Multiscale modeling of cruciform dwell tests with the uncertainty-quantified parametrically homogenized constitutive model

Abstract

This paper presents a novel multiscale approach for analyzing multi-axial stress-strain evolution in Ti-7Al cruciform specimens under dwell loading, through the use of an uncertainty-quantified, parametrically homogenized constitutive model (UQ-PHCM). The thermodynamically-consistent UQ-PHCM is built from rigorous upscaling of crystal plasticity FE models (CPFEM) using machine learning and uncertainty quantification. They explicitly incorporate microstructural information in the form of representative aggregated microstructural parameters (RAMPs). Uncertainty quantification accounts for uncertainty in model reduction, data sparsity and microstructural descriptors. This paper integrates advanced multiscale, multi-axial experiments with computational modeling at multiple scales to establish the UQ-PHCM as an effective tool for bridging the gap between laboratory specimen-scale experimental observations and micro-scale stresses and strains using CPFEM. The CPFEM is calibrated and validated by experimental data from surface strain measurements using digital image correlation (DIC) and grain-by-grain lattice strain measurements with in-situ far field high energy diffraction microscopy (ff-HEDM). A computational method is developed in CPFEM, to incorporate initial residual stresses consistent with measured lattice strains. The UQ-PHCM is validated with biaxial tensile dwell test results performed on the cruciform specimen with satisfactory prediction of gauge strain evolution in DIC measurements. Uncertainty in the strain field due to microstructural variability is also corroborated by the DIC measurements.