Physics-informed neural networks with learning-rate scheduling and domain decomposition for forward and inverse analysis of offshore flexible piles under complex loading

doi:10.1016/j.oceaneng.2026.125699

Physics-informed neural networks with learning-rate scheduling and domain decomposition for forward and inverse analysis of offshore flexible piles under complex loading

2026-04-28

0

PRE

AI

DOI：10.1016/j.oceaneng.2026.125699

Original

Original request for help

Save

Abstract

En 中文

• A forward and inverse physics-informed neural network framework for modeling slender pile in offshore environments. • The framework employs domain decomposition to handle subgrade reaction discontinuities in partially exposed piles. • The forward model predicts the performance of the offshore piles under complex loading conditions without monitoring data. • The inverse model accomplishes identifying the performance of the pile with sparse monitoring data.

Keywords:

physics-informed neural networks

domain decomposition

offshore piles

forward and inverse analysis

complex loading

Physics-informed neural networks with learning-rate scheduling and domain decomposition for forward and inverse analysis of offshore flexible piles under complex loading

Abstract

Journal

Organization