Geotechnical Lessons Learnt—Building and Transport Infrastructure Projects

Slope stability analysis is a branch of geotechnical engineering that is highly amenable to probabilistic treatment, and has received considerable attention in the literature. This paper tries to demonstrate how probabilistic methods can be applied to slope stability analysis. The probabilistic methods used in this paper range from simple First Order Second Moment method (FOSM), First Order Reliability Method (FORM) and Monte Carlo Simulations (MCS) to more advanced Random Finite Element Method (RFEM). The importance of considering the spatial variability of soil properties in slope stability analysis is highlighted by comparing these methods. The paper also demonstrates how to use the Response Surface Method (RSM) to improve the computational efficiency in probabilistic slope stability analysis.

Engineering design involves the application of scientific and mathematical principles to provide functional solutions to technical problems while considering the limitations imposed by regulations, safety, practicality, time and cost. Allowable Stress Design (ASD) traditionally has been used by Australian geotechnical practitioners, and the introduction of Limit State Design (LSD) in the Australian Standards has not resulted in a smooth transition from the conventional method and remains a topic of debate in the industry. Reliability Based Design (RBD) is the next generation of geotechnical design that is not only physically and mathematically self-consistent, but also is compatible with the use of RBD for structures which makes it ideal as an inter-disciplinary common language. Therefore, Reliability Based Design is actively being introduced into design codes around the world as the main concept for the development of the design codes, standards and guidelines. From personal experience and observations, as an Australian geotechnical practitioner, a few examples of the application of ASD and LSD methods in small- and large-size projects are presented. Examples include critical design issues such as misuse of ASD and LSD; poor identification of modes of failure, adoption of design parameters, interpretation of safety factors and interpretation of the contents of key Australian Standards, such as AS 2159, AS 5100.3 and AS 4678. Some thought-provoking psychological biases and fallacies are presented, then some showcase solutions by RBD are offered and finally some suggestions are provided to improve the Australian Standards to benefit from this design method. Some challenges in employing RBD in routine geotechnical practice are discussed; however, detailed descriptions are beyond the scope of this paper.

According to Sect. 4.1 of the Australian Standard AS 5104, structures and their components must be planned, built and upheld to meet their intended purpose throughout their design lifespan, while also being economically viable. Additionally, the standard emphasises assessing reliability based on potential failure consequences, as well as the costs, efforts and procedures required to mitigate such failure risks. In theory, an engineer could brief an asset owner, and hence adopt a risk profile to suit their needs and be compliant with Australian Standards and the National Construction Code. An example of using reliability-based methods to assess the slope of landfill liners and an example of assessing the probability of failure for solar farm foundations is presented. A key barrier to the adoption of reliability-based methods is an estimation of probability density functions for all of the relevant parameters when data is limited. The use of Bayesian methods to derive parameters for the landfill liner is discussed.

It is generally accepted that, in civil engineering construction projects, the largest element of financial and technical risk generally lies beneath the ground. Indeed, structural foundation failure, construction over-runs and delays can often be attributed to inadequate and/or inappropriate site investigations. Unfortunately, geotechnical engineers have, at their disposal, limited guidance when scoping the extent and nature of site investigations. Often, the scope of geotechnical investigations is not governed by what is needed to characterise appropriately the subsurface conditions but, rather, is driven by budgetary constraints. A pressing need is to arm geotechnical engineers with guidelines that link the scope of a site investigation to ground variability and the probability that the foundation will be under-designed, resulting in some form of failure, or over-designed, resulting in the foundation being larger and more costly than needed. This paper outlines research undertaken to develop such guidance, focusing on the design of pile foundations in variable ground using the probabilistic techniques of random field theory, Monte Carlo simulation and genetic algorithms (GAs). The GA analyses showed that, when the number of boreholes is less than or equal to the number of piles, the boreholes are best located coincident with the piles. A single borehole should be placed at the building’s centre-most pile. With two boreholes, they are best located at the sides of the building, and three boreholes should be placed to form an equilateral triangle, as much as possible, while still being near the piles.

Perhaps the static cone penetration test is one of the most popular in situ testing procedures to explore the geotechnical properties of soil layers. This device, however, cannot be easily employed in offshore sites where the soil is relatively inaccessible. On such sites, free-falling (dynamic) penetrometers are employed to provide information on the mechanical properties of the soil. These devices can provide the total time and depth of penetration as well as the deceleration characteristic of the penetrometer to infer soil properties. Nonetheless, explicit relations between the penetration characteristics and the soil properties do not exist in the literature, limiting the application of dynamic penetrometers in practice. Based on numerical methods and experimental tests, there are useful databases that relate the penetration parameters to the soil properties. In practical site investigation, the soil properties such as the shear strength, stiffness, strain rate dependency, and profile non-uniformity may be inferred from the dynamic penetration information by solving a complex inverse analysis problem. In this study, ten alternative machine learning techniques are employed to extract the soil properties from the dynamic penetration results obtained by experimental tests or numerical analysis. The results indicate that the extreme gradient boosting method outperforms its counterparts for the problem analysed in this study.

The use of numerical models for geotechnical analysis and design is now becoming routine. Numerical modelling is a complex process which can have enormous benefits but can also be misused which can lead to dangerous applications. This has been recognised within the industry through C791 published by the Construction Industry Research and Information Association (CIRIA). The guidance within C791 aims to improve the outcomes for modelling projects by providing advice on the development of clear and achievable objectives, and the required roles and responsibilities to achieve them. It also seeks to ensure the roles, responsibilities and experience of the project team are appropriate and that strategies for model calibration and testing of sensitivity to assumptions and robustness are identified and implemented. This paper summarises how the guidance was followed during a project to assess ground movement around an excavation for the proposed Atlassian Building Central, a 39-storey hybrid timber building adjacent to Central Station. The ground movements were to be used to assess risks to third-party assets around the excavation. Clear objectives were identified and individual models were developed to specifically satisfy these objectives. A design team was formed with the appropriate skills and experience which ensured key questions were asked and risks were identified and mitigated in a timely and efficient manner. Modelling complexity evolved sequentially and element calibration was performed to ground the analysis in reality. Critical inputs and variability of these inputs were identified and quantified. Sensitivity to variations in these inputs was explored, thereby identifying potential project risks and opportunities. The resulting design was transparent, efficient and verifiable. Risks and opportunities were easily communicated to the client, and third-party stakeholders were satisfied the risks to their assets could be managed effectively. At a time when numerical modelling and analysis techniques are becoming more prevalent, more complex and more relied upon, it is deemed critical that processes are put in place to reap the benefits while also managing the risks. This paper demonstrates how the processes and advice set out in C791 can be used to benefit all projects where advanced numerical analysis techniques are employed.

Density reconciliation of tailings deposits is an essential step in the design and operation of a tailings storage facility (TSF) because it provides the required information to forecast the storage capacity and life span of the facility. Density reconciliation is necessary because different stakeholders (e.g. designers, processing plant operators and TSF operators) inform their density predictions from various data sources. For geotechnical designers, density predictions can rely on evaluating the compressibility and hydraulic conductivity properties of mine tailings to predict filling rates and settlements of tailings deposits. This prediction is a challenging task because the conditions present during testing in the laboratory are generally very different from those existing at the field scale. It is customary to adjust tailings parameters and other inputs required to analyse settlements using engineering judgement to reconcile the model predictions with a limited number of observations from site performance. The adjusted inputs can then be used as the design parameters of the deposit. In this study, we discuss a Bayesian approach for the inference of the design parameters of the tailings that combine prior knowledge of the parameters with information from different sources, including laboratory testing data and observations from site performance. The deposit consolidation settlements are calculated with a commercially available one-dimensional large-strain consolidation model (FSConsol). The Bayesian approach is coded in Python, and the consolidation model is incorporated within the code with a response surface (RS). The method is illustrated with an example of an unnamed TSF that serves to highlight its advantages.

This paper investigates the impact of biases and logical fallacies on outcomes in the context of underground engineering. It examines the decision made by a coal mining contractor to modify excavation equipment that would enable continuous mining capabilities. Established and recent theoretical decision frameworks are reviewed and implemented in this decision analysis, revealing decision oversights and logical fallacies that resulted in project and organisational failure. This paper empirically validates the need for sequential and staged decision frameworks which identify and remediate the impact that biases have on decision processes that influence project outcomes.