Framework for the systematic characterization of uncertainties in geotechnical numerical simulations
Researcher: Renmin Pretell (PhD, 2022)
Abstract:
Numerical simulations in geotechnical applications, such as site response analyses (SRAs) and nonlinear deformation analyses (NDAs), require the use of soil and ground motion parameters. Most of these parameters are inherently spatially variable and thus challenging to measure and select for analyses, even when extensive geotechnical investigations have been conducted. This study will develop a framework for the characterization of uncertainties in numerical simulations to 1) prevent excessive or lack of conservatism in forward prediction of engineering demand parameters given the selection of spatially variable parameters, and 2) understand the sources of uncertainty and identify its reducible components by conducting further investigations or using more sophisticated numerical tools.
On its first stage, this research has focused on the investigation of the ability of 1D SRAs performed using state‑of‑practice approaches to capture a realistic 2D mean seismic response. State‑of‑practice approaches have generally led to either a single estimate of the seismic response, or to a mean seismic response lower than a realistic 2D estimate (see Figure 1, in terms of transfer functions). Results indicate that the 84th percentile of the seismic response estimated using the Toro randomization model (1995) reasonable captures a realistic mean seismic response.
References:
- Pretell, R., Abrahamson, N.A., and Ziotopoulou, K. (2021). Capturing 2D seismic response using 1D site response analyses. Earthquake Spectra (under review)
- Pretell, R., Ziotopoulou, K., and Abrahamson, N. (2019). Methodology for the development of input motions for nonlinear deformation analyses. 7th International Conference on Earthquake Geotechnical Engineering, Rome, Italy.