The inherent variability of the soil’s geotechnical parameters contributes to the variability and uncertainty in geotechnical structures, resulting in less economic designs with higher margins for errors. In this thesis, the variability and uncertainty in key geotechnical parameters as well as pile load-settlement models are investigated. For geotechnical variability quantification in a data-scarce region, the city of Quito was used as a case study. A geotechnical database for the region was built using publicly available sources including dissertations, reports, and books. Using this database, the observed variability in several geotechnical parameters was characterised using probability distribution functions. Results for that region suggested that Weibull distribution is suitable for data of SPT (N), plasticity index (IP), average shear wave velocity over upper 30 m of soil at a site (Vs30), peak friction angle from the direct shear test (ϕDS), and apparent cohesion from the triaxial test (cTX). The peak friction angle from the triaxial test (ϕTX) was best fitted with a truncated normal distribution. The results of this analysis can be beneficial for regional geotechnical design and hazard assessment. The variability between the north and south zones of the Quito region was investigated, with results indicating that there is a measurable difference between the soils in the north and south of the region. For the uncertainty in pile load-settlement models, a well-established database for piles in the UK (DINGO database) was utilised to assess the variability of eight published methods used for estimating the ultimate load utilising field load-settlement data. The results of this analysis suggest that the method presented in Chin (1970) was observed to overestimate the ultimate load while the method presented in DeBeer (1970) tends to underestimate the ultimate load. Also, the average ‘factor of safety’ for the piles used in this study (defined as the average ultimate load estimate using the eight methods, divided by the elastic limits of the piles) was estimated to be 2.63 using the eight methods, with 26.6 and 1.05 maximum and minimum values respectively. Similarly, a review of two methods (Fleming (1992) and Niazi and Mayne (2015a; 2015b)) can predict pile behaviour in terms of load-settlement curve estimation. Ultimate load estimation for the two methods was compared to ultimate load estimation using the Chin method. Results of this comparison suggest that the method presented in Fleming (1992) produces higher reliability in the ultimate load estimations compared to the method presented by Niazi and Mayne (2015a; 2015b), which can be attributed to the less reliability in estimating key geotechnical parameters required for the method presented in Niazi and Mayne (2015a; 2015b) compared to the reliability in estimating the parameters required for the method presented in Fleming (1992).
Use of Open-Access Databases to Characterise Geotechnical Parameter and Model Uncertainty.
Othman, M. M. A. (Author). 10 Dec 2024
Student thesis: Doctoral Thesis › Doctor of Philosophy (PhD)