Machine learning to inform tunnelling operations: recent advances and future trends

Abstract

The proliferation of data collected by modern tunnel-boring machines (TBMs) presents a substantial opportunity for the application of machine learning (ML) to support the decision-making process on-site with timely and meaningful information. The observational method is now well established in geotechnical engineering and has a proven potential to save time and money relative to conventional design. ML advances the traditional observational method by employing data analysis and pattern recognition techniques, predicated on the assumption of the presence of enough data to describe the physics of the modelled system. This paper presents a comprehensive review of recent advances and applications of ML to inform tunnelling construction operations with a view to increasing their potential for uptake by industry practitioners. This review has identified four main applications of ML to inform tunnelling – namely, TBM performance prediction, tunnelling-induced settlement prediction, geological forecasting and cutterhead design optimisation. The paper concludes by summarising research trends and suggesting directions for future research for ML in the tunnelling space.

Shaping the future of tunneling with data and emerging technologies

“…Furthermore, most research studies also do not publish underpinning data due to confidentiality restrictions imposed by data owners. Therefore, the research landscape has developed in a piecemeal fashion using different data sets of different sizes and relating to different unique project characteristics (Sheil et al, 2020). Since data are a fundamental building block of data-driven systems, this data availability issue limits the reproducibility, generalizability, transparency, and, thus, the trustworthiness of the developed models.…”

EPBM Advance Rate Prediction Using Hybrid Feature Selection and Support Vector Regression Modeling

“…To properly asses TBM performance, one needs to accurately forecast advance rate (AR) or penetration rate (PR), i.e., the speed of tunnel excavation, since a realistic prediction of AR directly affects our ability to estimate cost, times of completion and assess risk [5]. However, it is difficult to make accurate predictions of AR as AR is not directly controlled by operator but a result of a complex interaction between the ground and the TBM [6,7]. Operators, particularly their experience and ability to react to observed parameters, have also a great influence on TBM driving performance.…”

“…In earlier stages, traditional statistical methods as well as experimental methods were widely adopted for AR prediction. However, the highly complex and non-linear interactions between different TBM parameters cannot be captured by these methods [7,9]. Recently, machine learning algorithms have been implemented to develop AR prediction models for TBMs based on machine monitored data.…”

“…Based on recent research [5,7,20,21], most of the features used in previous studies on AR prediction are determined by researchers' experience rather than systematically, sometimes failing to consider the most influential and efficient parameters. Simply selecting parameters based on previous studies, sometimes referring to tunnels under different conditions, is not rigorous and may result in loss of accuracy of the model.…”

Prediction of Pipe-Jacking Forces Using a Bayesian Updating Approach

“…Zhang et al 2009b, 2010b, Yang et al 2011, the maximum likelihood method (e.g. Wang et al 2014) and more general artificial intelligence techniques such as support vector machines and artificial neural networks (Sheil et al 2020a).…”