Physics-Informed Optimization for Dynamic Soil–Structure Interaction Analysis of a Pile Partially Embedded in Nonlayered Soils

Abstract

This study proposes a novel physics-informed optimization approach to account for nonlayered soil conditions for the dynamic soil–structure interaction analysis that is critical for structural health monitoring involving soil–structure interaction. This approach can estimate soil properties based on a few measurements and then predict the natural frequency of structures, a key element that has been very often utilized for structural health monitoring and evaluation. A case study for frequency-based bridge scour detection has been presented for the demonstration of this approach, where a bridge pile is partially embedded in complex nonlayered soils. With this case study, we show that the nonlinear frequency–scour depth relationship can be accurately estimated using 2–4 measured data points, which is in favor of situations where conducting a field test is difficult or the demand for conducting field measurements needs to be reduced. The impact of errors resulting from measurement uncertainty on the performance of this approach is found to be insignificant. This physics-informed optimization approach thus can help analyze dynamic responses of engineering structures in complex nonlayered soil conditions involving soil–structure interaction.