# Cross-Scale Coupling Model of CPFEM and Thermo-Elasto-Plastic FEM for Residual Stress Prediction in TA15 Welds

## TL;DR

This paper introduces a new model that combines micro and macro simulations to better predict residual stresses in electron beam welding of TA15.

## Contribution

A dual-scale coupled model integrating CPFEM and thermo-elasto-plastic FEM for improved residual stress prediction in welds.

## Key findings

- The model reduces prediction error for molten pool morphology to within 16.3%.
- Peak longitudinal residual stress at the weld center is adjusted from 800 MPa to approximately 350 MPa.
- The model effectively captures shear stress components and reduces the 'M-shaped' stress distribution.

## Abstract

Existing macroscopic finite element models for electron beam welding (EBW) typically assume isotropic material behavior, often failing to accurately predict residual stresses induced by strong crystallographic textures. To address this limitation, this study established a sequential dual-scale coupled numerical model bridging micro-texture to macro-mechanics by combining the crystal plasticity finite element method (CPFEM) with thermal-elastic-plastic theory. Representative volume elements (RVEs) incorporating α and β dual-phase characteristics were constructed based on electron backscatter diffraction (EBSD) data from the TA15 weld cross-section. Through simulated tensile and shear calculations on the RVEs, homogenized orthotropic stiffness matrices and Hill yield constitutive parameters were derived and mapped onto the macroscopic model. Simulation results indicate that the proposed model maintains the prediction error for molten pool morphology within 16.3%, while effectively correcting the stress overestimation inherent in isotropic models. Specifically, it adjusts the peak longitudinal residual stress at the weld center from 800 MPa to approximately 350 MPa, significantly reducing the anomalous “M-shaped” stress distribution. By successfully capturing shear stress components, this work provides a high-fidelity computational approach for predicting complex stress states in welded joints, offering critical insights for structural integrity assessment.