Collaborative Research: 3D Ambient Noise Tomography (3D ANT) for Natural Hazards Engineering

尽管在医学成像方面取得了重大进展，但用于基础设施工程的地下成像远远落后。例如，许多工程分析仍然基于地下的一维剖面，或由几个一维测深构造的伪2D/3D剖面。当进行真正的3D成像时，勘探的深度和分辨率往往是有限的。虽然地下成像的问题相当复杂，但开发快速、逼真的地下三维图像的能力，以及伴随的工程特性(例如剪切模数)，将极大地促进工程设计，以实现更具弹性和可持续的基础设施。这项研究旨在开发一种新的3D地下成像方法，使用从表面传感器网格获得的环境噪声记录。新的3D环境噪声层析成像(3D ANT)方法将提供一种快速、非侵入性、健壮的方法，在地下表面顶部50至100米的m尺度上以3D方式成像。虽然基础设施工程中存在许多用于精确和深层3D地下成像的示例应用，但该项目将具体满足与自然灾害相关的两个需求：(1)开发用于地震地面运动研究的真实3D地下模型的需求，以及(2)用于异常(例如，空洞/天坑)检测的改进3D原位成像的需求。此外，开发更深层次、更高分辨率的3D地下成像方法将为社会带来预期和意想不到的重大和广泛的好处。利用环境噪声观察地球内部和检索快速可靠的模型的能力将影响到各种领域，如自然资源勘探、地下水文学、纯地球科学、考古学、地下开发、军事/安全研究和空间/行星探索。这项研究的智力优势在于验证了这样一个假设，即可以从地面环境噪声记录中提取出m尺度下精确的三维地下P波和S波速度模型，深度可达50至100米。为了验证这一假设，该研究将开发一种新的3D蚂蚁方法，并通过数值模拟和现场实验来验证该方法。使用环境噪声进行3D地下成像提出了与环境噪声的不可控频率、含量和传播方向相关的固有挑战。然而，环境噪声含有丰富的低频能量，允许比目前使用有源3D全波形反演(FWI)方法进行更深层次的成像。因此，当3D蚂蚁与有源3D FWI相结合时，将为目前无法获得的深度提供高分辨率图像。3D蚂蚁方法将需要从近距离表面传感器的2D网格收集环境噪声记录。噪声记录将被用来提取每对可能的传感器之间的实验关联函数。然后利用三维粘弹性波动方程得到综合相关函数，并将其与实验结果进行匹配，采用高斯-牛顿快速傅立叶变换提取三维地下模型。3D蚂蚁算法的优化将包括对现场测试配置(即传感器的数量和间距)和环境噪声特性(即频率含量和方位分布)的参数研究。最终，选择了两个具有良好特征的具有地面真实情况的现场，在现实世界的条件下测试该方法。只有当我们成功地培训学生将其带到未来，传播结果，并开发出实践者可以在行业中使用的工具时，这项研究才会产生最广泛的影响。我们的目标是解决这些挑战，同时扩大妇女对自然灾害工程的参与。灾害工程涉及改变民用基础设施的设计和修复方式，使社区和个人更能抵御自然灾害的破坏性影响，研究表明，当STEMM领域具有直接的社会影响时，妇女对这些领域更感兴趣。这项工作将帮助培养一群多样化和有才华的未来工程师的兴趣和目标的自然交集，他们的技能将对自然灾害工程的革命性至关重要，如编码、分析大数据和超级计算。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Despite significant progress in medical imaging, subsurface imaging for infrastructure engineering lags far behind. For example, many engineering analyses are still based on 1D profiles of the subsurface, or pseudo-2D/3D profiles constructed from several 1D soundings. When true 3D imaging is performed, the depth and resolution of exploration is often limited. While the problem of subsurface imaging is quite complex, the ability to develop rapid, realistic, 3D images of the subsurface, with accompanying engineering properties (e.g., shear modulus), would significantly advance engineering for more resilient and sustainable infrastructure. This research aims to develop a new 3D subsurface imaging method using recordings of ambient noise obtained from a grid of surface sensors. The new 3D Ambient Noise Tomography (3D ANT) method will provide a rapid, non-intrusive, robust way of imaging the subsurface in 3D at m-scales over the top 50- to 100-m of the subsurface. While numerous example applications for accurate and deep 3D subsurface imaging exist within infrastructure engineering, this project will specifically address two needs related to natural hazards: (1) the need for developing realistic 3D subsurface models for use in earthquake ground motion studies, and (2) the need for improved 3D in-situ imaging for anomaly (e.g., void/sinkhole) detection. Furthermore, significant and broad benefits for society, both anticipated and unanticipated, will result from developing deeper, higher-resolution 3D subsurface imaging methods. The ability to look inside the earth and retrieve rapid and reliable models using ambient noise will impact fields as diverse as: natural resource exploration, subsurface hydrology, pure earth science, archeology, underground development, military/security studies, and space/planet exploration. The intellectual merit of this research center around testing the hypothesis that accurate 3D subsurface P- and S-wave velocity models can be extracted at m-scales down to 50- to 100-m depth from surface recordings of ambient noise. To test this hypothesis, the research will develop a novel 3D ANT method and verify the method with numerical simulations and field experiments. The use of ambient noise for 3D subsurface imaging presents inherent challenges related to the uncontrollable frequency content and propagation direction of ambient noise. However, ambient noise is rich in low frequency energy, allowing for deeper imaging than what is currently possible using active-source 3D full waveform inversion (FWI) methods. Hence, 3D ANT, when coupled with active-source 3D FWI, will provide high resolution images to depths presently unobtainable. The 3D ANT method will require collecting ambient noise recordings from a 2D grid of closely spaced surface sensors. The noise recordings will be used to extract experimental correlation functions between every possible pair of sensors. 3D viscoelastic wave equations will then be used to obtain synthetic correlation functions, which will be matched with the experimental ones using a Gauss-Newton FWI approach for extracting 3D subsurface models. Optimization of the 3D ANT algorithm will include parametric studies on field testing configurations (i.e., number and spacing of sensors) and ambient noise characteristics (i.e., frequency content and azimuthal distribution). Ultimately, two well-characterized field sites with ground truth have been selected to test the methodology under real-world conditions. This research will only achieve its broadest impact if we are successful at training students to carry it into the future, disseminating results, and developing tools that practitioners can use in industry. We aim to tackle these challenges while simultaneously broadening the participation of women in natural hazards engineering. Hazards engineering is about transforming how civil infrastructure can be designed and rehabilitated such that communities and individuals are more resilient to the devastating effects of natural hazards, and studies have shown that women are more interested in STEMM fields when they have direct societal impacts. This work will help foster a natural intersection of interest and purpose in a diverse and talented group of future engineers with skills that will be critical for revolutionizing natural hazards engineering, such as coding, analyzing big-data, and supercomputing.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.