Moment-Based Sensitivity Analysis Framework for Stochastic Computational Models

Abstract

Computational modeling plays a critical role in engineering analysis and decision-making under uncertainty. When model inputs are described by random variables, the resulting outputs are inherently stochastic, making sensitivity estimation (SE) a fundamental step in quantifying the influence of input uncertainty. However, most conventional SE techniques are centered on input values, rather than the parameters defining the input distributions—limiting our understanding of how uncertainty at the distributional level propagates through complex systems. This presentation introduces a moment-based SE framework that directly evaluates the impact of input distribution parameters on key statistical characteristics of model outputs, including their moments and distributional forms. Sensitivity indices (SIs) are formulated with respect to the first four moments as well as the cumulative distribution function (CDF), enabling a direct assessment of probability-based performance metrics. A numerical post-processing approach is developed to efficiently compute these SIs following standard uncertainty quantification procedures, using moment-based models to approximate output distributions. The applicability and robustness of the method are demonstrated through several case studies, including nonlinear formulations and finite element models. The results highlight the method’s ability to capture subtle yet critical sensitivities, offering a broader and more informative perspective for stochastic modeling and reliability assessment.

Bio

Dr. Xuanyi Zhang is an Associate Professor at Beijing University of Technology and a Humboldt Research Fellow (2023). She received her Ph.D. in Civil Engineering from Central South University in 2019 and joined BJUT the same year, advancing to Associate Professor in 2022.

Dr. Zhang currently serves as Vice Chair of the Youth Committee of the International Association for Probabilistic Safety Assessment and Management, Deputy Secretary General of the International Society of Lifeline and Infrastructure Earthquake Engineering, and a core member of an innovation team supported by Beijing Municipality.

Her research focuses on structural reliability and risk management, with applications in high-speed railway infrastructure and lifeline engineering. Her work has been applied in engineering practice and supported by several national and municipal research projects. She is committed to fostering a positive academic environment and advancing engineering solutions through both theoretical development and real-world application.