Experiment Design and Optimization for Scenario Assessment Based on Computer Simulations

• Activity Number: 106
• Type: Invited
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Brian Williams*+ and Christine Anderson-Cook and Leslie M. Moore and Jason Loeppky and Cetin Unal
• Companies: Los Alamos National Laboratory and Los Alamos National Laboratory and Los Alamos National Laboratory and University of British Columbia Okanagan and Los Alamos National Laboratory
• Keywords: computer model ; calibration ; scenario assessment ; experiment design ; optimization ; uncertainty quantification

Abstract

We investigate experiment design and optimization strategies for scenario assessment with complex multi-physics codes. Component physics models (e.g. material strength, equation of state) contain uncertain parameters that are calibrated to experimental data. Other parameters allow for scenario definition, such as the geometry of the physical system or assumed operating conditions. We wish to identify regions of the scenario variable space in which at least one critical metric operates outside of its control bounds. Scenario assessment that properly accounts for uncertainty requires physics parameter uncertainty to be propagated through calculated performance metrics. It may be necessary to reduce this uncertainty through improved calibration of physics model parameters to expand the domain of acceptable scenario variation. Since we assume that multiple performance metrics will be utilized to determine the best new data to collect conditional on current understanding, we propose an approach that utilizes a comprehensive resource allocation framework that updates the current analysis and optimally reduces residual physics uncertainty and optimizes the regime of allowable scenarios.