publications Constrained neighbor lists for SPH-based fluid simulations Neighbor lists with hard memory bounds, by locally adjusting particle support radii to cap neighborhood size. Infinite continuous adaptivity for incompressible {SPH} Continuous, unrestricted spatial adaptivity for incompressible SPH, rather than adaptivity in discrete resolution levels. Multi-Level-Memory Structures for Adaptive {SPH} Simulations A hash map-based sparse data structure for highly adaptive SPH on GPUs. Multi-Level Memory Structures for Simulating and Rendering Smoothed Particle Hydrodynamics Extending multi-level memory structures to cover both simulation and rendering of adaptive SPH data. Semi-analytic boundary handling below particle resolution for smoothed particle hydrodynamics Boundary handling that stays consistent when boundary features are finer than the particle spacing. Optimized Refinement for Spatially Adaptive SPH Using optimization to decide how particles should be refined, instead of relying on hand-tuned refinement patterns. Visualizing Optimizers using Chebyshev Proxies and Fatou Sets Visualizing the behaviour of numerical optimizers by treating them as dynamical systems and rendering their Fatou sets. Lipschitz-agnostic, efficient and accurate rendering of implicit surfaces Rendering implicit surfaces accurately without needing a Lipschitz bound for the underlying function. Symmetric Basis Convolutions for Learning Lagrangian Fluid Mechanics Asking what the right convolution is for Lagrangian particle data, and building a symmetric basis formulation to answer it. workshopPapers Cross-Validation of SPH-based Machine Learning Models using the Taylor-Green Vortex Case Cross-validating SPH-based machine learning models on the Taylor-Green vortex, as a shared reference case. Physically-Motivated Machine Learning Models for Lagrangian Fluid Mechanics Practical lessons on building machine learning models for Lagrangian fluid mechanics that respect the physics. MoriNet - Machine Learning from a Mori-Zwanzig Perspective MoriNet — a machine learning architecture motivated by the Mori-Zwanzig decomposition. projects Spatially Adaptive Smoothed Particle Hydrodynamics Cumulative dissertation on spatially adaptive SPH, from neighbor lists through to boundary handling. diffSPH - Differentiable Smoothed Particle Hydrodynamics for Adjoint Optimization and Machine Learning A fully differentiable SPH framework in PyTorch, spanning incompressible to compressible schemes, built so that gradients flow through the solver. Solving Boundary Handling Analytically in Two Dimensions for Smoothed Particle Hydro-dynamics Exact algebraic boundary handling in two dimensions, to arbitrary order of accuracy. Learned Particle Shifting Learning the particle shifting step, one of the more expensive and more heuristic parts of an SPH solver. Mori-Zwanzig Formalism for Machine Learning Using the Mori-Zwanzig formalism to think about what a learned closure model is actually approximating. Transformers in The Loop Scaling learned Lagrangian models to larger simulations using transformer architectures. software OpenMaelstrom GPU solver for spatially adaptive, incompressible SPH — the framework behind the adaptivity papers. OmniSPH CPU solver for two-dimensional incompressible SPH, aimed at numerical experiments rather than scale.
-1400.webp)
-1400.webp)
-1400.webp)
-1400.webp)
-1400.webp)
-1400.webp)
-1400.webp)
-1400.webp)
