High-accuracy Computing Simulation Platform : Multi-functional, High-precision, and Flexible to Simulate Various Physical Phenomena

CONTRIBUTING TEAM： Chau-Lyan Chang, Fang-An Kuo, Kuan-Lin Chen

Fluid dynamics extensively permeates nature, science, and engineering, representing one of the fields most closely intertwined with human life and socio-economic development. Utilizing computational fluid dynamics (CFD) simulations not only aids in exploring the physical world of fluids but also serves as a powerful tool in addressing related scientific and engineering challenges.

The significance of a fluid simulation platform for scientific and engineering research cannot be overstated. From aerospace and defense to the automotive industry, energy sector, biomedical sciences, environmental science, civil engineering, and meteorological observations, anyone involved with material flow and energy transfer can employ this platform for simulation analysis. This facilitates faster product design and operational efficiency, providing precise and reliable data for research and development, and assisting in conducting experiments that would otherwise be costly and time-consuming in the real world. To support Taiwan in creating a high-accuracy simulation platform that is both autonomous and forward-looking, NCHC has independently developed it by combining high-fidelity numerical methods with high-efficiency parallel computing technologies to tackle complex fluid dynamics challenges.

 
 Vorticity and Q-criterion distribution of fluid flowing through an underwater tool(140 million cells)

From a functional perspective, the NCHC's high-accuracy computing simulation platform boasts four main features: unstructured mesh generator, high-accuracy space-time solver, highly parallel computing capacity, and innovative use of containerization technology to enhance cross-platform usability. First, the platform's unstructured mesh generator allows for the modeling of objects with various geometries to explore the interaction forces between the object and complex external physical phenomena and environments, accurately simulating objects ranging from simple shapes to extremely complex appearances. The built-in immersed boundary method solver allows users to directly compute flows over either rigid-body or two-phase objects without the need for grid generations. Secondly, the high-accuracy space-time solver, developed by NCHC based on the Space-Time CESE method, exhibits high computational precision in both time and space and supports multiple turbulence models, effectively handling various mesh types.
 

Thirdly, the platform's computational power is outstanding, with an efficiency of up to 84% for over 1000 CPUs and 64% for over 2000 CPUs, easily managing large-scale simulations across multiple computer cores. Fourth, the innovative containerization technology of NCHC packages software applications in container modules, highly compatible with underlying hardware such as the high-speed InfiniBand network, allowing users to easily and quickly deploy software across different platforms without the need for recompilation, significantly simplifying the installation process.




Density gradient distribution of a supersonic fluid passing through a rotating square object(Immersed boundary method @ Re=200)

The NCHC's high-accuracy computing simulation platform is applicable not only in fluid mechanics but also widely in different scenarios and physical conservation phenomena, including space research, marine energy development, submarine ecology, defense, and energy sectors. NCHC will soon collaborate further with the Taiwan Space Agency, using the platform's simulation analysis capabilities to develop rockets and space vehicles. In the marine domain, it will be used for ocean current analysis off the coast of Taiwan to promote marine energy development and conduct studies on acoustic and biological impacts on submarine ecology.
 

Multi-functionality and high accuracy make the platform a powerful tool. With internally embedded HPC resources, NCHC enhances further the platform's accuracy, flexibility, and usability, thereby improving the computational precision of domestic academic and research institutions and facilitating deeper analysis of physical phenomena, which provides a powerful driver for innovation and ideation. The platform is still in the testing phase and will be initially made available to collaborating entities. As it matures, it will be offered to scientific projects and the industry, hoping to expand the platform's functionality and application scope through collaboration, thereby accelerating the future development of science and engineering applications in our country.