EAGER: Mapping Fragmentation and Topology Optimization Concepts to GPUs

这一早期概念探索性研究补助金(AGER)的目标是创建外部内聚碎片和拓扑优化概念到GPU的有效算法映射。外聚力断裂框架将被用来详细研究脆性和准脆性材料的动态断裂不稳定性，以适当地解释这些材料的极限断裂速度以及随着断裂速度的增加而增加的断裂阻力。这一框架还将允许在介观尺度上对非均质材料进行多尺度研究，考虑到含有硬颗粒的软基质的大变形行为，包括具有界面裂纹的渐变相区的细节。用于拓扑优化的GPU框架将考虑可能影响材料设计的真实地面结构，例如极端材料(例如，伸展材料)的设计，这些材料是全球同质化的，但可能在局部(微结构)显示功能分级的材料架构。映射和并行化技术将使用NVIDIA的CUDA(计算机统一设备体系结构)框架来执行，但也可以使用和/或探索其他新兴的体系结构，如英特尔的MIC(多集成核心)。能够充分利用GPU硬件是一门艺术，它依赖于在不同层次上关联软件和硬件的算法映射的有效性。为此，将创建一个定制的拓扑数据结构，以支持在GPU上进行网格修改和邻接搜索。为了避免竞争条件，将研究适当的算法(例如网格着色)及其对并行化性能和并发问题的影响。这项研究将利用国家超级计算应用中心(NCSA)与Volodymyr Kindratenko博士(研究科学家，NCSA)合作进行。这项跨学科研究的更广泛结果源于这样一个事实，即GPU一直是一种颠覆性技术，在非图形应用程序中具有巨大的潜力，例如在计算力学中。这一研究将有助于理解显式和隐式算法，分别针对每种情况采用代理问题，即分段和拓扑优化。要解决的问题的规模有可能导致通过新的物理理解和洞察力进行计算发现。这项研究产生的概念将被改编成伊利诺伊大学香槟分校(UIUC)的课程。教育和研究成果将通过互联网广泛传播。此外，还将开展外展活动，以激励高中生从事工程研究和教育事业。

The objective of this Early-concept Grant for Exploratory Research (EAGER) is to create an effective algorithmic mapping of extrinsic cohesive fragmentation and topology optimization concepts to GPUs. The extrinsic cohesive fracture framework will be used for detailed investigation of dynamic fracture instability of brittle and quasi-brittle materials to properly explain the limiting crack speed in these materials as well as increased fracture resistance with crack speed. This framework will also allow multiscale investigations of heterogeneous materials at the mesoscale, accounting for large deformation behavior of a soft matrix with hard particles, including details of the graded interphasial zones with interfacial cracking. The GPU framework for topology optimization will consider realistic ground structures that can impact material design, such as the design of extreme materials (e.g. auxetic), which are globally homogenized but may locally (microstructurally) display a functionally graded material architecture. The mapping and parallelization techniques will be performed using NVIDIA's CUDA (Computer Unified Device Architecture) framework, however, other emerging architectures such as the Intel's MIC (many-integrated-core) can also be used and/or explored. To be able to fully utilize GPU hardware is an art that relies on the effectiveness of the algorithmic mapping associating software and hardware at various levels. To this effect, a tailored topological data structure will be created to support mesh modification and adjacency searches on the GPU. To circumvent race conditions, proper algorithms (e.g. mesh coloring) will be investigated together their impact on parallelization performance and concurrency issues. The research will make use of the National Center for Supercomputing Applications (NCSA) through collaboration with Dr. Volodymyr Kindratenko (Research Scientist, NCSA). The broader outcomes of this interdisciplinary research derive from the fact that GPUs have been a disruptive technology with great potential for non-graphics applications, such as in computational mechanics. This investigation will contribute to the understanding of both explicit and implicit algorithms by adopting surrogate problems for each case, namely, fragmentation and topology optimization, respectively. The scale of the problems to be addressed has the potential to lead to computational discovery through new physical understanding and insight. Concepts developed from this research will be adapted into the curriculum at the University of Illinois at Urbana-Champaign (UIUC). Educational and research findings will be disseminated broadly through the internet. Moreover, outreach activities will be conducted to motivate high-school students to pursue careers in engineering research and education.