Sensitivity Analysis and Bayesian Calibration of OpenSees Models Using quoFEM

Abstract

The NHERI SimCenter is a nine-year research project that aims to advance the simulation of natural hazard impact on the built environment and communities. The SimCenter is developing several open-source workflow applications and an underlying scientific application framework. All applications built on this framework provide an OpenSees interface that enables users to use their existing models in advanced simulation studies, such as local and regional performance assessment, and uncertainty quantification (UQ). SimCenter applications provide researchers an opportunity to explore different extensions of their models by lowering the interdisciplinary barrier and encouraging collaboration. Among the applications, quoFEM provides access to UQ analyses with an easy-to-use, standardized interface. This work demonstrates the research enabled by quoFEM through the example of model calibration using PM4Sand, a soil constitutive model available in OpenSees. After an initial sensitivity analysis, the model is calibrated using Bayesian inference based on observations of hysteretic soil response from cyclic direct simple shear tests. The uncertainty in the model parameters is used in forward propagation to explore plausible lateral spreading scenarios due to seismic liquefaction. The results demonstrate the utility of quoFEM to the OpenSees community as a UQ-enabling tool.