Real-Time Rate of Penetration Prediction for Hybrid Bit from Surface Drilling Data Using Artificial Intelligence Techniques

Abstract

Abstract Accurate prediction of rate of penetration (ROP) is essential for drilling optimization and cost reduction. Hybrid drill bits, which combine rolling cutters and fixed cutters, introduce complex rock–bit interaction mechanisms that challenge traditional ROP modeling approaches. This study presents artificial intelligence (AI)-based models to predict real-time ROP for hybrid bit drilling using surface drilling data. A dataset consisting of around 1,600 data points collected from six wells drilled in the Egyptian Western Desert was used. The input features included weight on bit (WOB), torque, rotary speed (RPM), flow rate (GPM), standpipe pressure (SPP), inlet and outlet mud weights (MW), and inlet and outlet mud temperatures. Statistical analysis, correlation heat maps, and histogram distributions were performed to understand feature influence and data characteristics. Five wells were used for training and validation, while the sixth well was reserved for blind testing. Two AI models were developed: Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS). The ANN architecture was optimized using two hidden layers by systematically tuning neuron numbers, training algorithms, and transfer functions. The ANFIS model was optimized using Gaussian membership functions (gaussmf), linear output functions, 250 epochs, and a cluster radius of 0.735. Results showed that ANN outperformed ANFIS. The ANN model achieved correlation coefficients (CC) of 0.87 and 0.84 for training and testing datasets, respectively. In contrast, ANFIS yielded CC values of 0.75 and 0.68. The findings confirm that ANN is more capable of capturing the nonlinear drilling dynamics associated with hybrid bit cutting mechanisms. The developed model enables real-time ROP prediction and supports drilling parameter optimization for improved drilling efficiency.