Validation of models used to predict the effects of lateral spreading by comparison of experimental and numerical response surfaces

Abstract

Following the LEAP-UCD-2017 and LEAP-ASIA-2019 exercises, an immense database of dynamic centrifuge data has been established for the purpose of validating numerical models that predict lateral spreading due to liquefaction. Using the experimental database and regression analyses, researchers have mapped a response surface to elucidate the relationship between soil density and shaking intensity to expected residual slope displacement. While it may not be practical to develop a database of experiments for every liquefaction scenario, the LEAP response surface may be conveniently used by others to assess that their numerical simulation procedures produce reasonable predictions for lateral spreading problems. To illustrate the value of response surfaces during the validation of numerical models, we developed a numerical response surface using the PDMY02 model implemented in the OpenSees finite element framework as an example test case. The methodology used to establish five PDMY02 constitutive model calibrations and the procedures to develop the numerical model are presented. One of the most important components of the calibration procedure, the attempted matching of the liquefaction triggering curves, is discussed in detail. Two input motions were considered for this work; one was a clean 1 Hz tapered sine wave, and the other included a 3 Hz component superimposed on the 1 Hz tapered sine wave. A numerical response surface was mapped for each input motion using the calibrations representing three relative density states. The greater transparency allows for comparisons between the numerical and experimental surfaces that illustrate the capabilities and limitations of numerical procedures, and therefore, this exercise should be integrated into future efforts to validate numerical models.