Advances the past decade in computer and computational science, especially developments in the fields of high performance computing and machine learning, have created new opportunities for simulation-based science for addressing in detail the uncertainties that impact performance predictions of different systems. Such advances are particularly relevant for natural hazard risk assessment. Proper quantification and evaluation of this risk might require integration/combination of complex numerical models, frequently at different scale and fidelity levels, consideration of diverse metrics related to the life-cycle performance of infrastructure systems, or even evaluation of responses at regional scale. Faithful consideration of all important sources of uncertainty impacting predictions in this context needs to leverage all advances coming from the broader UQ (uncertainty quantification) field for accomplishing computational efficiency. This is essential for avoiding simplifications of the problem, necessitated by computational-feasibility constraints.
This webinar will overview relevant UQ advances considered at the NHERI SimCenter for supporting computational efficient natural hazard risk assessment. It will start with an overview of a general probabilistic framework for risk quantification/assessment that allows seamless integration of UQ tools for efficient estimation of all relevant statistics. Discussion will focus on applications that utilize complex, black-box predictive models, perhaps with high-dimensional uncertainties. Stochastic simulation techniques will be considered for assessing risk for such applications and opportunities offered through different UQ computational tools will be overviewed, ranging from “standard” approaches such as parallel computing and enhanced Monte Carlo methods, to “advanced” Gaussian process metamodeling techniques. Some emphasis will be placed on the latter, discussing adaptive and multi-fidelity metamodel development or modifications that support implementation for applications that entail high-dimensional stochastic uncertainties. Examples discussed throughout the webinar will cover multiple hazards, ranging from earthquakes, to wind, to (time permitting) surge.
Dr. Alexandros Taflanidis is Associate Professor and the Frank M. Freimann Collegiate Chair in Structural Engineering in the Department of Civil and Environmental Engineering and Earth Sciences at the University of Notre Dame. He holds a concurrent position at the Department of Aerospace and Mechanical Engineering and he is also a Faculty Fellow at the Kellogg Institute for International Studies. He received his Bachelors (2002) and Masters (2003) in Civil Engineering from Aristotle University of Thessaloniki, Greece. He got his PhD in Civil Engineering with minor in Control and Dynamical Systems from the California Institute of Technology (2008). His research focuses on uncertainty quantification and uncertainty-conscious analysis/design, with main applications on natural hazard risk mitigation and sustainability/resilience of civil infrastructure systems. A special area of interest for his group is the integration of surrogate modeling techniques and machine learning tools in risk assessment/design and stochastic optimization.
Thank you for your submission.