Smart Observational Methods for Geomechanical Systems

风险是工程系统中不可避免的组成部分，尤其与涉及土壤和岩石的结构有关。众所周知，这些材料在所有尺度上都是不均匀的--从矿物颗粒接触处的微米级强度对比，一直到千米级的地震断裂带。土工结构面临其他外部风险因素，如气候变化和地震破坏，必须加以考虑。其中一些不确定性可以通过一致的监控来管理，以检测和表征性能变化。然而，现有的长期监测战略可能代价高昂，而且在时间和空间分辨率上受到限制。该研究方案的长期目标是开发具有成本效益和高分辨率的工具，用于表征和监测各种地质结构。该研究计划将通过使用静止和航空成像以及三维数字图像相关来实现广泛的空间覆盖，以1)校准斜坡的材料和几何特性，以及2)开发有监督的机器学习方法来处理这些图像数据并预测斜坡的不稳定性。这些工具将首先在西部大学最先进的滚筒离心机的斜坡模型上开发，然后在安大略省南部一个采石场的无人机图像上进行验证和更新。由此产生的方法预计将帮助土木工程和采矿工程师确定大面积斜坡的特征，并确定高速公路、大坝、自然资源和靠近渐进式滑坡的建筑物。该研究计划将利用对高频弹性波敏感的声发射传感器来检测、定位微米级的剪切和拉伸裂纹并对其进行分类，从而实现高时间分辨率的监测。这些微裂纹持续时间短，因此不能用传统的载荷、应变或孔压传感器检测到，因为这些传感器只对较低的频率扰动敏感。我们将开发机器学习方法，通过分析环形剪切实验中的声发射信号特征来识别和分类流体-土壤-岩石系统的变化，以及4)在缩尺离心机模型中研究地震表面破裂引起的离散声发射事件。这些声发射工具可以指导土木和能源工程师诊断水电大坝或地下核废料储存等基础设施项目中有问题的裂缝和泄漏，并优化地热系统中的流动网络。这些实验还将促进我们对地震断层和破裂机制的基础科学理解。我们在环形剪切装置中的流体-土壤-岩石实验将揭示富含流体的断裂带的频率特征，我们的离心机实验将揭示当大地震传播到地面时应力局部化的细节。

Risk is an unavoidable component of engineered systems and is particularly relevant to structures involving soil and rock. These materials are notoriously heterogeneous at all scales - from micrometer-scale strength contrasts at mineral grain contacts, all the way up to kilometer-scale earthquake fault zones. Geostructures face additional external risk factors such as climate change and earthquake damage that must be considered. Some of these uncertainties can be managed with consistent monitoring to detect and characterise changes in performance. However, existing long-term monitoring strategies can be costly and limited in temporal and spatial resolution. The long-term goal of the research program is to develop cost-effective and high-resolution tools for the characterisation and monitoring of various geostructures. The research program will achieve broad spatial coverage by employing stationary and aerial imaging in conjunction with three-dimensional digital image correlation to 1) calibrate for material and geometric properties of slopes, and 2) develop supervised machine learning methods to process these image data and predict slope instability. These tools will first be developed on scale models of slopes in Western University's state-of-the-art drum centrifuge, and then validated and updated on drone images of a quarry in southern Ontario. The resulting methodologies are expected to assist civil and mining engineers in characterising slopes over large areas and identify highways, dams, natural resources, and buildings in proximity to progressive landslides. The research program will achieve high temporal resolution monitoring by employing acoustic emission sensors, which are sensitive to high frequency elastic waves, to detect, locate, and classify micrometer-scale shear and tensile cracks. These micro-cracks have short durations and thus cannot be detected with conventional load, strain, or pore-pressure transducers which are only sensitive to lower frequency perturbations. We will 3) develop machine learning methods to identify and classify changes in fluid-soil-rock systems by analysing acoustic emission signal characteristics during ring-shear experiments, and 4) investigate discrete acoustic emission events originating from surface ruptures of earthquakes in scaled centrifuge models. These acoustic emission tools can guide civil and energy resource engineers in diagnosing problematic cracks and leaks in infrastructure projects such as hydroelectric dams or underground nuclear waste storage, and optimise flow networks in geothermal systems. The experiments will also advance our fundamental science understanding of earthquake faulting and cracking mechanisms. Our fluid-soil-rock experiments in the ring shear device will shed light on the frequency characteristics of fluid-rich fault zones, and our centrifuge experiments will reveal details of stress localisation when large earthquakes propagate to the ground surface.