Data-Driven Micromechanical Characterization and Mapping of Shale Rocks Using High Speed Nanoindentation
This study investigates the potential of high-speed nanoindentation in collaboration with data analytics and phase volume fractions to achieve micromechanical characterization of heterogeneous rocks. While micromechanical characterization can be performed using mechanical testing alone, integrating chemical analysis-such as elemental mapping techniques-provides essential phase identification. This enables more accurate interpretation of phase-specific mechanical properties. However, incorporating chemical analysis increases the complexity of the process. Hence, this study proposes data-driven micromechanical characterization and mapping of heterogeneous rocks based primarily on mechanical data and limited dependence on chemical analysis. In this study, Mancos shale rock is analyzed using high-speed nanoindentation to determine the mechanical properties at the microscale. Subsequently, a suite of unsupervised statistical learning techniques, such as Uniform Manifold Approximation and Projection (UMAP) with k-means Clustering, Gaussian Mixture Model (GMM), Dirichlet Process Mixture Model (DPMM), and Density-Based Spatial Clustering of Applications with Noise (DBSCAN), are applied to the nanoindentation data. Additionally, an automated image processing and segmentation technique was developed and tested. The results from each technique have been systematically compared against the conventional chemo-mechanical approach using two metrics: weighted error and spatial error. Based on the results, UMAP with k-means clustering is the most appropriate technique, while DBSCAN, DPMM, and image segmentation techniques are more suitable as secondary approaches. This study demonstrates the capability of high-speed nanoindentation combined with machine learning techniques for micromechanical characterization with reduced analytical complexity and improved workflow efficiency.