Machine learning techniques for predictive modelling in geotechnical engineering: a succinct review

Abstract

Abstract Geotechnical engineering plays a crucial role in evaluating seismic hazards, especially in areas prone to earthquakes, ensuring that infrastructure remains resilient. The advent of machine learning (ML) techniques has significantly enhanced the prediction of soil-structure interactions during seismic events, improving the accuracy of models. This review examines a variety of ML methodologies utilized in assessing seismic vulnerability, from large-scale evaluations to urban-specific analyses. By integrating ML with traditional geotechnical practices, decision-making is strengthened, leading to more resilient infrastructure designs. Key areas of focus include the prediction of foundation settlement, where various ML algorithms—such as regression models, hybrid approaches, and numerical analysis techniques—are emphasized for their contributions to real-time monitoring and risk management. The review highlights the necessity of accurate ground characterization in the design of foundations for tall buildings and presents innovative foundation systems that enhance resilience against fault rupture. Techniques such as aRVM, Random Forest (RF), PSO-ANN, Support Vector Machines (SVM), and numerical methods are discussed for their effectiveness in predicting settlement, building responses, and safety risks. The paper stresses the importance of addressing complex engineering challenges through hybrid algorithms and comprehensive ground assessments to advance predictive accuracy in geotechnical engineering across a range of infrastructure projects.

Keywords

• Machine learning (ML) techniques
• Predictive modeling in geotechnical engineering
• Geotechnical analysis
• Seismic hazard prediction
• Infrastructure performance prediction
• Data-driven modeling in geotechnics