Automated Lithology Analysis on HyLogger Data of Drill Cuttings with Multi-modal Machine Learning

Drill cuttings are small fragments of rock that are formed by the action of a bit in rotary drilling of a wellbore. The common practice of mudlogging, to identify the lithology from cuttings, is time consuming and subjective.

HyLogger-3, an automated drill cutting and core profiling system, provides both optical images and mineral information. In this study, a multi-modal machine learning (ML) approach is developed to automatically predict lithology from cuttings, incorporating both visual and mineral information from the HyLogger-3 dataset. The performance of the multi-modal ML model on blind test data demonstrates the robustness and accuracy of this approach.