Machine Learning for Lithology Percentages Prediction from Cutting Photos: Massive Validation, Results, and Challenges

Drill cutting samples are valuable data that cover the major part of drilled well compared to the core samples that cover only a limited depth interval. Therefore, accurate and objective cutting description plays major role in decision making while drilling, in the reservoir characterization studies, and in modeling workflows. We developed an Artificial Intelligence workflow to automatically predict cutting lithology percentages using cutting photos. The workflow can be applied in near-real-time as soon as photos are acquired.