Prediction of induced soil vibration during pile vibrodriving using Dynamic Mode Decomposition (DMD)

Abstract

This study investigates using the Dynamic Mode Decomposition (DMD) algorithm to perform approximations and time-ahead prediction of soil vibrations during the vibrodriving process. Geotechnical applications face challenges in modeling and predicting soil vibrations due to the soil’s heterogeneous nature. This study addresses this issue using a purely data-driven approach. Geophone data collected during pile installation using a vibrodriver were used to demonstrate the feasibility of the DMD algorithm. The research reveals that both the standard DMD and augmented DMD, which incorporate delay coordinates, can achieve accurate predictions, with the augmented DMD producing more accurate time-ahead predictions of the vibrations. The results emphasize the potential practical utility of data-driven methods for vibration prediction in geotechnical applications.