Quantifying spatial geotechnical uncertainty to advance the practice of risk assessment and decision-making in tunnel projects

Abstract

The largest element of technical and financial risk on any tunnel project lies in the ground conditions. The conventional practice of developing deterministic interpretations of ground conditions does not account for ground spatial variability nor quantify uncertainty. Given the large number of claims and litigations in the tunneling industry, it is evident that conventional practice is ineffective in assessing ground spatial variability and uncertainty. Geostatistics-based methodologies are developed in this thesis to provide advantages over the qualitative and subjective deterministic interpretation of ground conditions with the probabilistic assessment of ground conditions as outputs. Geotechnical site investigation data from varied geological settings are utilized to offer effective solutions for unique challenges in tunnel projects and improve ground awareness. As the first contribution of the thesis, a probabilistic geostatistical methodology is developed to quantify the soil transition location uncertainty in the longitudinal and transverse directions of tunneling. Results from application to a soil tunnel project reveal that soil transition locations identified from the methodology agree reasonably well with the ground truth transitions, estimated via the difference in the rates of chamber pressure dissipation at the tunnel springline. For a different soil tunnel project, the results from the methodology were found to be within three to eight rings of the soil transition location identified by a data-driven model developed to characterize as-encountered ground conditions. Next, a geostatistics-based methodology is developed to generate quantitative estimates of karstic features' size, number, occurrence probability, and occurrence location within the tunnel envelope. Applying the methodology to a mixed-ground tunnel project in Malaysia reveals a presence of 2% to 6% average volumetric karstic void fraction for every 50 m of tunnel excavation. The results are utilized to estimate an average cumulative grout volume of 4000 m3 for the complete tunnel section. Next, a risk-based methodology is developed to optimize geotechnical site investigations considering tunnel risks and geotechnical uncertainty. For the risk created by uncertainty in tool wear rate, geospatial assessments of geotechnical parameter uncertainty, tunnel risk consequences, and investigation accessibility are developed to identify priority locations for additional investigations on a soil tunnel project. The results revealed that additional investigations reduced geotechnical parameter uncertainty by about 40 % that led to a reduction in the location uncertainty of the first two cutterhead interventions by about 90 rings (~160 m). Lastly, a methodology is developed to quantify the geostatistical model accuracy in predicting soil conditions, emphasizing predicting soil transitions. The methodology was applied to two soil tunnel projects to investigate the effect of geotechnical SI density and geological environment on geostatistical model accuracy. The results revealed that for a 100 m borehole spacing, geostatistical model accuracy in capturing soil transitions in a sedimentary and sequenced stratigraphy is about 60 %, while that for the heterogeneous glacial environment is less than 50 %. The geostatistics-based methodologies presented in this thesis are developed with a vision to assist the tunneling community in (a) analyzing ground spatial variability and uncertainty and (b) tying the results from geostatistical modeling to tunnel risk assessment. The latter is expected to help the tunneling community realize the advantages of geostatistical modeling-based methodologies in improving tunnel construction performance.