Development of an AI Rock Type Classification and Fracture State Characterization Web-based Application

Conventional rock core logging and log checking methods, while dependable, can be labor-intensive and subjective, possibly prone to errors and inconsistencies. This project addresses these limitations by employing computer vision-based deep learning AI technology to improve the accuracy, objectivity, and productivity of rock core logging and log checking processes. A web-based application called PreLogging was developed, capable of classifying common Hong Kong rocks and decomposition grades, identifying fractures, and computing fracture state indices such as Total Core Recovery (TCR), Solid Core Recovery (SCR), Rock Quality Designation (RQD), and Fracture Index (FI). The development of PreLogging involved a multidisciplinary team of researchers, engineers, and industry stakeholders collaborating to produce a tool that addresses the needs of geotechnical engineering. The development process encompassed multiple stages, including the design of a user interface, creation of an alpha version, soliciting feedback from industry stakeholders, incorporating AI computation capabilities, and releasing a stable version. The project team found that collecting feedback from industry stakeholders was crucial in ensuring the application effectively meets the industry's requirements.