权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

HPMultiscale: High Performance Simulation of Space-Time Multiscale Nonlinear Problems

HPMultiscale：时空多尺度非线性问题的高性能仿真

基本信息

批准号：
357361040
负责人：
Professor Dr. Daniel J. Rixen
金额：
--
依托单位：
Lehrstuhl für Angewandte Mechanik
依托单位国家：
德国
项目类别：
Research Grants
财政年份：
2017
资助国家：
德国
起止时间：
2016-12-31 至 2023-12-31
项目状态：
已结题

来源：
https://gepris.dfg.de/gepris/projekt/357361040?language=en
关键词：
HPMultiscale Performance Simulation Space Time

项目摘要

Multiscale problems are often encountered in the simulation of structural dynamics problems such as for instance when cracks appear in large structures. Such simulations are extremely challenging since efficient solution techniques for such applications still need to be developed. In the first phase of the research project, we could improve the accuracy of the time-integration techniques for such models, while keeping the cost reasonable by allowing for different time-steps in different domains. To this end, we developed a software-framework where the FETI-method is exploited as parallel solver on High Performance Computers. For those problems, we developed new solution strategies, but also realized that the full potential of the new approach applied to multiscale problems can only be exploited if so-called recycling techniques are used. It is common to restart the parallel solution-process in each timestep and therefore expensive  work done in previous steps is usually thrown away. Recycling-methods are able to selectively reuse such valuable information in later  timesteps as well. Due to that reason, in this follow-up project, we want to develop efficient recycling- strategies for an optimal reuse of information in High-Performance-Simulations of multiscale transient structural problems.

在结构动力学问题的模拟中经常会遇到多尺度问题，例如大型结构中出现裂纹时。这样的模拟是极具挑战性的，因为有效的解决方案技术，这样的应用仍然需要开发。在研究项目的第一阶段，我们可以提高这些模型的时间积分技术的准确性，同时通过允许不同域中的不同时间步长来保持成本合理。为此，我们开发了一个软件框架，利用FETI方法作为高性能计算机上的并行求解器。对于这些问题，我们开发了新的解决方案策略，但也意识到，只有使用所谓的回收技术，才能充分利用应用于多尺度问题的新方法的潜力。在每个时间步中重新启动并行求解过程是很常见的，因此在前面的步骤中所做的昂贵的工作通常会被丢弃。 回收方法也能够在以后的时间步中有选择地重用这些有价值的信息。 由于这个原因，在这个后续项目中，我们希望开发有效的回收策略，在多尺度瞬态结构问题的高性能模拟中实现信息的最佳重用。 