CAREER: Accelerating Real-time Hybrid Physical-Numerical Simulations in Natural Hazards Engineering with a Graphics Processing Unit (GPU)-driven Paradigm

该奖项全部或部分由《2021年美国救援计划法案》（公法117-2）资助。该学院早期职业发展（Career）奖将通过为使用图形处理单元（gpu）的高保真模拟奠定算法基础，促进对自然灾害对建筑环境构成的风险的基本理解。复杂结构系统的高分辨率模拟需要以独特的方式求解详细的模型。例如，高层建筑在功能上非常重要，以至于它们“太大而不能倒”，因为适度的损坏可能难以修复，高层建筑功能的丧失严重影响事件后的恢复。然而，由于计算成本和物理测试的限制，对高层建筑地震反应至关重要的土-结构相互作用（SSI）的影响往往被忽视。为了实现更真实和更快的模拟，本研究将推导、演示和促进先进的数值方法，利用gpu的大规模并行性，即最初为图形渲染开发的实时计算机芯片，克服结构模拟中的计算瓶颈，特别是在高层建筑的实时混合模拟（RTHS）中。与此同时，这项研究的多学科组成部分将与更大的教育承诺相结合，以发展、传播和持续反思一种包容性的教学方法，以增强学生对计算的坚持和乐趣，培养学生掌握日益技术驱动的劳动力所需的技能。该奖项将有助于美国国家科学基金会（NSF）在国家减少地震灾害计划（NEHRP）中的作用。利用RTHS进行复杂结构的高分辨率模拟，将物理实验与实时数值模型相结合，以前是非常困难的。精细化的高保真模型可以获得更高的分辨率和精度，但也会增加运行时间，从而抑制RTHS的可行性。图形处理器将首次被用于加速RTHS，以实现更高保真度的“动态”土木结构模拟。土木结构的地震反应对全GPU加速提出了独特的挑战，包括异构单元公式，不同程度的非线性，以及依赖于直接求解的隐式积分方案。为了应对这些挑战，本研究将重新制定在gpu上组装和求解运动方程的方法：(i)大规模并行算法，（ii）半离散时间积分方案，以及（iii）事件驱动的gpu适应RTHS架构。这项研究将在一个分层测试程序中达到高潮，以模拟真实的高层建筑反应，包括SSI。该项目将在结构工程和科学计算的交叉领域建立多学科研究和指导。相互合作将用于综合三个nsf支持的自然灾害工程研究基础设施（NHERI）设施的专业知识，包括加州大学伯克利分校的计算建模和仿真中心、里海大学的实验设施和德克萨斯大学奥斯汀分校的DesignSafe网络基础设施。项目数据将存档并在NHERI数据仓库（https://www.DesignSafe-ci.org）.This)中公开提供，该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，认为值得支持。

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2).This Faculty Early Career Development (CAREER) award will advance fundamental understanding of the risks posed by natural hazards to the built environment by laying the algorithmic foundation for high-fidelity simulations using graphics processing units (GPUs). High-resolution simulation of complex structural systems requires that detailed models be solved in unique ways. For example, tall buildings are so functionally important that they are “too big to fail,” as moderate damage can be difficult to repair and tall building loss of function severely affects post-event recovery. Yet, the influence of soil-structure interaction (SSI), critically important to tall building response during an earthquake, is often neglected due to computational cost and physical testing constraints. To enable more realistic and faster simulations, this research will derive, demonstrate, and facilitate advanced numerical methods that harness the massive parallelism of GPUs, i.e., real-time computer chips originally developed for graphics rendering, to overcome computational bottlenecks in structural simulations, specifically in the real-time hybrid simulation (RTHS) of tall buildings. In parallel, the multi-disciplinary components of this research will be integrated with a larger educational commitment to develop, disseminate, and continuously reflect on an inclusive teaching pedagogy to enhance student persistence and joy in computation, training students with the skills needed for an increasingly technology-driven workforce. This award will contribute to the National Science Foundation (NSF) role in the National Earthquake Hazards Reduction Program (NEHRP). High-resolution simulations of complex structures using RTHS, which couples physical experiments with numerical models in real time, has previously been exceptionally difficult. Refined, high-fidelity models result in greater resolution and accuracy but also suffer increased run time, inhibiting the feasibility of RTHS. Graphics processors will, for the first time, be used to accelerate RTHS to enable higher-fidelity "on-the-fly" simulation of civil structures. The seismic response of civil structures poses unique challenges for full GPU acceleration, including heterogeneous element formulations, varying degrees of nonlinearities, and reliance on implicit integration schemes with direct solvers. To address these challenges, this research will re-formulate approaches to assembling and solving the equations of motion on GPUs with: (i) massively parallel algorithms, (ii) semi-discrete time integration schemes, and (iii) an event-driven GPU-adapted RTHS architecture. This research will culminate in a tiered testing program to simulate realistic tall building response, including SSI. This project will establish multi-disciplinary research and mentorship at the intersection of structural engineering and scientific computing. Mutual collaborations will be used to synthesize expertise across three NSF-supported Natural Hazards Engineering Research Infrastructure (NHERI) facilities, including the Computational Modeling and Simulation Center at the University of California, Berkeley, the experimental facility at Lehigh University, and the DesignSafe cyberinfrastructure at the University of Texas at Austin. Project data will be archived and made publicly available in the NHERI Data Depot (https://www.DesignSafe-ci.org).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.