System for Real-Time Rate of Penetration Optimization Using Machine Learning with Integrated Preventive Safeguards Against Hole Cleaning Issues and Stick-Slip

Abstract Drilling related costs can contribute 30-70% of operators’ capital expenditures for well construction. To reduce costs, operators can reduce bit-on-bottom time and flat time. This work describes a drilling optimization advisory system utilizing machine learning (ML) with integrated safeguards for preventing issues that might occur following drilling parameter changes intended to increase rate of penetration (ROP), such as hole cleaning (HC) issues which might lead to stuck pipe, or stick-slip that reduces drilling efficiency. This work builds on the authors’ previous publications on ROP optimization (OTC-31680-MS, SPE-214521-MS), incorporating modules targeted at prompt detection of stick-slip for timely mitigation, and ensuring advised drilling parameter changes do not potentially cause HC issues and pack-offs. The HC safeguard utilized a downhole Equivalent Circulating Density (ECD) estimation ML model (SPE-208675-MS), queried by the optimizer to estimate effects of proposed drilling parameter changes, and corresponding ROP, on the ECD. A configurable tolerance to (expected) ECD changes from baseline parameters ensured any ECD increases were acceptable. The stick-slip detector monitored the frequency spectra of surface rotary speed and torque measurements, and utilized a classifier to estimate the probability of stick-slip symptoms’ presence. The ROP optimization system with integrated ECD safeguard has been field-deployed in SE Asia since Q4 2023, with no stuck pipe incidents relating to pack-offs occurring since this version of the software has been in use. The further enhanced version with integrated stick-slip detection was deployed to field operations in Q2 2024; analysis of historical well data with torsional vibration issues demonstrates the detector identifies stick-slip with high performance, achieving a precision of 0.92 on holdout (unseen) drilling data intervals from five wells, with all stick-slip symptoms present in the data identified. With stick-slip identified based on the estimated probabilities, human monitoring staff are notified, and the ROP optimizer automatically alters its behavior to allow torsional vibrations to be mitigated in order to maintain high drilling efficiencies. The Literature contains many works on the separate topics of stuck pipe prevention, ROP optimization and vibration mitigation, however these have not previously been incorporated into a holistic system balancing these different, sometimes competing, objectives. This work demonstrates effective integration of modules for ROP optimization, and detection of pack-off risks and torsional vibrations, into a combined system enabling increased drilling efficiency while reducing risks leading to non-productive time, contributing to overall reduced well construction time.