Enhanced Rate of Penetration Prediction for Complex Bottom Hole Assemblies Using Advanced Artificial Intelligence Techniques

Abstract Accurate Rate of Penetration (ROP) prediction is critical in drilling operations to optimize performance and reduce costs. The use of eccentric reamer bottom hole assemblies (BHAs) adds complexity due to the dual cutting structures involved, increasing the challenge of accurately predicting ROP. Traditional models often fail to capture the intricate dynamics of such configurations. Recent advances in artificial intelligence (AI) offer new possibilities for improving ROP prediction and optimizing drilling parameters in real-time. The objective of this study was to develop robust AI models to predict real-time ROP for eccentric reamer BHAs using surface drilling parameters. Two AI techniques, Artificial Neural Networks (ANN) and Adaptive Neuro-Fuzzy Inference Systems (ANFIS), were employed. A dataset of over 17,000 data points from five wells in a Middle Eastern field, including parameters such as torque, weight distribution, and surface drilling metrics, was used to train and test the models. The dataset was divided into a 70/30 split for training and testing, with an additional unseen well used for validation purposes. The performance of both models was evaluated using correlation coefficients (R) and root mean square error (RMSE). The ANN model demonstrated superior accuracy in predicting ROP, achieving an R value of 0.97 for the training set and 0.88 for the validation set. The ANFIS model, while effective during training (R = 0.92), showed reduced performance in validation, with an R value of 0.71. The ANN model's better performance is attributed to its capacity to handle complex non-linear relationships between input parameters. By optimizing drilling parameters based on these predictions, the models ensure improved ROP and reduced wear on cutting structures, ultimately minimizing drilling costs. This work introduces a novel AI-based approach specifically tailored for eccentric reamer BHAs, which has not been extensively explored in prior studies. The use of ANN and ANFIS in tandem for real-time ROP prediction in such complex configurations represents a significant advancement in drilling optimization. The study's key innovation lies in incorporating detailed weight and torque distribution into the models, resulting in improved prediction accuracy and better performance in real-world drilling environments.