Bayesian calibration of a dislocation slip and twinning model for HCP metal strength

A recently developed hexagonal close packed (HCP) metal strength model relies on competition between dislocation slip and twinning deformation mechanisms to predict flow stress response. A combination of mechanistic and phenomenological elements form a generalizable model that is both parsimonious and reflects experimental data. The model is calibrated to a dataset of Kolsky bar and quasistatic experiments at a variety of strain rate and temperature conditions via Bayesian calibration, resulting in posterior distributions on the calibration parameters. Leave-one-out model performance quantifies the generalizability of the model and its associated uncertainty to unseen experimental conditions. Slip + twinning model performance is compared to competing models in terms of fit to experimental data, interpolation, and extrapolation behavior. Results indicate that consideration of both dislocation slip and twinning in the functional form allows for improved fit to data and more precise predictions.