Granular Media Research - Our Thrusts

Uncover the mechanisms of keyhole pore dynamics

Using the high-fidelity simulation tool developed, we have conducted comprehensive investigation into the lifecycle of keyhole instability and keyhole-induced porosity to uncovers new physics and understanding. Our study found that the dynamic instability of the keyhole and the subsequent evolution of pores are governed by five interdependent factors: vapor condensation, liquid vortex, recoil pressure, surface tension, and keyhole morphology. A central finding establishes vapor condensation within an enclosed pore as the primary driver for generating high-speed microjets, which directly cause pore collapse and splitting. This mechanistic understanding led to the classification of three pore deformation regimes based on the condensation rate: "splitting and leaving," "shrinkage and leaving," and "shrinkage and capture," and proposed condensation control as a pathway for pore elimination. Read this AM paper.

Physics-based adaptive control to avoid keyhole pore formation

Building upon these core discoveries, subsequent research has established a mechanism-based optimization strategy using adaptive laser power. The keyhole pore formation process is delineated into five stages: "J-like" keyhole formation, keyhole closure, pore collapse, pore splitting, and pore motion. Analysis indicates that reducing laser power at the onset of the "J-like" keyhole formation stage is an effective method to stabilize the keyhole and prevent pore generation. To facilitate real-time control, adaptive indices were developed to quantify keyhole fluctuations from 2D imaging data, forming the basis of an optimization criterion. Parametric studies demonstrate that a strategy employing a moderate laser power reduction at a judiciously selected threshold can effectively suppress expansion-induced pore formation while maintaining high manufacturing efficiency. We are committing with industrial partners for practical application of this powerful prediction tool. Read this AM paper.

Ongoing: Powder-based Direct Energy Deposition (PB-DED)

Powder-based directed energy deposition (PB-DED) has emerged as a promising additive manufacturing technology for large-scale metal components, yet its process fidelity is plagued by the complex interplay of particle-laden flow, laser energy distribution, and melt track hydrodynamics. Our ongoing effort is to further extend our tool developed for L-PBF to a novel semi-resolved Computational Fluid Dynamics and Discrete Element Method (CFD-DEM) coupling scheme that could address these limitations for high-fidelity simulations of PB-DED.

Stay tuned and visit back!

FSDs + 3D Random Gaussian Field

However, our early FSD-based 3D method suffered from artificial anisotropy and could not reproduce planar facets common in natural sands. Our following 2014 CMAME paper resolves these limitations by introducing a rigorous 3D random field formulation on spherical topology, where particle surfaces are represented as correlated Gaussian random fields. Critically, the authors develop a novel spectral-to-spatial mapping that ensures the resulting covariance matrix is positive definite while preserving the frequency content of the input FD spectrum. This allows for true isotropic 3D shape generation, where low-order modes control elongation and global irregularities, and high-order modes govern surface roughness, now without directional artifacts.

This mature 3D framework is seamlessly integrated with Constrained Voronoi Tessellation to pack particles into arbitrary containers while controlling size distribution, solid fraction, and fabric anisotropy. A key innovation is the introduction of a shape-fitting parameter that aligns particle morphology with the geometry of its Voronoi cell, enabling dense, facetted packings unattainable with purely random shapes. Together, these series FSD papers form a cohesive pipeline: from experimental image &rarr FD spectrum &rarr statistically faithful 3D particle &rarr controlled packing &rarr physically insightful DEM simulation. This paradigm shifts discrete modeling from idealized abstraction to quantitative, morphology-aware prediction of granular behavior.

Poly-superellipsoid-based approach

We have further proposed a poly-superellipsoid-based approach on particle morphology for DEM modeling of granular media (Read the paper). This poly-superellipsoid framework unifies geometric versatility with computational efficiency. Unlike conventional approaches that rely on clumped spheres or complex NURBS representations, the poly-superellipsoid uses only eight shape parameters to describe a broad spectrum of convex particle morphologies. This compact parametric representation enables accurate fitting of real particle geometries, such as those derived from CT scans, while avoiding the excessive computational cost associated with high-resolution clump models.

A key original contribution lies in the development of a robust and efficient contact detection algorithm tailored to poly-superellipsoids. We introduced a hybrid approach that combines the Levenberg-Marquardt (LM) optimization method with the Gilbert-Johnson-Keerthi (GJK) algorithm, leveraging the strengths of both: LM provides rapid convergence for general contacts, while GJK ensures robustness in degenerate cases (e.g., flat face-to-face contacts). Built upon the common-normal contact concept and supported by an analytically defined support function, this algorithm enables accurate computation of contact geometry, including penetration depth and contact normals, for arbitrary convex shapes. The method is validated through rigorous numerical tests and demonstrated in large-scale simulations of granular packing and triaxial compression, showing that particle shape directly governs macroscopic responses such as shear strength, dilatancy, and fabric evolution.

Collaborated with government sectors, we have been engaged in developing innovative computational tools to simulate debris flow mitigation by flexible ring barrier. We coupled computational fluid dynamics (CFD) with discrete element method (DEM) to capture the multi-phase, multi-way, multi-physics interactions among the debris solids, the debris fluid and the different deformable components of the flexible ring barrier are captured, including ring-ring and ring-cable frictional contact and sliding. Read our latest paper on this subject.

We are able to have the same ring flexible barrier subjected to the impact of different geophysical flows which may occur in a natural setting, ranging from dry rock avalanche, debris avalanche, debris flow, debris flood to mud flow, to investigate the impact mechanisms. We found that the transition from pileup to run up mechanisms is found closely related to flow features (including Froude number, solid volume concentration and fluid rheology) and barrier responses (barrier components including barrier net, cables and brakes as well as their coordinated interactions in response to the impact). The insights on the two impact mechanisms and their transitions gained in the study may provide a useful reference for future design of flexible barriers in mitigating hazardous geophysical flows. For detail please reader our paper published in Engineering Geology (see also PDF).