Monday  |  Salon 10  |  09:00 AM–09:20 AM

#15594, Interlaced Material Characterization and Model Calibration for Improved Computational Simulation Credibility

Computational simulation is increasingly relied upon for high-consequence engineering decisions, especially when exhaustive physical testing is not possible or practical. A foundational element to solid mechanics simulations such as finite element analysis (FEA) is a credible constitutive or material model. However, the selection, calibration, and validation of material models is often a discrete, multi-stage process that is decoupled from material characterization activities. Furthermore, current characterization and calibration methods often use limited data (e.g. load-displacement curves), leading to an incomplete description of material phenomenology. Altogether, current characterization-calibration approaches are time-consuming, expensive, and potentially inadequate.

We are developing a paradigm-shifting approach by interlacing experimental characterization and model calibration in a real-time feedback loop, in order to describe material phenemonology more completely, reduce model parameter uncertainty, and reduce cost and time required for calibration. Our proposed Interlaced Characterization and Calibration (ICC) paradigm has five core components: First, a multi-axial load frame subjects the specimen to complex stress states that reflect real-world loading conditions, instead of simplified, uniaxial, homogeneous loading. Next, advanced, full-field diagnostics such as digital image correlation (DIC) deliver rich characterization data compared to traditional load-displacement data, while spectral decomposition of the field data reduces its dimensionality for more efficient fusion with simulations. Third, a suite of material models is calibrated with quantified parameter uncertainties using Bayesian inference, as opposed to a deterministic calibration of a single model. Fourth, Bayesian Optimal Experimental Design (B-OED) is used to determine the loading regime that will reduce parameter uncertainties and bias, and the load frame actively drives the test specimen to this state. Finally, because Bayesian inference is prohibitively expensive with FEA, surrogate models are pre-built offline prior to entering the feedback loop. By interlacing characterization and calibration to actively drive the experiment in a feedback loop, the ICC approach will align data captured with data needed, increase confidence in calibrations, and produce quantified uncertainty of model parameters.

SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Denielle Ricciardi   Sandia National Laboratories
D. Tom Seidl   Sandia National Laboratories
Brian Lester   Sandia National Laboratories
Amanda Jones   Sandia National Laboratories
Elizabeth Jones   Sandia National Laboratories