Predicting and Analysis of the Ground-Borne Vibrations Generated by Pile Driving Utilizing LSTM

Abstract

Existing subsurface and ground structures around the pile drive site are affected by ground-borne vibrations. These vibrations have an impact on surrounding structures and could be problematic. This work suggests and discusses the relevance of long short term memory (LSTM) deep learning (DL) algorithm for predicting and analyzing ground-borne vibrations generated by pile driving. More precisely, we investigate the ground-borne vibration characteristics’ predictability, potential prediction success, and improvement of the precise prediction time scales. We examine two of the most popular pile driving methods: impact pile driving and vibratory pile driving. We demonstrate that for each of the aforementioned driving types, the LSTM can effectively predict ground-borne vibration characteristics such as transverse (x) velocity, longitudinal (y) velocity, vertical (z) velocity, force, and inertia. Other vibration data types in soil dynamics and, more broadly, other vibration types found in engineering can also be predicted and analyzed using the LSTM-based approach suggested in this study. © 2025 Elsevier B.V., All rights reserved.