Landslides are devastating natural disasters that occur frequently around the world, resulting in significant economic losses and casualties. Statistics show that the most frequent and widespread damaging type of landslides are triggered by precipitation. Machine learning techniques have been applied to predict landslides based on past events. However, there are a number of challenges associate with these techniques, including appropriate selection and design of machine learning algorithms, input features, and training datasets. In this work, we present an enhanced machine learning modeling strategy for landslide hazard assessment. Average landslide susceptibility map of California is generated based on the developed model, and is shown to align well with the historical observations.

Recurrent Neural Networks (RNNs) have been shown able to learn and predict complex mechanical behavior of geomaterials. In this talk, I will show that RNNs also possess the capacity to learn rate-and-state friction laws from synthetic data, another relevant phenomenon in the geotechnical setting.

The data employed for training the network is generated through the application of traditional rate-and-state friction equations coupled with the aging law for state evolution. A novel aspect of our approach is the formulation of a loss function that explicitly accounts for initial conditions, the direct effect, and the evolution of state variables during training. It is found that the RNN, with its GRU architecture, effectively learns to predict changes in the friction coefficient resulting from velocity jumps, thereby showcasing the potential of machine learning models in understanding and simulating the physics of frictional processes.

Geo-energy engineering and energy geotechnics will induce temperature changes in the soil grounds. It has been widely observed in laboratory and field investigations that the mechanical behaviour of clays is sensitive to temperature variations. For clayey soils, plastic deformation and time-dependent behaviour are especially concerning. In this study, a series of temperature-controlled laboratory tests, including oedometer tests, constant-rate-of-strain consolidation tests, and triaxial tests on both normally consolidated and over-consolidated clays. The results demonstrated that temperature changes have significant effects on the creep behaviour of clays. The viscoplastic strain rates are accelerated upon heating, which causes additional settlements in drained conditions, and excess pore pressure generation or effective stress relaxation in undrained conditions. Based on experimental observations, a new viscoplastic model of clayey soils is developed to account for the thermal effects. A virgin heating line is introduced into the existing equivalent-time model framework, providing a state-based theory for depicting the viscoplastic strain rates of clayey soils. The new model is further enhanced with a series of advanced features, including the plasticity anisotropy, structuration of sensitive clays, as well as the sub-loading behaviour of over-consolidated clays. The thermal parameters of the model can be easily calibrated through a heating-cooling test in oedometer conditions. The model is validated using laboratory test data with complicated temperature and loading conditions. Finally, the constitutive model is implemented in a finite element programme to conduct thermo-hydro-mechanical analysis on the long-term behaviour of energy foundations in clayey soil ground.

When a pile is embedded in a newly or recently reclaimed site, negative skin friction (NSF) is mobilized due to larger soil settlement compared to pile settlement. The generation of NSF imposes an additional axial force on the pile, which has a detrimental rather than beneficial effect. However, the impact of soil creep on the long-term development of NSF remains poorly understood. This study employs numerical analysis to investigate this effect. Initially, an elasto-viscoplastic model with an enhanced time integration algorithm is proposed and successfully implemented into the finite element package ABAQUS. Subsequently, a two-dimensional axisymmetric pile-soil interaction model is established and calibrated using a known field case. Parametric studies are then conducted to examine the varying degrees of creep effect on the evolution of NSF and the neutral plane (NP, which represents the position where soil settlement coincides with pile settlement), during both primary and secondary consolidation periods. The findings indicate that a higher creep coefficient of the soil leads to an increase in NSF and a descending trend of the NP. The delayed development of NSF due to creep is attributed to an increase in excess pore pressure during the early stages of consolidation. Moreover, the position of the NP undergoes significant variation at the onset of consolidation when considering creep effects. Finally, an exponential prediction model is proposed to account for the time dependence of the NP location.

Saturated loose alluvial deposits exhibit phenomena like soil liquefaction and alteration of ground motion during strong seismic event which cause damage to structures built on them. Though the saturated soils contain a significant amount of fines and still they are susceptible to liquefaction and hence their behaviour must be assessed through site-specific nonlinear ground response analysis (GRA). This study considers the response analysis of Bengal Geosynclinal basin filled with riverine sediments and are the part of the region around the city of Kolkata, India. The subsurface profile of these deposits has been characterized by drilling boreholes and consists of layers of very dense sand and silty layers having higher thickness above the hard surface. These layers have been overlain by the silty sand with loose to medium density and silty layers right up to the ground surface. The effective stress based nonlinear GRA is carried out for the selected basin sediment deposits using DEEPSOIL v7.0 to account for the shear strength degradation and to estimate the pore pressure generation. Critical-state-based constitutive models viz. PM4Sand and PM4Silt are also employed within the framework of fully-coupled finite element analysis to characterise the liquefaction response of the deposit. These models have been calibrated using a single-element modelling of the Cyclic Direct Simple Shear (CDSS) test. In the analyses, spectrum compatible time histories are used and responses in the form of variation of shear strain and pore pressure within the deposit with time is obtained. Though the deposit has appreciable amount of fines, the soil experienced liquefaction and thereafter the softening behaviour is noticed. This information is very useful in damage assessment of structures built on the liquefiable deposits having higher fines content. For future constructions in the Geosynclinal basin region, this information has to be used accordingly in the damage assessment.