An Approach for the Classification of Rock Types Using Machine Learning of Core and Log Data

Classifying rocks based on core data is the most common method used by geologists. However, due to factors such as drilling costs, it impossible obtain samples from all wells, which poses challenges for accurate identification of rocks. In this study, authors demonstrated application an explainable machine-learning workflow using and log identify rock types. The type determined utilizing flow zone index (FZI) first, then collection, collation, cleaning well data, four supervised learning techniques were correlate with types, prediction models constructed. optimal machine algorithm classification selected a 10-fold cross-test comparison AUC (area under curve) values. accuracy rate results indicates that proposed can greatly improve SHapley Additive exPlanations (SHAP) was rank importance various logs input variables types provides both local global sensitivities, enabling interpretation solving “black box” problem associated algorithms. study reliably predict solve hard problems in geological research. Furthermore, provide consistent arising lack while providing powerful tool trajectory optimization. Finally, system aid selection intervals be completed and/or perforated.