Search

Understanding the variation of geological interfaces plays a crucial role in the analysis and design of infrastructure systems. Generally, there are two classes of techniques for predicting geological interfaces, for example, interpolation/regression-based techniques and machine-learning-based techniques. In this paper, a Multi-scale Meta-learning Model ([M.sup.3]) methodology is proposed. The new methodology improves the current state-of-the-art techniques by fusing two levels of information: (i) generic characteristics of the sampling locations, for example, coordinates, and (ii) location-specific characteristics, for example, local-scale predictions. The implementation starts from using an array of classic interpolation/regression-based techniques as base learners to provide first-level predictions at a local scale. These predictions are then combined with generic characteristics to train a meta-learner following the stacking ensemble learning framework. In this manner, the location-specific information from the base learners can be simultaneously considered with the generic information in the training process. The variation of rockhead elevation is predicted using the [M.sup.3] methodology and a comprehensive borehole dataset in Singapore. A detailed comparative study involving several existing methods is also carried out to rigorously validate the [M.sup.3] methodology. The results show that the [M.sup.3] methodology achieves 20% improvement in the model performance compared to existing methods, indicating its promising potential in geotechnical site characterization. Key words: geological interface, rockhead, geostatistical modelling, machine learning, meta learning, ensemble learning, 1. Introduction It has been reported that geological uncertainties can pose significant impacts on the performance of underground civil infrastructure systems (e.g., Li et al. 2016; Qi et al. 2016; [...]

A common source of error in bender element testing lies in the determination of the shear wave arrival time arising from the near-field effect, which is attributed to shear wave–like motion of the compression wave obscuring the arrival of the shear waves, and can be mitigated by using a sufficiently high travel-distance-to-wavelength (L/λ) ratio. However, the L/λ ratio also is relevant for signal interference due to side reflections, and its role in this important mechanism is far less understood. Stiff materials such as cement-treated soils, which have a very short shear wave travel time, are likely to exacerbate these errors. This paper examined the causes of travel time errors in cement-treated soil using bender element tests of standard-sized triaxial specimens, large block samples, and three-dimensional finite-element analyses. The results show that the reflection of shear waves from sample boundaries generates a second pulse in the received signal, resulting in signal distortion. This can be ameliorated at sufficiently high L/λ ratios. The reflection of the compressional waves generates spurious prearrivals and distorts the profile of the direct shear wave even at high L/λ ratios. Shear and compressional wave reflections can be mitigated by increasing the specimen diameter to decouple the latter from the direct shear wave arrival. [ABSTRACT FROM AUTHOR]

In this paper, a small strain numerical procedure, the press-replace method (PRM), is adopted to model the monotonic, quasi-static installation of a jacked-in pile under an undrained condition. Numerical simulations are carried out using the small-strain geotechnical finite-element software PLAXIS 2D (version 20.03.00.60). The PRM implementation for modeling jacked-in pile installation is described. This paper aims to demonstrate and assess the prediction potential and limitations of the PRM for modeling the pile installation process followed by the subsequent load testing, using published information from two well-known field case studies. It is shown that the combination of the PRM with the modified Cam-clay constitutive model can not only capture the penetration resistance experienced by the pile during jacking but also simulate realistic stress and pore-pressure changes around the pile caused by the jacking process. The pore-pressure response during postinstallation consolidation, which leads to the pile setup phenomenon, can also be reasonably captured. Subsequently, the importance of incorporating installation effects in the numerical simulations of pile load tests is demonstrated, using a wished-in-place (WIP) pile model and another with the installation process simulated via the PRM. The PRM model predicts a significantly higher pile capacity than the WIP model. The different load-settlement behavior of the two piles can be attributed to the different soil states present in the two models at the start of the load test, in which the WIP model does not capture the changes occurring in the stresses, overconsolidation ratios, and void ratios of the surrounding soil caused by the pile jacking process. [ABSTRACT FROM AUTHOR]

This paper presents the reliability-based design of a strip footing against bearing capacity failure in spatially variable soils using a deep learning approach. In this method, Convolutional Neural Networks (CNNs) is modified to construct, in the presence of soil spatial variabilities and uncertain loads, a single global response surface to emulate the behaviour of the geotechnical system over the entire design space. To improve the efficiency and accuracy of the deep learning approach, an adaptive training strategy is adopted. With sufficient training samples, the CNN-based global response surface can be used in lieu of the time-demanding finite-element model for efficient reliability-based design of the geotechnical system. The integration of the deep learning approach into the conventional double-loop framework is also demonstrated. Some key limitations of the double-loop framework are addressed using the deep learning approach. The results highlight that (i) the CNN-based global response surface successfully captures the behaviour of the strip footing over the entire design space, (ii) it can be seamlessly integrated into the conventional double-loop framework, (iii) it can effectively accommodate changes in target failure probability without excessive computational costs, and (iv) it successfully achieves a significant reduction in computational costs compared to other techniques. [ABSTRACT FROM AUTHOR]

Discrete element method (DEM) is a numerical technique that models the material as a combination of discrete particles which interact with each other through the contacts springs. While modelling soil in finite element method (FEM), approximations are made which considers soil as a continuous material. Hence certain properties such as void ratio which are intrinsic properties of the soil cannot be directly modelled or monitored. FEM also requires the failure criteria to be defined to predict the post elastic behaviour of the soil. DEM overcomes such drawbacks and the particle contacts are the only parameters that has to be defined. In the present work the contact parameters for sand grains are estimated for the purpose of being modelled in DEM, using direct measurement approach. The parameters required for defining the contact springs between the particles are the young's modulus, coefficient of restitution, friction coefficient, Poisson's ratio and density. The estimated parameters from the experiments were used to model the soil behaviour in an oedometer test. The obtained results from the DEM modelling are compared with the experimental results to benchmark the parameters estimated for the sand grains.