PREDICTING THE WEIGHT AND TYPE OF DRILLING MUD BY MACHINE LEARNING METHOD

Selecting the optimal drilling fluid, defined by its weight and chemical type, is critical for preventing costly wellbore instability and catastrophic accidents. Traditional methods often rely on trial-and-error, past experience or simplified models that fail to capture the complex rock-fluid interactions. While data mining offers a promising alternative, a research gap exists in simultaneously predicting both mud weight and type. This study introduces a novel machine learning framework that concurrently predicts these essential properties. Utilizing a comprehensive dataset extracted from 50 years of daily drilling reports across 20 oil wells, we trained and compared three nature-inspired algorithms: Ant Colony (ACO), Bee Colony (BCO), and Emperor Penguins Colony (EPC) optimization. The results demonstrate that all models achieved high predictive accuracy, with the Bee Colony Optimization (BCO) algorithm emerging as the most precise, yielding a correlation coefficient (R²) of 0.9841 and a root-mean-square error (RMSE) of 0.0245. Furthermore, sensitivity analysis revealed that the Rate of Penetration (ROP) is the most influential parameter on mud properties, surpassing other drillin