Enhanced loss of circulation prediction using Sea Lion-optimized machine learning models

Lost circulation is a major drilling challenge that increases cost and non-productive time and is influenced by different parameters. This study develops and evaluates machine learning models for predicting lost circulation severity during drilling operations. Three machine learning algorithms, Decision Tree, K-Nearest Neighbors (K-NN), and Random Forest, were optimized using the Sea Lion Optimization algorithm to enhance prediction accuracy. Feature importance analysis identified azimuth as the most significant predictor, followed by measure depth, average bit RPM, hook load, and torque, indicating that directional and mechanical drilling parameters outweigh mud properties and formation data in predicting fluid loss. Correlation analysis further confirmed strong relationships among these key parameters. The results indicate that azimuth and depth have a stronger correlation with the target parameters. All optimized models demonstrated good performance in classifying fluid loss severity. The Random Forest model achieved the best performance among the evaluated models. The K-NN and Decision Tree models also showed strong performance. Confusion matrix analysis validated all models' capability to reliably distinguish between different fluid loss severity categories (Seepage, Partial, Severe, and Total Loss) on both training and testing datasets. The findings of this study can help to provide an effective approach for predicting lost circulation severity, offering potential for reducing non-productive time, cost, and improving drilling safety.