Investigating Bayesian Parameter Identification Using Non-Standard Laboratory Specimens

This work investigates the applicability of Bayesian inverse analysis for identifying parameters from non-standard aluminum specimens with notches that induce stress concentrations. Unlike conventional standardized specimens, the notched samples used in this work are typically unsuitable for direct parameter extraction due to geometric irregularities and size effects. The experimental procedure involved tensile tests conducted using a universal testing machine, with deformation data collected via LVDT sensors and optical measurements with digital image correlation. The numerical simulations were performed using a quadrilateral finite element model with embedded strong discontinuities to capture the complete material response, including elastic, plastic, and fracture behavior. The proposed identification procedure successfully provided reliable posterior parameter estimates on aluminum rectangular and single-notch specimens. Furthermore, the identified parameters were validated on a double-notch specimen made of the same material. The results highlight the importance of parameter interpretation and show that the Bayesian framework can reliably identify key material and model-dependent parameters from non-standard specimens while accounting for uncertainty in both measurements and model formulation.