Identification of Rock Crack Damage Regions using Machine Learning-data

Dentification of macrocracks and their evolution under external loads is pivotal in studying rock failure. The formation of macrocracks can be attributed to the progression of microcracks. One significant challenge in studying microcracks is their location within the rock, rendering direct observation infeasible. This paper presents a methodology to characterize hypothetical damage regions formed by microcracks using Acoustic Emission (AE) data. Initially, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm was employed to filter out less significant AE events, thereby pinpointing critical clusters of microcracks. The spatial distribution of microcracks was then modeled using a Gaussian Mixture Model (GMM), with the Expectation-Maximization (EM) algorithm utilized to compute the GMM components. Confidence regions of the Gaussian distributions were calculated to delineate the damage regions. The results demonstrated that this approach effectively identifies damage regions associated with macrocracks. The generated damage regions accurately indicate the location and orientation of macrocracks. This investigation enhances the understanding of the distribution of microcracks and their relation to macrocracks. Moreover, this approach can be considered an automated method for quantitatively assessing cracks.宏观裂纹的识别及其在外载荷作用下的演化规律，是岩石破坏研究的关键课题。宏观裂纹的形成可归因于微裂纹的扩展。当前微裂纹研究面临的一项显著挑战在于其埋藏于岩石内部，无法直接观测。本文提出了一种基于声发射（Acoustic Emission, AE）数据表征微裂纹所致假想损伤区域的方法。首先，采用基于密度的带噪声应用空间聚类（Density-Based Spatial Clustering of Applications with Noise, DBSCAN）算法滤除显著性较低的声发射事件，进而精准定位微裂纹的关键聚类簇；随后，利用高斯混合模型（Gaussian Mixture Model, GMM）对微裂纹的空间分布进行建模，并通过期望最大化（Expectation-Maximization, EM）算法求解高斯混合模型的组分参数；进一步计算高斯分布的置信区域以划定损伤区域。研究结果表明，该方法可有效识别与宏观裂纹相关的损伤区域，所生成的损伤区域能够精准反映宏观裂纹的位置与取向。本研究深化了对微裂纹分布规律及其与宏观裂纹关联机制的认知，此外，该方法亦可作为一种自动化的裂纹定量评估手段。