Hybrid Physics-Field Data Approach Improves Prediction of ROP / Drilling Performance of Sharp and Worn PDC Bits

Abstract

This study presents a hybrid approach that combines data-driven and physics models for worn and sharp drilling simulation of polycrystalline diamond compact (PDC) bit designs and field learning from limited downhole drilling data, worn state measurements, formation properties, and operating environment. The physics models include a drilling response model for cutting forces, worn or rubbing elements in the bit design. Decades of pressurized drilling and cutting experiments validated these models and constrained the physical behaviour while some coefficients are open for field model learning. This hybrid approach of drilling physics with data learning extends the laboratory results to application in the field. The field learning process included selecting runs in a well for which rock properties model was built. Downhole drilling measurements, known sharp bit design, and measured wear geometry were used for verification. The models derived from this collaborative study resulted in improved worn bit drilling response understanding, and quantitative prediction models, which are foundational frameworks for drilling and economics optimization.

Well Control Space Out: A Deep-Learning Approach for the Optimization of Drilling Safety Operations

“…Having multiple sensors require the implementation of communication protocols, such as Bluetooth, Wi-Fi, 6Lo, ZigBee, among others, as well as synchronization protocols to time stamp the data before storing them in a database for further analyses. The processed data is then fed into the ML/DL models (data-, physics-based, or hybrid) to identify hidden patterns in the data and predict anomalies associated with drilling before they occur [63].…”