Bayesian Identification of Soil‐Structure Systems Using Seismic Response Measurements: A Case Study on a Field Test Structure

ABSTRACT The state‐of‐the‐practice soil‐structure interaction (SSI) analysis, referred to as the substructure approach, relies on simplified models of kinematic and inertial interaction. While computationally efficient, these models often lack accuracy in estimating impedance functions and foundation input motion (FIM) due to oversimplified assumptions about the soil‐structure system behavior. To overcome these limitations, this study presents a probabilistic soil‐structure system identification framework based on Bayesian inference to estimate substructure model parameters and FIM using recorded structural response data. The framework consists of two steps: the first estimates superstructure parameters, and the second jointly estimates impedance function parameters and FIM, thereby mitigating bias due to parameter interdependencies. The framework is applied to the Garner Valley soil‐foundation‐structure‐interaction test facility using two recorded seismic events for parameter estimation and a third for validation. The validated model demonstrates reasonable predictive capability. When applied across multiple soil‐structure systems, this framework can inform improved SSI modeling and analysis practices, leading to more accurate seismic response predictions for structural systems.