A New Automated Carrying Capacity Index Model Optimizes Hole Cleaning Efficiency and Rate of Penetration by Applying Machine Learning Technique

Abstract

Hole cleaning is a major factor to drill hydrocarbon wells in safe and competent manner with cost effectiveness. Hole cleaning contributes to minimizing drilling troubles such as stuck pipe incidents and avoid well loss if hole cleaning was not managed properly. Ensuring adequate hole cleaning efficiency will help drill fast with smooth rate of penetration (ROP) with desired drill cuttings transport. In this paper, the development of a new real time hole cleaning model to evaluate and monitor hole cleaning effectiveness while drilling and ensure drilling efficiency optimization with high quality and economics in safe and environmental manner. Moreover, Artificial intelligence (AI) tool that is artificial neural network (ANN) was applied for confirming and validating selected parameter of model to show similar real time profile. The methodology to develop real time hole cleaning model is based on carrying capacity index that was developed earlier only for vertical wells. The original carrying capacity index will be optimized and enhanced to consider other mechanical drilling parameters and drilling fluid theological properties. The model will account the wellbore inclination, average hole cleaning annular, cuttings, hydraulics velocities based on the impact of cuttings accumulation, jetting drill bit nozzles, mud rheology, gravity, ROP, temperatures, and pressures and altered drilling fluid viscosities with rig and bit hydraulics. Many offset mechanical drilling parameters and drilling fluid properties were collected for studying the influences and relationships on hole cleaning efficiency and rate of penetration. The developed model will be developed and fed by real time values of sensors of drilling rig and generate real time profile of hole cleaning efficiency for evaluating, monitoring, and improving ROP with allowing immediate intervention by drilling team while drilling operations. The model can be used in panning phase and different drilling scenarios to have an evident imagination of downhole cleaning effectiveness. On the other hand, ANN application was run by selecting inputs of mud pump flow rate (Q), standpipe pressure (SPP), rate of penetration (ROP), plastic viscosity (PV), yield point (YP), mud weight (MW) and low shared yield point (LSYP) were collected and used of total number 5563 real time readings. The newly developed real time model was applied in the field in vertical and directional hole sections with water base mud and oil base mud to improve rate of penetration (ROP) and evaluate mud theological properties capability to have effective drill cuttings transport. The drilling efficiency was obtained, and ROP improved by 55%. While ANN model showed regressions (R2) 0.961 & 0.956 with absolute average percentage error (AAPE) 2.595 & 2.621 for training and testing validations respectively. The real time model was applied as well in real time offset wells parameters and confirm the importance of real time hole cleaning model. The real time hole cleaning model can ensure consistency of evaluation, monitoring, and optimization for the drilling operation in real time bases by using real time values of sensors available in all drilling rigs. The model can interpret the downhole measurements and give clear indications about hole cleaning efficiency.