Research

Understanding the cracking behavior (i.e. initiation, propagation, and coalescence) of rocks in the subsurface is important for rock engineering such as enhanced geothermal systems (EGS). More specifically, in EGS, hydraulic fracturing is used to enhance the permeability of the often tight crystalline rocks in geothermal reservoirs by creating a system of open and well-connected fracture networks. However, in the field, one can rarely directly observe the cracking processes and must rely on indirect information such as micro-seismicity induced by the cracking processes. It is often challenging to infer the cracking behavior from micro-seismic data given the complex conditions (e.g. stress condition, lithology) that may affect their relations. As an experimentalist, I make used of well-controlled experimental tests to understand the cracking mechanisms and their relations with lab-scale microsesimicity, AE of rocks under different conditions. My research provides the basis for interpreting the microseismicity induced by rock cracking in these rock engineering projects.

A comprehensive study on the nucleation, growth, and interaction of microcracks (i.e. microcracking) and the associated fracture process zones (FPZs) of rocks is crucial for better understanding and predicting subsequent macrocracking processes. My doctoral project focused on the microcracking mechanisms of granite using a multi-scale crack characterization method integrating AE and microscopic observation techniques such as scanning electron microscopy and petrographic microscopy. Since granite is a common host rock in EGS geological disposal of nuclear waste, my research could contribute to managing the efficacy of EGS or the safety of geological disposal of nuclear waste.

Acoustic emission (AE) technique is a powerful tool for monitoring microcracking and microdamage in geomaterials such as rocks and cement. It finds broad application across disciplines, including Geotechnical Engineering, Geoscience, and Structural Engineering. While AI is revolutionizing seismic data analysis, its application to the lab-scale counterparts, namely AE remains relatively limited. The main challenge lies in the limited availability of labeled data—manually selected, picked, and weighted waveforms—within the AE community compared to the vast labeled datasets available in seismology. We created the Stanford Acoustic Emission Database (SAED), which includes ~ 50,000 manually picked waveforms. To our knowledge, SAED is the largest open-labeled dataset in the AE field. Based on the SAED, we trained and tested AE-PNet, a deep-learning model designed to pick P-wave arrivals both accurately and efficiently. Moving forward, our research will focus on extending AI applications to process continuous AE data for tasks such as denoising, P-phase picking, and event association. Continuous AE data provides deeper insights into microcracking mechanisms but presents significant challenges for traditional algorithms. In short, an AI-powered analysis workflow promises to enhance the reliability of AE results, unlocking the AE technique’s full potential.