A Physics-Data Hybrid Model for Predicting Earth Pressure Evolution in Buried Horizontal Cylindrical Tanks

Physics-informed neural networks for predicting ultimate bearing capacity of layered soils under complex loading conditions

Seismic Response Prediction of a 2D Single-Story Steel Frame Using Physics-Guided Support Vector Regression Ensemble

Predicting the nonlinear seismic response of structures that have entered the plastic range under strong ground motions is severely constrained by data scarcity and computational cost. In this article, to address this dual challenge, we propose a physics-guided ensemble model based on Support Vector Regression. A finite element model of a single-story steel structure was created, and 500 nonlinear time-series analyses were generated using Incremental Dynamic Analysis for 50 different natural ground motions, at 10 levels of PGA intensity. Using an innovative feature engineering strategy, the 16 original ground motion parameters were decomposed into intensity, waveform and interaction features, thereby expanding the input space to 47 physically meaningful dimensions. Subsequently, the 35 features with the greatest information richness were extracted using a selection process based on mutual information. A systematic comparison with benchmark models demonstrated that Support Vector Regression (SVR) with Radial Basis Function (RBF) kernels offered significantly superior performance to Deep Learning with a reduced number of samples for this task, thus confirming the superiority of the structural risk minimization principle under conditions of limited data. Furthermore, the proposed two-level stacked ensemble achieved the lowest Mean Absolute Error among all evaluated models, demonstrating improved robustness in reducing prediction deviations and suppressing extreme errors in nonlinear seismic response estimation. These results demonstrate that the combination of physics-guided feature engineering and kernel-based learning provides an efficient surrogate approach for rapid seismic response prediction of steel structures under previously characterized ground-motion conditions.

A physics knowledge-based surrogate model framework for predicting multiple time-dependent responses of surrounding rock induced by tunnel excavation

A Physics-informed Neural Network Approach for Robust Buckling Load Prediction and Reliability-Based Design of Thin Truncated Conical Shells

Thin-walled truncated conical shells are widely used in aerospace, marine, offshore, and lightweight infrastructure systems due to their high strength-to-weight ratio and geometric efficiency. Their buckling resistance under axial compression, however, is highly sensitive to geometric imperfections, manufacturing tolerances, material variability, and nonlinear instability effects. Conventional design procedures rely on conservative knockdown factors (KDFs), such as those recommended in NASA SP-8019, which do not explicitly account for shell geometry, fabrication quality, data uncertainty, or target reliability. This study develops a physics-informed neural network (PiNN) framework for predicting critical buckling loads of thin truncated conical shells and integrates the trained surrogate within a reliability-based design (RBD) formulation. The model combines geometric and material descriptors with mechanics-informed features derived from shell stability theory and the localized reduced stiffness method (LRSM). A physics-informed loss function penalizes mechanically inadmissible predictions exceeding the theoretical elastic buckling load. The framework is trained and evaluated using 133 experimental Mylar conical shell tests under axial compression. Compared with a conventional deep neural network (DNN), the PiNN improves predictive accuracy, reduces mean absolute error, and enhances physical consistency. The trained PiNN is then used to evaluate reliability indices and calibrate safety-consistent KDFs for prescribed target reliability levels. Results demonstrate that the PiNN-RBD framework provides an efficient approach for uncertainty-aware design of imperfection-sensitive shell structures.

Integrating deep neural network with elastoplastic analysis: a hybrid approach for slope stability analysis

Conventional data-driven methods for slope stability analysis often exhibit an over-reliance on data while neglecting underlying physical principles. To address this limitation, this study proposes a physics-informed neural network (PINN) framework that integrates a neural network surrogate with the elastoplastic deformation mechanism of soil slopes. The governing equations incorporating the elastoplastic constitutive model based on the Mohr–Coulomb yield criterion, along with the boundary conditions, are embedded into the training framework of the PINN. The PINN functions as a surrogate model that requires no pre-constructed training dataset and automatically satisfies both the governing equations and the boundary conditions. Consequently, the developed PINN can directly predict the displacement field of a slope and automatically derive the associated stress–strain fields that comply with the deformation mechanism. These outputs are coupled with the multi-initial point sequential quadratic programming (MSQP) algorithm and the slip surface stress analysis (SSSA) method, enabling the efficient identification of the critical slip surface and the calculation of the corresponding factor of safety (FOS). The proposed method is validated through two illustrative examples. Comparisons of the results with those from commercial software confirm the high accuracy of the proposed method in predicting the stress–strain response and the FOS. This study provides a data-driven and physics-informed paradigm for slope stability analysis grounded in clear physical mechanisms.

Coupled CFD-DEM Modeling for Volumetric Sand Production Prediction and Proactive Wellbore Stability

Sustaining petroleum production targets requires developing marginal, weakly consolidated offshore reservoirs. However, continuous volumetric sand production causes severe borehole collapse and surface facility erosion. Legacy analytical Critical Drawdown Pressure (CDP) baselines assume an ideal post-yield elastoplastic state, ignoring transient rock-fluid degradation and yielding forecasting errors exceeding 20%. To resolve this multi-scale gap, this study presents a coupled simulation framework split into two computational nodes. In Node 1, an offline two-way coupled Computational Fluid Dynamics-Discrete Element Method (CFD-DEM) Digital Core tracks discrete particle kinematics and pore-scale hydrodynamics. A 1,200-run synthetic dataset generated via Latin Hypercube Sampling (LHS) mapped three sanding regimes, capturing wormhole propagation where localized porosity spikes from 0.24 to 0.85. In Node 2, a lightweight Physics-Informed Machine Learning (PIML) neural network is trained on this dataset. Unlike unconstrained black-box models, the PIML surrogate embeds fluid-solid mass conservation and Navier-Stokes partial differential equations (PDEs) directly into its loss function via automatic differentiation. When validated against a 24-month blind historical dataset from the Gulf of Guinea, the PIML surrogate achieved near-perfect fitment with an R2 of 0.97, an RMSE of 1.2 lb/1000 bbl, and an AAPRE of 7.4%, marking a 78.3% error reduction over legacy baselines without non-physical artifacts. Retroactively deployed as a dynamic virtual choke advisor, the surrogate minimized unexpected downhole cleanouts to zero, yielding a net 70.3% ($2.6 million) lifecycle operating expenditure (OPEX) reduction. The framework proves that physics-bounded intelligence successfully replaces reactive workflows to safely optimize drawdowns in unconsolidated assets.