Intelligent Solution for Auto-Detecting Lithology Scheme While Drilling by Machine Learning

Abstract

Abstract Predicting the lithology type of drilled formation is a critical task in the drilling operations as it significantly affects the drilling program and the drilling operation's total cost. Hence, it is highly recommended to detect the lithology variation while drilling to be able to optimize the drilling parameters based on the penetrated lithology type. Currently, the lithology changes are estimated from the geological data and logs which are considered as operations limitations and drawbacks of this approach as cost and lag time. The current study proposes an intelligent machine learning solution for auto-detecting the formation tops and lithology types of drilled formations while drilling in real-time utilizing drilling surface data. Machine learning techniques are technically employed for developing real-time prediction models for the drilled rock lithology from the surface rig sensor data as weight on bit, drillstring speed, hook load, mud pumping rate, torque, pumping pressure and rate, and rate of penetration as model input data to predict the drilled lithology class. Different ML techniques Decision Tree, K Neighbors Classifier, and Bagging Classifier were tested through the methodology to assess the computational power for classifying and auto-detecting the drilled lithology while drilling by feeding real-time drilling data to the models. The data set represents a complex lithology of five different drilled formations (dolomite formation, anhydrite, dolomitic limestone composition, limestone, and shale formation), while the dataset is utilized for training and testing purposes. The methodology approach presents a comprehensive layout for data collection, preprocessing, data statistics and analytics, feature engineering, model development and parameters optimization, and prediction performance evaluation. The results showed a high prediction performance for the models for training and testing with an overall accuracy higher than 98 through detecting complex lithology schemes. Predicting the drilled formation's tops and lithology while drilling in real-time will provide a technical guide for optimizing the drilling parameters for better drilling performance and optimized mechanical-specific energy.