【摘要】 以CPU计算为主研究立管涡激振动的商用及开源CFD软件效率偏低，而基于GPU的高性能计算软件尚不成熟。为此，利用Python编程语言和JAX框架，应用浸没边界-格子玻尔兹曼方法，开展了GPU集群加速的流固耦合数值计算方法研究，开发了立管涡激振动GPU集群模拟程序。与现有研究结果对比表明，该计算程序准确模拟了涡激振动的力学特性，具有较高的精度。通过增加GPU数量，计算效率可显著提升。同时，依托JAX框架提供的即时编译技术显著降低了计算程序开发的难度。论文提出的立管涡激振动GPU集群模拟方法可为相关研究提供借鉴。更多还原

【Abstract】 Computational fluid dynamics methods are crucial for studying vortex-induced vibrations(VIV) in risers.However, traditional commercial and open-source software primarily relies on less efficient computations of CPU, while GPU-based high-performance computing software remains underdeveloped. To address this, we used Python and JAX framework to investigate GPU-cluster-accelerated numerical methods of fluid-structure interaction, developing a GPU-cluster simulation program for the riser VIV. Comparisons with existing researches indicate that the computational program can accurately simulates the mechanical characteristics of VIV with high precision. Increasing the number of GPUs significantly enhances computational efficiency.Additionally, leveraging JAX’s Just-In-Time(JIT) compilation technology greatly reducing the difficulty of program development. Thus the proposed GPU-cluster simulation method for riser VIV can be a valuable reference for related research.更多还原