Quantifying tunneling risks ahead of TBM using Bayesian inference on continuous seismic data

Abstract

Seismic tunnel look-ahead methods, used during Tunnel Boring Machine (TBM) excavation, provide key advantages in forecasting potential tunneling risks and improving operational reliability and efficiency. However, these methods often grapple with data uncertainty due to the complex subsurface medium and incomplete data, causing tunnel engineers to question seismic prediction reliability. The analysis of uncertainty for seismic tunnel look-ahead methods remains largely unexplored. In this paper, we delve into these uncertainties using field tunnel look-ahead seismic data. We propose a Bayesian update workflow, grounded in a novel real-time seismic tunnel look-ahead method, to quantify tunneling risks and mitigate seismic uncertainty using continuous data measurement. We construct probability density functions of potential risks from seismic prediction outcomes and introduce Bayesian inference to update the probability with multiple observations. We demonstrate this proposed method’s application in various geological conditions in an urban Singapore area through several field experiments. Results indicate that our method enhances geological interface prediction reliability, minimizes interference from abnormal bodies adjacent to the tunnel, and demonstrates the ability to integrates interpreted geological information to refine tunnel look-ahead predictions.