Abstract
AbstractDrilling is challenging in hard and heterogeneous sandstones because drill bits can experience excessive wear, leading to short footage. Consequently, choosing the correct bit via the accurate prediction of bit progressive wear and corresponding drilling response for pre-well and real-time is essential for drilling optimization. An advanced progressive wear prediction tool has been developed to overcome this challenge, which includes a multi-well data learning engine. A description of the individual cutting structure and its wear is included in the model, enabling optimization of bit designs and drilling parameters that reduce total well construction costs.This work builds upon 3D drilling simulation at sharp or worn bit states. Progressive wear is included to simulate the evolution of wear flats with the distance drilled. The thermomechanical physics modeling is based on decades of cutting mechanics research and laboratory findings. To make this simulation tenable, a reasonable jump-in-distance was drilled, and corresponding wear was implemented based on incrementing the cutting location with the highest wear rate. The simulation is integrated into a multi-well learning engine for formation specific progressive wear model, using an evolution-based optimization algorithm. In the optimization process, the error metric included both drilling response error and the final wear area distribution error.Corrected downhole drilling data and final dull state were processed for the multi-well data learning of the progressive wear model. The approach is first demonstrated on five runs in three different offset wells for pre-well optimization. The results obtained from the optimized performance on multiple offset training wells and the validation of the test runs achieved accurate prediction results. The pre-trained progressive wear model was then deployed in a real-time field trial. The results show the wear model successfully predicted both the progressive wear response and final dull state distribution in real-time for multiple field deployment tests with significant performance improvement. The error of the wear model prediction is smaller than IADC 0.5. The methodology implemented in this study demonstrates that physics and laboratory-based constraints imposed on the progressive wear model ensure that the model does not encounter overfitting and generalization issues during the data learning process due to the limited scarcity or bias in the training data.This study presents a 3D representation of bit design, cutting structure details, and their progressive wear simulation in a holistic framework. The framework is applicable to challenging formations, where substantial amounts of data are not available for conventional data-driven models. The hybrid data-physics approach is leveraged to optimize the model parameters with runs from the available multi-well training dataset. This area-specific drilling response simulation with progressive wear is a critical tool for the engineer in improving designs and developing a parameter roadmap tailored for the proposed bit in the upcoming well plan.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have