SCEC Award Number 19135 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Output-Only Bayesian Nonlinear Site Characterization using Geotechnical Downhole Array Data
Investigator(s)
Name Organization
Ertugrul Taciroglu University of California, Los Angeles Pedro Arduino University of Washington
Other Participants Farid Ghahari, University of California, Los Angeles
SCEC Priorities 4c, 4b, 2b SCEC Groups EEII, CS, GM
Report Due Date 04/30/2020 Date Report Submitted 04/30/2020
Project Abstract
 Dynamic soil properties are key ingredients of analysis for predicting/assessing soil-structure interaction (SSI) and site response effects under seismic excitations. While there is already a large amount of valuable data recorded by the numerous geotechnical arrays around the world, there isn’t a reliable technique that enables the extraction of dynamic nonlinear/hysteretic properties of soil layers. In this project, we proposed a stochastic filtering technique to estimate nonlinear soil properties from earthquake data recorded in geotechnical arrays. To do so, a 1D-3C Finite Element Model of a site is modeled and the parameters of the soil’s constitutive model are identified using Unscented Kalman Filter techniques. The method was verified using synthetic examples and validated using data recorded in the Lotung site.
Intellectual Merit Dynamic soil properties are key ingredients of analysis for predicting/assessing soil-structure interaction (SSI) and site response effects under seismic excitations. Although there are various techniques and tools to carry out forward site response analysis with various levels of complexity, there could be significant uncertainty due to the available uncertainties in nonlinear soil models and their parameters. In this project, within a proposed general framework, we developed, tested, and validated an input-output identification algorithm to identify nonlinear soil properties form data recorded in the geotechnical array. For verification and validation, the Lotung site and its data were extensively studied.
Broader Impacts This project stimulated the development of a novel method for model selection and parameter estimation of nonlinear properties of soil layers from geotechnical arrays. While the focus was 1D site response, the method is easily extendable to the more complex 2 and 3D problems
One postdoctoral scientist worked in the development of efficient matlab scripts to combine site response analyses performed using a 1D -3C FE tool, developed as part of a related SCEC project, and a Bayesian framework.
The project has helped strengthen cooperation between the UW and UCLA research groups.
This project adheres to the initiative for a Technical Activity Group to coordinate SCEC research on nonlinear shallow crust effects. In this context, this project prompted cooperation and coordinated research activities between project 18020 led by Domniki Asimaki and our group. The Asimaki group used the same constitutive model but implemented in the Hercules framework.
Exemplary Figure Figure 2. The two-step solution.

Figure 14. The identification results using 5 parameters. Left: identification progress, right: comparison between recorded (simulated) and predicted responses in the EW direction at the ground surface and depth 6m.

credits: S. F. Ghahari, W. Zhang, P. Arduino, E. Taciroglu