The mechanical response of crystalline materials is strongly influenced by their microstructure, particularly grain morphology and crystallographic texture, which are shaped by manufacturing processes and can evolve during service. In welded components, spatial variations in texture can lead to significant scatter in plastic deformation behaviour, yet this variability is rarely quantified in a systematic way. Crystal plasticity finite element (CPFE) modelling provides a physically based description of texture-dependent plasticity, but its high computational cost limits its direct use for uncertainty and sensitivity analyses. In this study, the influence of texture variability on the plastic response of electron beam welded 316L austenitic stainless steel is investigated using an experimentally informed uncertainty framework. Large-area electron back-scatter diffraction measurements are used to quantify realistic bounds of dominant texture components across the weld. These texture variations are propagated through a calibrated CPFE model using a surrogate based on polynomial chaos expansion, which acts as an analytical extension of the underlying crystal plasticity simulations. The surrogate model enables efficient evaluation of the mean response, uncertainty bounds, and sensitivity of the macroscopic stress–strain behaviour with respect to individual texture components, without the need for repeated finite element simulations. The results show that only a subset of texture components exerts a dominant influence on plastic response, while others have a negligible effect within the experimentally observed ranges. This work demonstrates how experimentally measured texture variability can be linked quantitatively to plastic deformation behaviour, providing a practical and physically grounded route for uncertainty quantification in microstructure-sensitive materials modelling. • Integrated CPFE–PCE framework for texture-driven uncertainty analysis. • EBSD data used to define realistic bounds of texture variability. • Surrogate model enables rapid prediction without repeated CPFE runs. • Sensitivity analysis identifies dominant texture components (Cube, Goss). • High predictive accuracy validated with low RMSE and relative error.
Kumar et al. (Fri,) studied this question.