Bayesian Model Updating for Soil-Structure System Identification Using Forced-Vibration Test Measurements

The state-of-practice soil-structure interaction (SSI) analysis is based on analytical models with simplifying assumptions regarding the soil-structure system behavior. There is, however, a lack of validation studies of the available analytical solutions. In this study, we propose validating the existing analytical models through identifying soil-structure systems, using their real-world measured responses, and comparing the identified models to their analytical counterparts. The proposed soil-structure system identification framework is based on a Bayesian time-domain model updating and parameter estimation technique. In this paper, the proposed framework is employed for identifying a test structure located in Garner Valley, California, using its recorded response during a forced-vibration test. The estimated values of the soil-structure model parameters show a notable discrepancy with their analytical equivalents. The potential sources of this observed discrepancy are discussed, and a refined solution is proposed.