Physics-guided neural network with engineering priors for subgrade compaction prediction

doi:10.1016/j.autcon.2026.107202

Physics-guided neural network with engineering priors for subgrade compaction prediction

2026-08-13

0

PRE

AI

DOI：10.1016/j.autcon.2026.107202

Original

Original request for help

Save

Abstract

En 中文

• A physics-guided deep learning framework is proposed for subgrade compaction prediction. • Engineering priors are introduced as differentiable constraints without requiring explicit governing equations. • The proposed model improves prediction accuracy, convergence stability, and cross-scenario generalization. • Feature-interval analysis reveal the mechanisms underlying model improvement in complex working conditions.

Keywords:

Subgrade compaction

Physics-guided learning

Compaction state evolution

PG-Bi-GRU

Physics-guided neural network with engineering priors for subgrade compaction prediction

Abstract

Journal

Organization