Drilling through shales with higher clay content is highly challenging with water-based drilling fluids. Numerous shale swelling inhibitors such as salts, polymers, and other chemical additives are used in drilling muds to stabilize shale formations. The inspection of the swelling inhibition capabilities of these additives is of utmost importance before its field application. Capillary Suction Timer instrument has been used to evaluate the shale inhibition performance of three categories of shale stabilizers for water-based drilling fluids. Each category of additive contains three different types of chemicals: salts (Potassium Chloride, Potassium Formate, and Choline Chloride), polymers (polyN-[3-(dimethylamino)propyl]methacrylamide, poly[2-(acryloyloxy)ethyl]trimethylammonium chloride and Polyvinylpyrrolidone) and ionic liquids (1-Butyl-3-methylimidazolium chloride and two other deep eutectic solvents). Capillary Suction Time for each chemical additive has been determined for various concentrations at room temperature. This test helps assess the clay swelling capability of the drilling mud in the presence of different additives at various concentrations. The Capillary Suction Time is dependent on the type and concentrations of chemical additive exhibiting a relative change in clay water interaction inhibition. The ionic liquids area significant category for clay swelling inhibition showing a minimum increase in Capillary Suction Time after 24 h. This indicates shale inhibition durability of the additive by resisting the water intake by clay. In addition, Fourier Transform Infra-Red and X-ray Diffraction techniques helped in characterizing the clay under consideration. The present work focuses on modeling the capillary suction time using machine learning algorithms. This technique provides a robust model for assessing the swelling capabilities of the drilling mud based on the concentration and types of additives. Of the different algorithms used, the Gradient Boosting algorithm provides the best model for capillary suction time estimation with an accuracy of 98%.
Read full abstract