Hybrid Data Driven Intelligent Algorithm for Stuck Pipe Prevention

Abstract

Abstract In today's drilling industry, it is essential to utilize both downhole and surface real-time sensor systems along with physics-based models to detect drilling hazards at an early stage and take timely measures to mitigate drilling related risks reducing non-productive time (NPT) and invisible lost time (ILT). Stuck pipe events are a major cause of NPT with an estimated cost to the oil and gas industry in the range of hundreds of Millions USD per year. Given its impact to drilling operations, stuck pipe prevention has high relevance and attention in the industry. In fact, in the last decade, the number of initiatives using drilling data coupled with advanced algorithms (combination of artificial intelligence (AI) and machine learning (ML)) to better understand and prevent stuck pipe events has greatly increased. Different approaches utilizing surface data are available ranging from data driven models, physics-based models, or even hybrid approaches combining physics-based models with data driven models. In any case, the outcome is an algorithm tasked with the identification of stuck-pipe events before they happen. If such an algorithm is deployed at the rig site (i.e., running on an edge device), ingesting surface and downhole data in real-time, the potential to improve the drilling process in terms of performance and safety greatly increases. Furthermore, safer drilling operations have an impact not only on the reduction of the overall capital expenditure (CAPEX) for well construction, but also on associated carbon emissions. The approach presented in this paper is based on drilling automation applications focused on the integrity of the drilling process. The system includes a set of advanced algorithms coupled with digital twins (e.g., physics-based models of the wellbore) running on an edge device deployed at the rig site, to create a comprehensive monitoring and alert solution for surface hookload. The monitoring system consists of three main components: a reference environment given by digital twins, which provides safe operating envelopes (SOE) defined by overpull and buckling as boundaries; a set of algorithms to detect and sample common indicators of torque and drag automatically, such as pick-up (PU), slack-off (SO), rotation off bottom (ROB) torques and loads; and a higher layer to identify trends and deviations between the samples to create early warnings related to stuck-pipe symptoms. The monitoring system implemented can be deployed for all kinds of drilling operations (i.e., drilling, tripping, circulating). By providing early warnings of stuck pipe like symptoms, the system enables users (i.e., rig crew, drilling operations, drilling optimization engineers) to mitigate such symptoms in time, hence avoiding costly consequences.