The Optimization of Geotechnical Site Investigations for Pile Design in Multiple Layer Soil Profiles Using a Risk-Based Approach

Abstract

The testing of subsurface material properties, i.e. a geotechnical site investigation, is a crucial part of projects that are located on or within the ground. The process consists of testing samples at a variety of locations, in order to model the performance of an engineering system for design processes. Should these models be inaccurate or unconservative due to an improper investigation, there is considerable risk of consequences such as structural collapse, construction delays, litigation, and over-design. However, despite these risks, there are relatively few quantitative guidelines or research items on informing an explicit, optimal investigation for a given foundation and soil profile. This is detrimental, as testing scope is often minimised in an attempt to reduce expenditure, thereby increasing the aforementioned risks. This research recommends optimal site investigations for multi-storey buildings supported by pile foundations, for a variety of structural configurations and soil profiles. The recommendations include that of the optimal test type, number of tests, testing locations, and interpretation of test data. The framework consists of a risk-based approach, where an investigation is considered optimal if it results in the lowest total project cost, incorporating both the cost of testing, and that associated with any expected negative consequences. The analysis is statistical in nature, employing Monte Carlo simulation and the use of randomly generated virtual soils through random field theory, as well as finite element analysis for pile assessment. A number of innovations have been developed to assist the novel nature of the work. For example, a new method of producing randomly generated multiple-layer soils has been devised. This work is the first instance of site investigations being optimised in multiple-layer soils, which are considerably more complex than the single-layer soils examined previously. Furthermore, both the framework and the numerical tools have been themselves extensively optimised for speed. Efficiency innovations include modifying the analysis to produce re-usable pile settlement curves, as opposed to designing and assessing the piles directly. This both reduces the amount of analysis required and allows for flexible post-processing for different conditions. Other optimizations include the elimination of computationally expensive finite element analysis from within the Monte Carlo simulations, and additional minor improvements. Practicing engineers can optimise their site investigations through three outcomes of this research. Firstly, optimal site investigation scopes are known for the numerous specific cases examined throughout this document, and the resulting inferred recommendations. Secondly, a rule-of-thumb guideline has been produced, suggesting the optimal number of tests for buildings of all sizes in a single soil case of intermediate variability. Thirdly, a highly efficient and versatile software tool, SIOPS, has been produced, allowing engineers to run a simplified version of the analysis for custom soils and buildings. The tool can do almost all the analysis shown throughout the thesis, including the use of a genetic algorithm to optimise testing locations. However, it is approximately 10 million times faster than analysis using the original framework, running on a single-core computer within minutes.