|About this Abstract
||2018 TMS Annual Meeting & Exhibition
||Computational Method and Experimental Approaches for Model Development and Validation, Uncertainty Quantification, and Stochastic Predictions
||Uncertainty Quantification in Materials Strength Models Using Bayesian Inference
||David Rivera, Jason Bernstein, Katie Schmidt, Nathan Barton, Ana Kupresanin, Jeff Florando
|On-Site Speaker (Planned)
The quantification of uncertainty in materials strength models plays a key role in the development and design of new materials. In this work an uncertainty quantification (UQ) methodology based on Bayesian statistics is applied to estimate the uncertainty linked to calibrating a strength model with limited experimental data. Specifically, the mechanical threshold stress (MTS) model is calibrated using Taylor anvil test data and the uncertainties in the parameters determined through the generation of their posterior distribution. The model is then applied to predict the deformation of Taylor impact experiments conducted outside the range of the data used in the calibration and the propagation of uncertainty demonstrated. Furthermore, the proposed methodology is compared with alternate UQ techniques such as perturbation methods. The results illustrate an approach to UQ which provides the materials design process with an awareness of inherent model uncertainty when making decisions based on limited experimental data.
||Planned: Supplemental Proceedings volume