arXiv subjects

Amiya Prakash Das

Publications and source records attributed to Amiya Prakash Das.

3 recordsLinked to original sources

arXiv2026

A variational physics-informed graph neural network for heterogeneous solid mechanics

Stress localization in heterogeneous solids is governed by the bimaterial interface, where the displacement field remains $C^0$-continuous, while in-plane stresses jump due to the stiffness mismatch. Coordinate-based physics-informed neural networks (PINNs) represent this jump via a prescribed regularization width or a weighted interface penalty, making their accuracy sensitive to how phase-contrast changes are handled. This work presents a variational, label-free physics-informed graph neural network (PI-GNN) in which the heterogeneity is carried by the discretization rather than by the trial field. The solver operates on a conforming adaptive mesh graph, assigns constitutive behavior per element, and minimizes the discrete total potential energy as a single unweighted objective in which only first derivatives appear. The discrete energy on piecewise-linear elements coincides with the finite element (FE) Ritz functional. Dirichlet conditions are enforced by construction, with no penalty term, no interface weight, and no prescribed transition width. Using one fixed architecture, optimizer, and loss across small-strain elasticity and finite-strain Neo-Hookean hyperelasticity in two and three dimensions, the von Mises error remains below $3.58\%$ across a stiffness-contrast sweep spanning $(E_{\mathrm{inc}}/E_{\mathrm{mat}}\in[10^{-2},10^{2}])$, where a strong-form PINN degrades to $5.58\%$, and its displacement error reaches $7.66\%$ against $0.49\%$ for the PI-GNN. A trained network halves the ($σ_{xx}$) error of an energy-based PINN ($5.01\%$ versus $10.94\%$). Training cost exceeds a single FE solve by more than an order of magnitude, so the construction is a variationally consistent, penalty-free interface representation for parametric surrogates and inverse identification rather than a replacement for a one-off FE analysis.

math.NA↗

arXiv2026

Hydrodynamic Behavior of Non-spherical Particles in Confined Vertical Flows: A Resolved CFD-DEM Study

We investigate the sedimentation and vertical hydraulic transport of non-spherical polymetallic nodules (PMNs) using resolved computational fluid dynamics-discrete element method (CFD-DEM) with multisphere particles spanning $98 < Re_\text{p} < 2904$. Shape effects induce 1.8-2.0 times drag enhancement relative to volume-equivalent spheres, arising from 50\% larger frontal areas and wake asymmetry, reducing terminal velocities by 27-29\%. Vertical transport exhibits velocity-driven transitions from intermittent settling to stable convection, as demonstrated by residence-time and drag-force statistics. While PMNs exhibit enhanced rotational-translational coupling and broader force fluctuations, the regime progression qualitatively resembles that of volume-equivalent spherical particles. Drag variance evolution reveals contrasting behavior: small particles $(d/D=0.082)$ show narrow distributions and wake suppression at higher velocities, while large particles $(d/D=0.22)$ exhibit non-monotonic variance. These findings elucidate shape-confinement interactions in vertical transport and establish bounds on the applicability of volume-equivalent spherical particles in reduced-order models.

physics.app-ph↗

arXiv2025

A DEM-driven machine learning framework for abrasive wear prediction

Particle-induced wear is a critical concern in bulk material handling systems, where abrasive interactions accelerate equipment degradation, increase maintenance needs, and raise operational costs. The Discrete Element Method (DEM) and Archard's wear model are widely adopted for predicting particle-surface wear processes. However, DEM is computationally prohibitive for real-time design and predictive maintenance, often requiring hours to days for a single parametric analysis. We propose a DEM-machine learning (ML) framework to address this limitation that combines physics-based simulations with data-driven efficiency. A dataset of 200 DEM simulations is generated by systematically varying particle size, material, and contacting plate geometric parameters. A few ML models -- linear regression, Lasso and Ridge regularization, decision trees, and a genetic algorithm-optimized artificial neural network (GA-ANN) -- were trained and evaluated. Feature selection revealed that Archard's wear constant, particle size, plate angle, and impingement velocity are the dominant predictors of wear. While linear models offered interpretability, their accuracy was limited. The GA-ANN achieved the highest performance $(R^2 = 0.91)$, effectively capturing nonlinear wear dynamics while reducing computational cost by orders of magnitude. This study demonstrates that physics-informed ML provides a scalable pathway for accurate, real-time wear prediction, enabling predictive maintenance and optimized design in bulk material handling industries.

physics.app-ph↗