NASA missions often involve the development of new vehicles and systems that must operate in harsh domains with a wide array of uncertainties and operating conditions. These missions involve high-consequence safety-critical systems for which experimental data are either very sparse or prohibitively expensive to collect. Limited heritage data may exist but, typically, are also sparse and may not be directly applicable to the system of interest, making uncertainty quantification (UQ) andmodel validation extremely challenging. Furthermore, NASAmodeling and simulation standards require estimates of uncertainty and descriptions of any processes used to obtain these estimates. The NASALangley Research Center developed a UQ challenge problem in an effort to focus a community of researchers toward key technical challenges common to many practical applications. The problem statement can be found in work by Crespo et al. (“The NASA Langley Multidisciplinary Uncertainty Quantification Challenge,” 16th AIAA Non-Deterministic Approaches Conference, AIAA Paper 2014-1347, Jan. 2014), and the computational models for it are available at http://uqtools.larc.nasa.gov/nda-uq-challenge-problem-2014/. The challenge problem features key issues in model calibration, uncertainty quantification, global sensitivity analysis, and robust design using a disciplineindependent formulation. The problem formulation is indeed discipline independent, but the underlying model, as well as the requirements imposed upon it, describe a realistic aeronautics application. A key aspect of the problem that makes it novel and challenging is the presence of both aleatory and epistemic uncertainties in a setting requiring consistency between their qualitative and quantitative prescriptions. As such, some uncertain parameters aremodeled as randomvariables, others as unknown constants lyingwithin known intervals, and others as probability boxes. This invited edition compiles the responses of 11 research teams to the challenge problem. The responses were generated by key discipline experts from U.S. research laboratories, industry, and academia. Participants included Sandia National Laboratories, Los Alamos National Laboratory, University of Southern California, University of Florida, Vanderbilt University, Supelec, University of Liverpool, Southwest Research Institute, GE Global Research, NASAAmes Research Center, and the Swiss Federal Institute of Technology. The breadth and depth of the solution strategies proposed constitute a survey of the state of the practice on uncertainty quantification from a practical engineering-like perspective. The usage of alternative and often dissimilar UQmethodologies for tackling the very same problem enables assessing their strengths and limitations on a unifying and fair setting.We hope that the scope of the challenge problem, as well as the responses to it presented herein, will keep inspiring scientists and engineers to develop more effective and efficient UQ technologies for improving the credibility and consistency of simulation-based analyses and designs.
Read full abstract
Mar 28, 2023
Andrey O Kuzmin + 2
Open Access