EAGER: Exploration of an Interdisciplinary Approach to Resolving a Critical Issue in Evaluating Liquefaction Hazard of Challenging Soil Sites

这个探索性研究（EAGER）项目侧重于开发计算方法，以提高用于研究具有挑战性的土壤地点地震诱发液化风险的数据的准确性和分辨率。液化是松散的沙质土壤在地震中失去强度的过程，造成严重的破坏。这项研究将通过改进目前的液化危害评估方法，帮助优化减轻液化风险的努力，从而为全世界的整个社会带来利益。具体的重点将是提高锥贯入测试（CPT）测深数据的准确性和分辨率，这是岩土工程师开发土壤特性深度剖面的常用方法。CPT测深提供了一种被称为贯深计的尖端仪器在被推入地下时所经历的尖端阻力和侧面摩擦的测量剖面。然而，所测得的阻力和摩擦力代表了穿透器周围土壤体积的影响。因此，CPT数据被测量过程的物理原理“平滑”或“模糊”。当工程师试图定位薄层土壤时，这是一个问题，这可能是“平滑”的，但这对理解液化机制和风险至关重要。来自成像科学的先进计算模型和优化技术将用于“去模糊”这些数据。如果成功，这项研究将产生强大而高效的计算框架和相关软件，用于校正数据以解析薄的次表层。分辨率和可靠性的提高将使岩土工程师能够识别以前无法识别的地下薄层。这将代表一种重要的方法进步，使研究薄层在液化中的基本作用成为可能。这项研究的结果将与该专业直接相关，并且通过共同pi参与弗吉尼亚理工大学岩土工程研究与实践中心，将加快该专业对研究成果的采用。该跨学科项目将支持弗吉尼亚理工大学的两名研究生研究人员，一名研究数学，一名研究土木工程。共同负责人R. Green为退伍军人建立了一个外展项目，迄今为止已经在他的研究小组中招募了三名退伍军人。PI和copi将利用该项目进一步努力与退伍军人合作。在2010-2011年新西兰坎特伯雷地震序列（CES）期间，由砂质土壤组成的淤泥和粘土层互层地表液化表现的预测与观测严重程度的比较，突出了目前使用的液化评估程序的重大局限性。一个潜在的问题是CPT在识别和正确表征影响整个剖面液化响应的薄层方面的局限性。多层互层对测量的CPT尖端阻力和套筒摩擦具有“平滑”作用，导致砂层密度严重低估，细粒层刚度高估。虽然有建议纠正CPT尖端阻力的“薄层效果”的程序，但大多数这些程序都是手动的，不适用于多个薄层效果。最近程序开发占多个薄层的影响,提出这是一个反问题,假设测量CPT等于“真正”的CPT数据与与深度有关的空间滤波器卷积后一个简单的一维模型(雅德琼,2018),但我们的分析表明本程序不能正确的对薄层的规模利益(例如,几厘米或薄层),通常是缓慢收敛,如果达到收敛。此外，更重要的是，当应用于来自CES的液化案例历史的大型数据库时，总的来说，该程序产生的预测比没有对CPT测深应用薄层校正的预测更不准确。该项目的目的是探索一种跨学科的方法来解决这一关键问题，以评估具有挑战性的土壤场地（即具有多个薄互层不可液化土壤的砂土剖面）的液化危害。将开发一种基于总变分（TV）最小化方法的可靠的、可扩展的计算技术，以反演这些站点的“真实”锥突测试（CPT）测深数据（即，校正多个薄层效应）。为了评估所开发的反演算法的有效性，将使用物质点法（MPM）来模拟在具有一系列感兴趣特征的具有挑战性的土壤剖面中执行的CPT；MPM模型将首先根据校准室测试数据进行验证。这将使我们知道？真的？和测量?多互层剖面CPT测深数据。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This EArly-concept Grant for Exploratory Research (EAGER) project focuses on developing computational methods to improve the accuracy and resolution of data used to study earthquake-induced liquefaction risk at challenging soil sites. Liquefaction is the process by which loose, sandy soils lose their strength during earthquake shaking, causing significant damage. This research will help optimize efforts to mitigate liquefaction risk by improving the current methods for liquefaction hazard assessment, resulting in a benefit for society at large, worldwide. The specific focus will be improving the accuracy and resolution of data from cone penetration test (CPT) soundings, a common method for geotechnical engineers to develop a depth profile of the soil properties. CPT soundings provide measured profiles of tip resistance and side friction experienced by a pointed instrument called a penetrometer as it is pushed into the ground. However, the measured resistance and friction represents the influence of a volume of soil surrounding the penetrometer. Thus, CPT data are "smoothed" or "blurred" by the physics of the measurement procedure. This is problematic when engineers attempt to locate thin layers of soil, which might be "smoothed out," but that are critical to understanding liquefaction mechanisms and risk. Advanced computational models and optimization techniques from imaging science will be used to "deblur" these data. If successful, this research will result in robust and efficient computational framework and associated software that corrects data to resolve thin subsurface layers. The improvements in resolution and reliability will ideally allow geotechnical engineers to identify thin layers in the subsurface that were not previously identifiable. This will represent a significant methodological advance, enabling studies of the fundamental role of thin layers in liquefaction. The results from this research will be of direct interest to the profession, and the adoption of the research findings by the profession will be expedited by the Co-PIs' involvement in the Center for Geotechnical Research and Practice at Virginia Tech. This interdisciplinary project will support two Virginia Tech graduate student researchers, one in mathematics and one in civil engineering. Co-PI R. Green has established an outreach program for military veterans and has to date recruited three veterans in his research group. The PI and CoPIs will use this project to further this effort in working with veterans. Comparison of predicted versus observed severity of surficial liquefaction manifestations at sites comprised of sandy soils with interbedded silt and clay layers during the 2010-2011 Canterbury, New Zealand earthquake sequence (CES) highlights significant limitations of currently used liquefaction evaluation procedures. One potential issue is the limitation of CPT to identify and properly characterize thin layers that impact the liquefaction response of the entire profile. Multiple interbedded layers have a "smoothing" effect on the measured CPT tip resistance and sleeve friction, resulting in significant underestimation of the density of sand layers and an overestimation of the stiffness of fine-grained layers. While procedures have been suggested to correct CPT tip resistance for "thin layer effects," most of these procedures are manual and are not applicable for multiple thin layer effects. A recent procedure was developed to account for multiple thin layer effects by posing it as an inverse problem, assuming the measured CPT data equal the "true" CPT data convolved with a depth-dependent spatial filter following a simple 1D model (Boulanger and DeJong, 2018), but our analysis indicates this procedure cannot correct for thin layers at the scale of interest (e.g., layers of a few centimeters or thinner), and is often slow to converge, if convergence is achieved at all. Furthermore and more significantly, when applied to a large database of liquefaction case histories from the CES, the procedure yields, in totality, less accurate predictions than if no thin layer corrections were applied to the CPT soundings. The objective of this project is to explore an interdisciplinary approach to resolving this critical issue in evaluating liquefaction hazard of challenging soil sites (i.e., sand soil profiles having multiple, thin interbedded layers of non-liquefiable soil). A robust, computationally scalable technique will be developed to invert for "true" Cone Penetration Test (CPT) sounding data (i.e., correct for multiple thin layer effects) for these sites that is based on the total variational (TV) minimization method. To assess the efficacy of the developed inversion algorithm, the Material Point Method (MPM) will be used to simulate CPT performed in challenging soil profiles that have a range of characteristics of interest; the MPM model will first be validated against calibration chamber test data. This will allow us to know both the ?true? and ?measured? CPT sounding data for the multiple interbedded layer profiles.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.