Wellbore Instability Mechanism Evaluation by Computer Vision on Cavings

Summary Geomechanics is basic to ensure safer drilling operations. This study focusses on predicting wellbore instability conditions based on the caving texture while drillling. Both Artificial Intelligence and Computer Vision are used to automatize the process of caving recognition and classification. The textural and geometrical information extracted with Computer Vision is used to train a model that reported 68.85% accuracy when classifying cavings into the categories tabular, blocky and splinter. This tool has potential to be deployed on site, representing a cost-effective approach to anticipate wellbore instability in real time and make better decisions while drilling.