Abstract
Drilling fluid viscosity is a critical parameter influencing the efficiency and safety of drilling operations, yet real-time monitoring under dynamic conditions remains challenging due to varying temperature, pressure, and flow rates. This study develops algorithms for continuous, automated viscosity monitoring of drilling fluids to optimize drilling processes. The primary objective is to design robust algorithms capable of accurate viscosity predictions in real-time, addressing limitations of traditional manual methods like the Fann 35 viscometer, which are labor-intensive and fail to capture rapid rheological changes. The methodology involves testing water-based and oil-based muds using inline viscometers, pressure transducers, and temperature sensors. Signal processing techniques, including low-pass filtering, and machine learning models, such as linear regression and neural networks, were developed to process sensor data and predict viscosity. The neural network model achieved accuracy (RMSE of 0.42 cP for water-based muds, 0.67 cP for oil-based muds in lab tests) compared to linear regression (RMSE of 0.85 cP and 1.12 cP). Tests confirmed robustness, with the neural network maintaining an RMSE of 0.58 cP and 0.79 cP. Statistical analysis using ANOVA and regression validated the results, demonstrating significant improvements over existing methods. The system enables real-time adjustments, enhancing drilling efficiency and safety. Future work includes improving algorithm adaptability and exploring edge computing for remote applications.