Developing a Data-Driven Operation Classifier and Stand Counter Tool to Improve Drilling Operations

Abstract Oil and gas exploration in onshore and offshore environments demands precision and rigorous control across all drilling stages. Drilling rigs utilize a wide range of surface sensors to collect critical data, including Weight on Bit (WOB), Hook Load (HL), Standpipe Pressure (SPP), and Revolutions Per Minute (RPM). However, the complexity of drilling operations combined with the volume of data generated presents significant challenges for drilling data interpretation, as raw sensor data is not typically encoded by initial measurement systems with important context describing operations carried out while drilling. These challenges highlight the necessity of advanced approaches to support decision-making during operations. Effective monitoring and management of drilling operations is crucial for optimizing performance and ensuring safety. Despite the availability of sensor data, the manual interpretation of these signals can lead to inefficiencies and missed opportunities for improvement. Automated classification systems offer a potential solution by providing insights into drilling activities, enabling better operational control and optimization. Furthermore, in the context of data driven solutions, knowing the type of operation is often the first step towards developing more complex data-based and physics-driven solutions. Previous studies have demonstrated the effectiveness of machine learning models in similar contexts, suggesting their potential for enhancing drilling operations. The problem lies in the difficulty of accurately and efficiently identifying and classifying key drilling operations using traditional methods. This issue becomes more pronounced as operations scale in complexity and the volume of data increases. Without automated systems, critical analyses, such as predictive analytics and early issue detection and crew performance evaluation, are hampered, leading to potential delays, increased risks, and suboptimal performance.