Predicting Swab and Surge Pressures for Wellbore Instability Analysis Using Machine

Abstract

This study applies machine learning (ML) techniques to predict surge and swab pressures during tripping operations with the aim of enhancing wellbore stability in drilling environments. Using real operational data obtained from the Sponge Bob 1 well, a Water-Based Mud (WBM) system, key drilling parameters such as mud weight, plastic viscosity, yield point, and true vertical depth were collected, processed, and engineered into a structured dataset of 70 observations. Five supervised ML models—Artificial Neural Network (ANN), Random Forest, XGBoost, Decision Tree, and K-Nearest Neighbors—were developed and evaluated using Root Mean Squared Error (RMSE) and Coefficient of Determination (R²). Results showed that ANN provided the highest predictive accuracy with RMSE values of 5,666.42 psi and 3,904.43 psi, and R² values of 0.89 and 0.91 for surge and swab pressures, respectively. The analysis confirmed mud rheology and depth as the most influential predictors of transient pressure behavior, while graphical comparisons demonstrated strong alignment between actual and predicted values. The study concludes that ML offers a robust, data-driven alternative to conventional hydraulic models and recommends its integration into real-time drilling systems to support proactive wellbore pressure management and reduce non-productive time.