| 宁民亮,范宣华,王柯颖,陈璞.大规模多点基础激励随机振动并行计算研究[J].计算力学学报,2022,39(1):13~20 |
|
| 
码上扫一扫！ |
| 大规模多点基础激励随机振动并行计算研究 |
| Parallel computing of large scale multipoint-base-excitation random vibration |
| 投稿时间：2020-11-10 修订日期：2021-03-28 |
| DOI：10.7511/jslx20201110002 |
| 中文关键词: 大规模有限元分析 多点基础激励 随机振动 PANDA平台 并行计算 |
| 英文关键词:large-scale FE analysis multipoint-base-excited random vibration PANDA platform parallel computing |
| 基金项目:国家自然科学基金面上项目（11872059）；国家重点研发计划课题（2016YFB0201005）资助项目 |
|
| 摘要点击次数: 788 |
| 全文下载次数: 247 |
| 中文摘要: |
| 解决大规模装备响应预测问题的有效途径是发展高性能结构动力分析软件。本文针对多点基础激励随机振动问题，推导并优化了相关求解公式。依靠JAUMIN框架和PANDA平台开展算法设计，研发出了多点基础激励随机振动并行求解模块。应用实例验证了算法和研发模块的精确性，在天河机群上测试了该模块的并行可扩展性。研究结果表明，研发模块具备与商业有限元软件相同的计算精度，其并行可扩展能力可达十亿自由度，CPU扩展至万核时并行效率仍高达80%，大规模计算能力远超现有商业软件。相关研究对于实现国内力学分析软件自主可控以及开展复杂装备评估具有重要意义。 |
| 英文摘要: |
| An effective method to solve the response prediction problem of large-scale complex equipment is to develop high-performance structural dynamic analysis software.Theoretical derivation was carried out for the multipoint-base-excited random vibration problem in this paper.Combined with the JAUMIN framework and PANDA platform to carry out algorithm design, and developed a multipoint-base-excited random vibration parallel analysis module.The correctness and parallel scalability of the algorithm and modules were verified through typical examples.The research results show that the analysis module can achieve the same calculation accuracy as some commercial FE software.Its parallel scalability can reach one billion degrees of freedom, and the parallel expansion efficiency of 10000 cores can reach more than 80%, far beyond the computing power of some commercial FE software.Related research is of great significance for achieving independent and controllable analysis software and carrying out the evaluation of complex equipment. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
