation process of the Laves phase NbCr₂/Nb two-phase alloy. In addition, it serves as theoretical basis for rational design of forging process and accurate calculation of the deformation force of the alloy.