Rigorous modeling of frictional pressure loss in inclined annuli using artificial intelligence methods

Abstract

One of the challenging issues during underbalanced drilling (UBD) is the prediction of frictional loss in the presence of three-phases of drilling fluid, cuttings, and air. In the current work, three innovative machine learning-based algorithms based-on Gradient tree boosting (GTB), Adaptive neuro-fuzzy inference system (ANFIS), and Extreme learning machine (ELM) have been suggested to calculate the frictional loss in gas-based drilling fluids containing cuttings in inclined annuli. A number of 216 real frictional pressure loss data in terms of hole inclination, pipe rotation, rate of penetration, and flow rates of each phase have been collected to train and validate the frictional pressure loss models. The visual and statistical comparisons of the frictional pressure loss models and actual values reveal that models have a fascinating ability in prediction of the frictional pressure loss for inclined wells. Besides, the GTB model has the best performance with R2 = 1, RMSE = 0.0031, MRE = 0.415, STD = 0.0023, and MSE≈0 in comparison with actual frictional pressure loss values. Furthermore, an exciting sensitivity analysis has been used to identify the effectiveness of each operational parameter on frictional pressure loss for inclined annuli. Due to these facts, the current work can be applied as an assistant in the development of drilling simulators in complicated field conditions.