Jeffrey Florando | 17-ERD-029
Project Overview
At its core, material specification and materials qualification require an understanding of the uncertainties in relation to a given performance metric. When simulations are used to inform the design and evaluation of these metrics, it is essential to understand the uncertainties in the underlying models and parameters that feed into those simulations. In this work, we created a statistical framework to address uncertainties in the materials-strength modeling used in integrated simulations. The framework is based on the Bayesian methodology, which allows for uncertainties to be updated as new data become available. The results of this work show how model parameters and their uncertainties can be updated as new and different types of strength data are added, as well as methods for determining which future experiment has the greatest potential to reduce the uncertainty. The statistical framework developed in this project will be utilized to assess and propagate the uncertainties in the strength models used in programmatic simulations and to create a way to formally update those uncertainties as new data become available.
Mission Impact
Uncertainty quantification is a growing field within Lawrence Livermore National Laboratory's Stockpile Stewardship Program, and the tools and methods developed during this project offer the potential to be applied to specific mission challenges, including Livermore's Verification and Validation Program, which anticipates adopting this methodology. The uncertainty quantification of strength models described in this report is relevant to the NNSA mission to manage the nuclear weapons stockpile, as well as DOD efforts, including the Joint Munitions Program.
Publications, Presentations, and Patents
Bernstein, J., et al. 2019. "A Comparison of Material Flow Strength Models Using Bayesian Cross-Validation." Computational Materials Science 169 (18: 109098). doi: 10.1016/j.commatsci.2019.109098. LLNL-JRNL-758785
Bernstein, J. 2017. "Mixed Effects Material Strength Models with Model Discrepancy." Joint Statistical Meetings 2017, Baltimore, MD. LLNL-POST-735172
——— 2018a. "Bayesian Comparison of Material Strength Model Predictiveness." Data Science Institute (DSI) Workshop, Livermore, CA, 2018. LLNL-POST-754165
——— 2018b. "Comparing the Predictive Performance of Material Strength Models." Conference on Data Analysis 2018 (CoDA 2018), Santa Fe, NM, March 2018. LLNL-POST-746782
Florando, J. 2018. "Understanding Uncertainties for Materials Modeling." JOWOG-Mat 32, Lawrence Livermore National Laboratory, Livermore, CA, 2018. LLNL-PRES-761357
——— 2019, "Calibrating Strength Model Parameters Using Taylor Anvil and Stress-Strain Data." Center for Materials in Extreme Dynamic Environments Mach Conference 2019, Annapolis, MD, April 2019. LLNL-PRES-771647
Muyskens, A. 2020. "Error Propagation in Recursive Multi-fidelity Emulators." Conference on Data Analysis 2020 (CoDA 2020), Santa Fe, NM, March 2020. LLNL-POST-804597
Rivera, D. 2018. "Bayesian Calibration for Materials Strength Models." The Minerals, Metals & Materials Society TMS 2018 Annual Meeting and Exhibition, Phoenix, Arizona, March 2018. LLNL-PRES-747576
——— 2019. "Taylor Impact and Materials Strength Model Calibration." The Minerals, Metals & Materials Society TMS 2018 Annual Meeting and Exhibition, San Antonio, TX, March 2019. LLNL-PRES-76936
——— 2020. "Calibrating Strength Model Parameters Using Multiple Types of Data." The Minerals, Metals & Materials Society TMS 2020 Annual Meeting and Exhibition, San Diego, CA, February 2020. LLNL-PRES-807278
Schmidt, K. 2017. "Sensitivity Analysis of Strength Models Using Bayesian Adaptive Splines." 20th Biennial Conference of the Topical-Group of the American Physical Society on Shock Compression of Condensed Matter (SCCM), Saint Louis, MO, July 2017. LLNL-PRES-734153
——— 2018. "Parameter Subset Selection for Mixed-Effects Models." NNSA Joint Statistical Meetings 2018, Vancouver, Ontario, Canada. LLNL-PRES-751652
——— 2019a. "CNEC Research Summary and Transition to Lawrence Livermore National Laboratory." 2019 CNEC Workshop & Advisory Board Meeting, Raleigh, NC. LLNL-PRES-766720
——— 2019b. "Calibrating Strength Model Parameters Using Taylor Anvil Data." NNSA Joint Statistical Meetings 2019, Denver, CO. LLNL-PRES-781761
——— 2019c. "Uncertainty Quantification for Material Strength Models" Statistical Perspectives on Uncertainty Quantification Conference 2019, Chapel Hill, NC. LLNL-POST-774141
——— 2019d. "Validation and Uncertainty Quantification of Complex Models." DATAworks Workshop 2019 Springfield, VA. LLNL-PRES-771318
——— 2020a. "Sensitivity Analysis: Quantifying What Matters." Women in Data Science (WiDS), Livermore, CA, March 2020. LLNL-PRES-805661
——— 2020b. "Sequential Design for the Calibration of Materials Strength Models." Conference on Data Analysis 2020 (CoDA 2020), Santa Fe, NM, March 2020. LLNL-POST-805270
Schmidt, K., et al. 2017. "Sensitivity Analysis of Strength Models Using Bayesian Adaptive Splines." 20th Biennial Conference of the Topical-Group of the American-Physical-Society (APS) on Shock Compression of Condensed Matter (SCCM), Saint Louis, Missouri, July, 2017. LLNL-PROC-737638
