This paper presents a detailed review of existing and emerging deep learning algorithms for time series forecasting in geotechnics and geoscience applications. Deep learning has shown promising results in addressing complex prediction problems involving large datasets and multiple interacting variables without requiring extensive feature extraction. This study provides an in-depth description of prominent deep learning methods, including recurrent neural networks (RNNs), convolutional neural networks (CNNs), generative adversarial network, deep belief network, reinforcement learning, attention and transformer algorithms as well as hybrid networks using a combination of these architectures. In addition, this paper summarizes the applications of these models in various fields, including mining and tunnelling, railway and road construction, seismology, slope stability, earth retaining and stabilizing structures, remote sensing, as well as scour and erosion. This review reveals that RNN-based models, particularly Long Short-Term Memory networks, are the most commonly used models for time series forecasting. The advantages of deep learning models over traditional machine learning, including their superior ability to handle complex patterns and process large-scale data more effectively, are discussed. Furthermore, in time series forecasting within the fields of geotechnics and geosciences, studies frequently reveal that deep learning methods tend to surpass traditional machine learning techniques in effectiveness.

This paper considers centrifuge modelling of free-fall piezocones in soft clay, and interprets the test data to estimate the undrained shear strength and coefficient of consolidation. Data from the free-fall piezocones, which involved dynamic embedment in a normally consolidated kaolin clay sample from various drop heights, were compared with equivalent data from piezocone tests. The undrained shear strength, as interpreted from the acceleration trace of the free-fall piezocone, was found to be in good agreement with that derived from the piezocone penetration tests. The free-fall piezocone is also found to produce identical pore pressure dissipation behaviour as for a statically penetrated piezocone, despite a significant rise in the pore pressure at the start of the dissipation phase.

The finite-element analysis of free-falling objects penetrating soil deposits is one of the most sophisticated and challenging problems in geomechanics. A robust numerical method will be described here for dealing with such complex and difficult problems. The approach is based on the arbitrary Lagrangian–Eulerian (ALE) method of analysis, the main features and challenges of which are described briefly in the paper. The ALE method is then employed to perform a parametric study of a perfectly smooth penetrometer free-falling into a uniform layer of clay, which deforms under undrained conditions. The effect of the mechanical properties of the clay soil on the penetration characteristics is presented, and an approximate, closed-form expression is derived for the dynamic penetration factor, N_dp. Comparisons are made between the deduced values of N_dpand published values of the conventional cone factor, N_c, and comparisons are made with experimental data to validate the approach.

This paper investigates the uplift capacity of horizontal plate anchors in sand through finite element analyses and centrifuge experiments. The finite element simulations adopt a sophisticated bounding surface plasticity model that accounts for stress- and density-dependent behaviour, as well as loading- and fabric-related anisotropic effects in sands. Failure mechanisms at peak anchor capacity show that failure occurs progressively, with a marked decrease in mobilised friction angle within the shear bands close to the anchor edge. Numerical simulations of a large set of centrifuge experiments on rectangular, strip, and circular plates at different relative densities and stress levels are in good agreement for dense conditions, but perform more poorly for loose conditions due mainly to the open cone yield surface in the bounding surface model. Equivalent comparisons with current limit equilibrium methods highlight the challenges in direct application of element-level strength equations. Finally, the paper proposes a modified limit equilibrium solution based on a “rigid-block” failure mechanism extending to the soil surface, but with anchor factors that encompass the results from the finite element simulations. The modified solution provides a higher level of agreement with results from a large database of plate and pipeline test data than existing limit equilibrium methods.

This paper describes a centrifuge study using a range of penetrometer tests (T-bar, piezocone and free-fall piezocone) to explore strength changes in a reconstituted, normally consolidated, natural calcareous silt. Various penetrometer test procedures were applied to measure the penetration resistances including monotonic, cyclic and twitch-type movements as well as pauses for pore pressure dissipation. These mobilised combinations of partial or full remoulding, strain softening, consolidation and viscous rate effects. The penetrometer resistance – representing a proxy for strength – reduced by a factor of 4·1 from drained to undrained conditions (at the lowest fully undrained penetration rate). In undrained conditions, viscous enhancement of the penetration resistance raised the tip and shaft resistance in free-fall piezocone tests by ∼2·8 and ∼3·6 times, respectively. The ‘restart’ resistance immediately after the dissipation tests was ∼2·5 times higher than the resistance prior to dissipation, giving an indication of consolidation-induced strength gain. The ‘twitch’ test (using sequential steps decreasing the velocity) captured drainage and viscous rate effects, and also gave a ‘restart’ resistance that showed even greater consolidation effects than from a dissipation test. Overall, the different penetrometer test types and procedures measured resistances in the same soil sample that varied by a factor exceeding 20 from highest to lowest, resulting from different penetration rates and history, due to strain rate, strain level (or remoulding) and consolidation. An expression for the monotonic penetration resistance combining drainage and viscous rate effects was fitted to the response of all tests, spanning >7 orders of magnitude in strain rate.

The Casagrande thread-rolling method for determining the plastic limit of fine-grained soil is heavily dependent on operator judgement and can often give inconsistent or unreliable results. This paper presents an energy-based approach used in the development of an improved testing procedure for the plastic strength limit. A 0·727 kg cone is allowed to fall freely through 200 mm before contacting the surface of the test specimen, with the plastic strength limit determined for a cone penetration depth of 20 mm. For ten mineral clays of intermediate to very high plasticity tested, the plastic limits deduced using the cone were in good agreement with the measured Casagrande plastic limits. The values deduced using an 8 kg contacting cone were consistently lower than the Casagrande limits.