Jong-Hyun Kim* Mintae Na** Junho Kim** Sejin Yun** Jung Lee*** Sun-Jeong Kim**
(* : Inha University, ** : Hallym University, *** : Hanbat National University)
PLOS ONE 2026
Abstract : This paper presents a GPU-accelerated visualization framework for interactive exploration of fusion-plasma simulation data represented as scalar fields and particle trajectories. The primary technical contribution is a capacity-safe isosurface-extraction pipeline that combines inclusive slab partitioning with a two-pass counting and exact allocation strategy. This design prevents slab-boundary omissions and output-buffer overflow while supporting large scalar grids. The generated positions and normals are written directly to CUDA–OpenGL interoperable buffers, enabling the standalone renderer to display explicit isosurfaces without an intermediate CPU-side copy. The extracted geometry and scalar-volume data are additionally integrated into Unreal Engine through a separate staging interface. This integration provides a unified environment for volume rendering, explicit-isosurface inspection, particle-trajectory visualization, clipping, and VR-based spatial exploration. These engine-side functions are used as interaction and visualization mechanisms rather than being presented as new rendering algorithms. Experiments on electromagnetic-potential fields with resolutions up to 5123 demonstrate stable isosurface extraction across different isovalues. In the most geometrically complex tested case, the method generates approximately 42.81 million vertices in 44.7 ms, uses less than 11 GB of GPU memory, and renders the resulting geometry at 28.6 FPS. The results indicate that the proposed framework provides a practical explicit-surface visualization path for large plasma datasets while complementing direct volume rendering and established scientific-visualization toolkits.
[paper]