SimCenter | Uncertainty Analysis of Structural Seismic Response Parameters, June 28, 2017

SimCenter Series: EARLY CAREER RESEARCHER FORUM

ABSTRACT -- Uncertainty Analysis of Structural Seismic Response Parameters Using High-Throughput Computing: Reliability-based seismic design of structures requires an ensemble of nonlinear time history analyses (NLTHA) based on a nonlinear finite element (FE) model of the structure of interest. This ensemble of NLTHAs typically accounts for the seismic record-to-record variability, but can also consider the variability (uncertainty) of the FE model parameters. For detailed nonlinear FE structural models, a single NLTHA is computationally intensive (runtime on the scale of hours or days). To statistically quantify the variability of the structural response, Monte Carlo simulation (and its various derivatives) can be employed to set up the ensemble analysis. Monte Carlo simulation, known for being one of the most robust yet computationally expensive methods to propagate uncertainty through a numerical analysis, is made computationally feasible via parallelization on a supercomputer.

A 5-story three-dimensional steel moment building frame subject to both model parameter uncertainty and ground motion record-to-record variability is analyzed in parallel using a hybrid statistical/structural analysis software. This software was created by coupling two open-source software frameworks: Dakota, developed by Sandia National Laboratories, for uncertainty quantification and OpenSees (Open System for Earthquake Engineering Simulation) for advanced modeling and analysis of structural and geotechnical systems subjected to earthquakes. Some of the capabilities of this coupled framework will be illustrated through this application example.