EAPSI: Improving Computer Simulations of Earthquakes

EAPSI：改进地震的计算机模拟

• 批准号: 1414950
• 负责人: Brian Skjerven
• 金额: $ 0.51万
• 依托单位: Skjerven Brian
• 依托单位国家: 美国
• 项目类别: Fellowship Award
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-06-01 至 2015-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1414950&HistoricalAwards=false
• 关键词: EAPSI; Improving; Computer; Simulations; Earthquakes

Predicting earthquakes is a difficult task, due largely to a limited understanding of the complex physical processes involved. Computer simulations allow researchers to better understand such processes, but the current models are limited in terms of the problem size that they can handle. Improving these computer models could provide insight into seismic events, such as earthquakes, and possibly aid in prediction. This project will focus on improving an existing software package for modeling Episodic Tremor and Slip (ETS) events. This research will be conducted at the Chinese Academy of Sciences in collaboration with Dr. Zongyan Cao, an expert with the Xeon Phi Coprocessor, Intel's new chip that will be used in the computer simulations.ETS events occur where continental plates meet, and they are characterized by weak, long-term tremors with acute seismic events occurring periodically. A novel mathematical model has been proposed to describe ETS, and subsequent computer simulations closely match observed GPS data. This model describes a system of shearing and fluid release. As two plates slide along each other, heat is produced, ultimately causing a dehydration reaction to occur. The result is a release of high-pressure fluid, which acts as a lubricant, resulting in a slip event. Additionally, the fluid absorbs heat, and the whole system cools and returns to it's original state. The code for this chemo-mechanical oscillator model is limited in size and scale; it currently only runs in one dimension and in serial. This research seeks to develop a parallel version of the code that is capable of handling multiple dimensions and multiple time and length scales. Intel's Xeon Phi Coprocessor will be used, in part due to its ease of programmability, but also because the next generation of HPC systems will make use of accelerator technologies like the Xeon Phi. This NSF EAPSI award is funded in collaboration with the Chinese Ministry of Science and Technology.

预测地震是一项艰巨的任务，这在很大程度上是由于对所涉及的复杂物理过程的有限理解。计算机模拟使研究人员能够更好地理解此类过程，但是当前模型在可以处理的问题大小方面受到限制。 改善这些计算机模型可以提供对地震事件（例如地震）的见解，并可能有助于预测。该项目将着重于改进用于建模情节震颤和滑移（ETS）事件的现有软件包。这项研究将在中国科学院与Xeon Phi Phi Pocrosessor的专家Zongyan Cao博士合作，该专家是Intel的新芯片，该芯片将用于计算机模拟中。SETS事件发生在大陆板相遇的情况下，其特征是弱，长期的震颤，急性地震事件发生了定期。已经提出了一个新型的数学模型来描述ETS，随后的计算机模拟与观察到的GPS数据密切匹配。该模型描述了剪切和流体释放的系统。当两个板相互滑动时，产生热量，最终导致脱水反应发生。结果是释放高压流体，该液充当润滑剂，导致滑动事件。此外，流体会吸收热量，整个系统冷却并返回其原始状态。这种化学机械振荡器模型的代码尺寸和规模有限；目前，它仅在一个维度和串行中运行。这项研究旨在开发能够处理多个维度以及多个时间和长度尺度的代码的并行版本。 Intel的Xeon Phi协处理器将被使用，部分原因是它的可编程性易用性，而且还因为下一代HPC系统将利用诸如Xeon Phi之类的加速器技术。该NSF EAPSI奖与中国科学技术部合作。