Abstract
Pipe stuck accident is one of the primary causes contributing to NPT (Non-Productive Time) in the drilling industry, incurring global financial losses amounting to hundreds of millions annually. For decades, Practices and techniques have been formulated and developed to avoid and tackle this downhole accident, while the early detection and proactive management are critical for effectively preventing this issue at its early stage.
This paper presents the development and implementation of a novel digital Twin system, which amalgamates physics-based modeling and machine learning to facilitate real-time monitoring and early warning of pipe stuck while drilling. The system performs a real-time simulation of hook load, torque, and SPP (Standpipe Pressure) under various drilling states based on T&D and wellbore hydraulics models, which undergo verification and calibration within predefined computing cycles in order to enhance accuracy and reduce error margins. The PSRI (Pipe Stuck Risk), ranging from 0-100%, is based on ROC (Rate of change) of these three actual drilling parameters and the deviation of the simulated value from actual ones under different drilling states, is proposed to provide a comprehensive risk assessment over time series regardless of drilling state. A Continuous, real-time assessment of pipe stuck risk, categorized into low (green), medium (yellow), and high (red) levels, can be realized with thresholds for parameter deviation and ROC derived from historical data of wells with and without pipe stuck accidents and RFA (Random Forest Algorithm).
The system is designed to alert users such as drillers or monitoring specialists with visual and auditory signals when a high-risk scenario is detected, prompting immediate preventative or corrective actions. Case studies have shown that the system could provide early warnings of high-risk of pipe stuck typically between 30 minutes to 2 hours ahead, particularly during back-reaming and tripping operation. This capability demonstrates the system’s effectiveness in early identification and detection of this downhole accident under different drilling states.
The system exemplifies a paradigm shift towards more efficient, safe, and predictive operational modes for early detection and warning of pipe stuck, demonstrating digital twin-driven technique could enhance the capture and utilization of digital data for cost containment and operational efficiency gains while drilling, thereby minimizing NPT and enhancing drilling optimization.
