ABSTRACT
Sensing and information technologies, adopted now routinely on underground urban infrastructure projects, have enabled the production of massive streams of spatial-temporal data. Yet, geotechnical risk remains uncomfortably high and more often than not drives the schedule, cost and overall risk of such projects. In short, geotechnical risk is not decreasing in proportion to the uptake in big data. The reasons for this include a lack of methods to translate the increased spatial-temporal data into a commensurate reduction in geotechnical risk. This paper presents research-driven methods to reduce geotechnical risk using the increased data now available. A typical urban TBM project environment is presented and data from multiple actual tunnel projects is used to demonstrate method efficacy. The paper first addresses the quantification of baseline spatial geotechnical parameters and their uncertainty, and the consequential risks that exist during tunnel construction, e.g., ground and building deformation, stuck TBM, clogging, groundwater inflow, etc. Because these risks are quantified based on apriori site investigation data, optimization of geotechnical site investigation to best decrease risk uncertainty is addressed and demonstrated. Geotechnical monitoring data and TBM data collected during tunnel construction is used with back-analysis techniques to improve estimation, including decreasing uncertainty, of geotechnical parameters. This is demonstrated using 3D computational modeling of ground deformation and machine learning/empirical model-based estimation of clogging and TBM cutterhead wear risk. The paper also presents the reduction in geotechnical risk uncertainty by implementing a machine-learning approach to characterize the ground through which the TBM advances. A Bayesian approach to updating estimate magnitude and uncertainty of tunneling-induced ground deformation is presented by using the monitoring data and back-analysis.
