Modeling & Intelligent Control of a Drilling-String System of Oil & Gas Wells

Abstract

This dissertation tackles the persistent challenge of Stick-Slip Oscillations (SSO) in drilling systems by developing an intelligent control framework that combines advanced modeling, optimal control, and machine learning. The study begins with a review of drilling dynamics and SSO suppression methods, identifying limitations in conventional approaches. A high-fidelity 4-Degree-Of-Freedom (4-DOF) torsional model is then developed to accurately capture SSO behavior, followed by the design of a Linear Quadratic Regulator (LQR) enhanced with integral action to eliminate steady-state errors. To address practical implementation challenges, a Luenberger observer is integrated for real-time state estimation. The control strategy is further refined using computational intelligence techniques. Genetic Algorithms (GA) optimize the LQR weighting matrices to improve performance across diverse operating conditions, while an Artificial Neural Network (ANN) is trained to dynamically adjust control parameters in response to down-hole uncertainties. Simulation results demonstrate significant improvements in SSO suppression, with the ANN-adaptive LQR outperforming both conventional and GA-optimized LQR controllers, particularly under unpredictable disturbances and different operating conditions of the drilling system. This work provides a robust, adaptive solution for mitigating torsional vibrations in drilling operations, bridging the gap between theoretical control design and real-world applicability.