Physics-informed machine learning in geotechnical engineering: a direction paper
Физически информированное машинное обучение в геотехнической инженерии: обзор перспективных направлений
2025-05-22
SCID: 54.1/wk3bgwek
Discuss with AI
digital twinsgeotechnical engineeringphysics-informed machine learningslope stability analysissoil mechanics
Figures from the paper
Abstract (AI)
This direction paper explores the evolving landscape of physics-informed machine learning (PIML) methodologies in the field of geotechnical engineering, aiming to provide a comprehensive overview of current advancements and propose future research directions. Recognising the intrinsic connection between geophysical phenomena and geotechnical processes, we delve into the intersection of physics-based models and machine learning techniques. The paper begins by elucidating the significance of incorporating physics-informed approaches, emphasising their potential to enhance the interpretability, accuracy and reliability of predictive models in geotechnical applications. We review recent applications of PIML in soil mechanics, hydrology, geotechnical site investigation, slope stability analysis and foundation engineering, showcasing successes and challenges. Furthermore, we identify promising avenues for future research in geotechnical engineering, including the integration of domain knowledge, model explainability, multiphysics and multiscale problems, complex constitutive models, as well as digital twins and large AI models within PIML frameworks. As geotechnical engineering embraces the paradigm shift towards data-driven methodologies, this direction paper offers valuable insights for researchers and practitioners, guiding the trajectory of PIML for sustainable and resilient infrastructure development.
Key Findings
1
Digital twins and large AI models are identified as promising components for extending PIML frameworks in geotechnical engineering.
2
Future PIML research should integrate geotechnical domain knowledge, explainable modeling, multiphysics and multiscale formulations, and complex constitutive models.
3
PIML is positioned as a pathway toward more sustainable and resilient infrastructure through data-driven geotechnical methodologies.
4
Physics-informed machine learning can improve the interpretability, accuracy, and reliability of predictive models in geotechnical engineering.
5
Recent PIML applications span soil mechanics, hydrology, geotechnical site investigation, slope stability analysis, and foundation engineering, with demonstrated successes alongside unresolved challenges.
Research Object
physics-informed machine learning methodologies in geotechnical engineering
Research Subject
the interpretability, accuracy, reliability, applications, challenges, and future research directions of physics-informed machine learning for geotechnical processes and applications
Publication Details
Publication Date
2025-05-22
Journal
Publisher
ISSN
Cited by
86
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai7
Exploiting Generative AI to Scale up Intelligent Tutoring Systems2023
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks2015
Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting2021
Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks2021
Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s Next2022
Physics-Informed Neural Networks for Heat Transfer Problems2021
ChatGPT and Open-AI Models: A Preliminary Review2023