Chapter 2: AI, Computational Geomechanics & Environmental Engineering¶

Chapter Context & Source Papers

This chapter synthesizes technical papers from Session 1 (AI and Computational Geomechanics) and Session 2 (Cold, Coastal and Environmental Geotechnical Engineering) presented at GAIC 2025 and published in GeoVadis: The Future of Geotechnical Engineering (Volume 3), CRC Press / Taylor & Francis (2026), DOI: 10.1201/9781003645955.

1. AI and Surrogate Modeling in Geotechnical Engineering¶

1.1 Strip Footing on Spatially Variable Slopes via Surrogate Models (Sharma & Pain)¶

Priyanka Sharma and Anindya Pain addressed the uncertainty of bearing capacity factors (\(N_c\), \(N_q\), \(N_\gamma\)) for strip footings adjacent to cohesive-frictional soil slopes characterized by spatial random fields:

Random Finite Element Method (RFEM): Cohesion \(c\) and friction angle \(\phi\) are modeled as log-normal and normal 2D random fields with anisotropic scale of fluctuation \(\theta_x, \theta_z\):

\[\rho(\Delta x, \Delta z) = \exp \left( - \sqrt{\left(\frac{2\Delta x}{\theta_x}\right)^2 + \left(\frac{2\Delta z}{\theta_z}\right)^2} \right)\]

Surrogate Machine Learning Model: Using an Extreme Gradient Boosting (XGBoost) and Gaussian Process Regression (GPR) framework trained on 2,500 Monte Carlo FE simulations, ultimate bearing capacity \(q_{ult}\) is evaluated in milliseconds with \(R^2 > 0.96\):

\[q_{ult} = c N_c^* + q N_q^* + \frac{1}{2}\gamma B N_\gamma^*\]

where modified factors \(N^*\) directly account for slope angle \(\beta\), setback distance \(d/B\), and spatial coefficient of variation \(COV_{c,\phi}\).

1.2 Extraterrestrial Regolith Drilling Force Mechanics (Raghav et al.)¶

V. Raghav and co-authors developed an analytical cutter-rock interaction model for lunar regolith coring mechanisms:

Drilling Mechanics: Total thrust force \(F_T\) and torque \(T_Q\) consist of three decoupled components: cutting face resistance, frictional rubbing along cutter wear flat, and chip evacuation clearing forces.
Low-Gravity Influence: Under lunar gravity (\(g = 1.62\text{ m/s}^2\)), chip packing ratio and effective confining pressure \(\sigma'_3\) decrease, significantly reducing cutter bit wear but increasing the risk of borehole wall sloughing in low-density regolith (\(< 1.5\text{ g/cm}^3\)).

1.3 State of AI in Geotechnical Practice: Frontiers and Pitfalls (Samui)¶

Prof. P. Samui delivered an objective appraisal of machine learning in geotechnics:

Bottlenecks: Small training datasets (scarce geotechnical boreholes), data imbalance, spatial autocorrelation neglect, and lack of physical interpretability ("black-box" models).
Physics-Informed Neural Networks (PINNs): Loss function regularized by governing differential equations (equilibrium \(\nabla \cdot \boldsymbol{\sigma} + \mathbf{b} = \mathbf{0}\) and seepage continuity \(\nabla \cdot (k \nabla h) = 0\)):

\[\mathcal{L}_{total} = \mathcal{L}_{data} + \lambda_{phys} \mathcal{L}_{PDE} + \lambda_{BC} \mathcal{L}_{BC}\]

2. Rock Mass Uplift Anchor Mechanics (Parab et al.)¶

G.S. Parab, V.V. Dandage, and A.V. Sharma analyzed high-capacity uplift rock anchors embedded in fractured basalt:

Load Transfer Mechanism: Field tension tests instrumented with multi-point borehole extensometers (MPBX) demonstrated non-uniform shear stress distribution \(\tau(z)\) along the grouted tendon length:

\[\tau(z) = \tau_{\text{peak}} \exp(-\alpha z)\]

Finite Element Verification: Numerical 3D FEM confirmed that anchor cone breakout failure angles deviate from the theoretical \(45^\circ\) inverted cone when sub-horizontal jointing dominates, flattening to \(30^\circ - 35^\circ\) and dictating deeper minimum embedment depths \(h_e \ge 6\text{ m}\).

3. Environmental Geotechnics & Barrier Engineering¶

3.1 Bentonite Backfill Unconfined Compressive Strength (Nayak et al.)¶

B.P. Nayak, S. Kumar, and R. Bag examined engineered clay backfills for deep geological disposal and cut-off walls:

Mix Proportioning: Sand-bentonite-fly ash mixtures. Hydraulic conductivity \(k < 1 \times 10^{-9}\text{ m/s}\) was achieved when bentonite content exceeded 15% by dry weight.
Unconfined Compressive Strength (\(UCS\)): Curing time and dry density \(\rho_d\) controlled post-swelling strength; excessive fly ash addition caused brittleness and reduced crack self-healing capacity under chemical leaching.

3.2 Red Mud Effects on Bentonite Shrinkage Potential (Shaikh et al.)¶

J. Shaikh and co-investigators stabilized bentonite liners using industrial bauxite residue (red mud):

Shrinkage Limit Modification: Bentonite suffers high volumetric shrinkage upon desiccation. Addition of 20% to 30% red mud introduced non-swelling ferric oxides and alumina, increasing shrinkage limit from 11.2% to 19.8% and eliminating desiccation crack networks.

3.3 Nanomaterial Mitigation of Alkali-Induced Swelling in Red Soil (Kumar et al.)¶

T. Aravind Kumar, P. Hari Prasada Reddy, and S.K. Vindula investigated high-alkali industrial effluents (\(\text{pH} > 12\)) impacting red clayey soils:

Swelling Mechanism: Hydroxide attack dissolves silica tetrahedra, triggering severe swelling and loss of fabric integrity.
Nano-Silica and Nano-Alumina Stabilization: Incorporating 1.0% nano-additives promoted pozzolanic calcium-silicate-hydrate (C-S-H) gel growth, reducing alkali-induced swelling index from 68% to under 14%.

4. Bio-Mediated Geotechnics: Low-Cost MICP (Rawat & Satyam)¶

Vikas Rawat and Neelima Satyam formulated a cost-effective Microbial Induced Calcite Precipitation (MICP) technique substituting expensive commercial nutrient broth with sugarcane molasses:

Bacterial Kinetics: Sporosarcina pasteurii urease activity in diluted sugarcane molasses broth (3% v/v) matched standard nutrient broth within 48 hours (\(U \approx 18\text{ mM urea/min}\)).
Calcite Precipitation & Strength:

\[\text{CO(NH}_2)_2 + 2\text{H}_2\text{O} \xrightarrow{\text{Urease}} 2\text{NH}_4^+ + \text{CO}_3^{2-}\]

\[\text{Ca}^{2+} + \text{CO}_3^{2-} \rightarrow \text{CaCO}_3 \downarrow\]

Geotechnical Enhancement: Sand specimens treated with molasses-MICP exhibited \(UCS\) increases from \(0\text{ kPa}\) to \(1,840\text{ kPa}\) and a 2.5-order-of-magnitude reduction in hydraulic conductivity, slashing treatment costs by 62%.

5. Marine Geotechnics: Floating Offshore Wind Turbines (James & Haldar)¶

M. James and S. Haldar simulated 15-MW semi-submersible floating offshore wind turbines (FOWT) moored in deep marine strata:

graph TD
    Turbine["15-MW Semi-Submersible Platform"] --> Mooring["Taut & Catenary Mooring Lines"]
    Mooring --> Anchor["Suction Caisson Anchors in Clay"]
    Anchor --> Soil["Cyclic Degradation of Undrained Shear Strength (su)"]
    Soil --> Pullout["Ultimate Inclined Pullout Capacity (H-V Interaction)"]

Coupled Aero-Hydro-Servo-Elastic-Geotechnical Modeling: Mooring line tension amplitudes under 100-year storm conditions and seismic shaking triggered cyclic pore pressure generation \(\Delta u / \sigma'_{v0} > 0.45\).
Suction Caisson Capacity Envelope: Yield surface under combined horizontal (\(H\)), vertical (\(V\)), and moment (\(M\)) loading:

\[\left(\frac{H}{H_{\text{ult}}}\right)^a + \left(\frac{V}{V_{\text{ult}}}\right)^b \le 1.0\]

The analysis emphasized that dynamic cyclic softening reduces available pullout resistance by up to 22%, requiring caisson aspect ratios \(L/D \ge 3.5\).