On Thursday, Ulsan National Institute of Science & Technology (UNIST) announced that the research team led by Professor Yoon Eui Sung from the Department of Nuclear Engineering has developed a groundbreaking algorithm that rapidly identifies collision points of high-speed particles in virtual nuclear fusion devices.
According to the research team, the new algorithm when applied to V-KSTAR, boosted detection speeds by up to 15 times compared to previous methods. V-KSTAR is a digital twin that replicates KSTAR, South Korea’s thermonuclear experimental reactor, in a 3D virtual environment.
Nuclear fusion power generation, often known as the artificial sun, relies on injecting high-speed neutral particles to create sun-like temperatures inside the reactor. However, if these particles veer off course and collide with the reactor walls, they can cause damage or disrupt the fusion process.
To address this issue, the research team ingeniously adapted collision detection algorithms used in the gaming industry. Their novel approach outperforms the traditional Octree method by a factor of 15.
Unlike the Octree method, which pre-divides space and checks each section for particles, this new algorithm performs calculations only when necessary.
Previously, it had to constantly monitor up to 300,000 particles potentially colliding with a wall divided into 70,000 triangles. The new algorithm eliminates about 99.9% of these calculations through simple arithmetic.
Moreover, the triangular division technique enables accurate collision detection even in complex 3D fusion reactor geometries.
The algorithm visualizes high-heat areas on the V-KSTAR display, allowing even non-experts to easily identify potential danger zones.
Professor Yoon emphasized that this collision algorithm is crucial not only for tracking neutral particle beams but also for expanding V-KSTAR’s overall 3D capabilities.
This study was supported by the Korea Institute of Fusion Energy (KFE) and the Korea Institute of Energy Technology Evaluation and Planning, utilizing KAIROS supercomputing resources at KFE.
The findings were published in the April issue of Computer Physics Communications, a prestigious international journal.