UCLA

The Shear Wave Velocity Profile Database (VSPDB) was developed in prior work to archive and disseminate data from sites with shear wave velocity (VS) profiles, including seismic velocities, soil stratigraphic information, and penetration resistances. The VSPDB is organized as a structure query language (SQL) relational database and is populated to date mainly with data from California. The first major component of this project was to augment the database with information from 30 bridge sites in California with 251 borehole logs and 88 VS profiles. The information was obtained from the California Department of Transportation (Caltrans) and translated into unified machine-readable formats via a computed program (UNIFY). The second major component of this project was related to VS prediction models. A number of models were reviewed and conditioned on penetration resistance, depth, and soil type parameters. A major distinguishing characteristic of available models is whether they allow for different levels of VS and penetration resistance scaling with depth (or overburden stress). I selected a model by Brandenberg et al. (2010) for validation against VSPDB data, which includes differential scaling features. That SPT-based model was developed and recommended by Building Seismic Safety Council for VS prediction for California sites. Querying of the VSPDB led to 2453 data pairs (1415 for sand, 641 for clay, and 397 for silt) consisting of all predictor variables (penetration resistance, depth, soil type) and VS values. Plots of data trends relative to model predictions were examined, and residuals analyses were performed to investigate potential model biases and whether the data requires different levels of scaling with N60-values or overburden pressure. The results indicate that the model overpredicts VS for all three soil types (sand, clay, and silt) but mainly for clay. The scaling of VS with N60 is stronger from the data for sand but not the other soil types. No adjustments to overburden scaling were required. Adjusted model coefficients based on the data analyses are provided.