Conditional random fields for simulating spatial variability of geotechnical parameters in tunnelling

Abstract

In times of fast urbanization, underground infrastructure has taken primacy for servicing the city development. Shield tunnel construction is facing significant challenges due to geological uncertainty and limited availability of site-specific test data. As proven by recent works, tunnel design can be done more precisely within the probabilistic frameworks by considering the soil spatial variability, rather than using conventional deterministic-based approaches. In spatial variability modelling, the problem of incomplete utilization of known borehole soil parameters within the unconditional random fields can be overcome by generating conditional random fields. This paper presents a comprehensive review of the CRF application for spatial variability assessments in tunneling in the last five years. It finds that, when applying CRF-based probabilistic approaches, parameter uncertainty is undeniably reduced, and tunnel performance assessments are improved. Since CRFs have rarely been used for tunnel performance assesment, the paper recognizes research gaps that are of great interest to be further investigated in the future.