Lessons Learned From Using Simple Supervised Learning Tools on Small‐Ensemble Data—Applicability to Tunnel Design and Monitoring

ABSTRACT Integrating interdisciplinary strategies with artificial intelligence (AI), particularly machine learning (ML), is an effective way of addressing urgent engineering challenges. Therefore, a thorough evaluation of existing methodologies is essential, taking into account their respective strengths, limitations and opportunities. This paper presents the main findings from exploratory research conducted through a variety of case studies. Based on the insights gained from these case studies, the paper critically examines three key areas of tunnelling. First, the challenges related to acquiring, generating and storing data, particularly for ML applications, are addressed. Emphasis is placed on ensuring that data are stored securely and are accessible for straightforward analysis. Second, the paper examines the application of ML to small datasets, providing insight into tunnelling requirements. It reviews ensemble methods and demonstrates their applicability using examples of small datasets. Third, the paper discusses the importance of interpretable tools in tunnel projects. Transparent and interpretable models help engineers understand model outputs, so it is important to consider this type of model wherever possible. The use of symbolic regression for estimating the long‐term closure of tunnels is presented. Finally, the paper summarises the key findings and considers the future prospects of this interdisciplinary approach. The aim is to encourage further development in this area.