Managed Pressure Drilling significantly improves tripping safety in challenging wells, by providing a closed-loop system with controlled flow parameters and volume calculations. A key tool in MPD monitoring is the Virtual Trip Tank (VTT), which calculates fluid volumes using the closed-loop Coriolis flowmeter measurements to detect anomalies like kicks and losses. However, current VTT systems suffer from calibration drift and lack reference volumes, leading to inaccuracies and false alarms. This article summarizes a new methodology for enhanced VTT monitoring with automatic calibration, and introduces a pattern-recognition approach for improved kick and loss detection, reducing reliance on traditional methods.
The Problem with Current VTT Systems
Traditional VTT algorithms rely on flow-in and flow-out measurements. While Coriolis flowmeters provide accurate flow-out data, flow-in is often estimated based on a fixed pump displacement and a constant pump efficiency. This assumption is problematic because pump efficiency, and thereby displacement, is affected by operational parameters such as pump pressure differential, pump velocity, and fluid rheological properties. Therefore, pump efficiency changes throughout the drilling operation, with an output decay as pumping pressure increases (Fig. 1). Consequently, using a fixed efficiency factor to compute the flow-in introduces errors that accumulate over time, causing the VTT signal to drift (Fig. 2). This drift can lead to misinterpretations, such as falsely indicating a kick when conditions are stable.
Moreover, distinguishing true well control events from normal operational disturbances during tripping is difficult. Pipe movement during the trip operation inherently generates flow disturbances, making it challenging to rely solely on delta-flow analysis (i.e., flow-in vs. flow-out comparison). These inaccuracies result in a high number of false alarms, eroding the drilling crew's trust in the system. The issue is compounded by factors like noisy flow-out measurements due to heave and choke effects, and the inability of digital twins to capture transient mud behavior accurately.
Consequently, even though tripping operations account for over 27% of globally reported well control incidents (Gillard et al., 2022, Fig. 3)—and despite the availability of advanced tools such as Coriolis flowmeters for real-time volume measurement— many operations in 2025, even on state-of-the-art drillships, still rely on conventional trip tanks and Excel-based tallies. This outdated approach often results in invisible lost time (ILT) due to repetitive operations caused by measurement uncertainties and second-guessing the actual trend. Furthermore, misinterpretation of influxes and losses remains common, compounded by the fact that these inaccurate readings are still used by operators and service providers as key references in well control decision-making.
Project Objectives
Therefore, based on these identified issues, the objectives for this project were:
- Remove the VTT measurement drifting, by developing an auto-calibration routine for the flow-in measurement (delta-flow needs to be zero for stable periods)
- Provide a reference displacement volume to validate tripping progression, by modeling the expected displacement (like an automated tripping tally)
- Provide model based early event detection while tripping, to assist decision making beyond the basic volume displacement comparison
A Novel Methodology for Enhanced VTT Performance
1. Flow-In Calibration
Real-time FLOW-IN calibration is grounded in mass conservation principle. If no additional fluid sources or sinks are present in the well—giving rise to influxes or losses—and there is no operational activity that could disrupt mass balance (e.g., drilling, moving the drill pipe, or adding/removing drill pipe), then the flow-out of the well should accurately reflect the flow-in. The basis for the flow in calibration is then to find steady-state parameters, where both flow-in and flow-out have the least possible deviation, and using the flow-out as ground truth generate a corrected flow-in value (Fig. 4).
The proposed methodology tackles these challenges through several integrated components:
1. Rig-State Detection Algorithm
This algorithm classifies rig operational states (tripping, in-slips, circulation) to identify suitable steady-state intervals for calibration (Fig. 5). By understanding the rig's real-time context, the system can differentiate between normal operational signatures and actual anomalies. Machine learning models are trained to achieve high accuracy in classifying these rig states, ensuring robust performance even with slight class imbalances.
2. Pattern-Recognition to find suitable calibration periods
Based on the rig-state detection results, the algorithm looks for periods where there is no pipe movement inside the well, the flow-in and flow-out are expected to remain stable (e.g., in-slips with pumps on, circulating). Then the suitable periods for calibration are filtered by computing the similarity between flow-in and flow-out vectors, and selecting only the closest looking ones, done using the Symbolic Aggregate Approximation (SAX) and Edit Distance computation. This technique transforms flow-in and flow-out signals into symbolic representations, allowing for easy comparison and identification of deviations (Fig. 6).
Based on the filtered values, the flow-in calibration curve is fitted, where the flow-out measurement is used as a ground truth to estimate the flow-in signal. A second-degree polynomial regression model corrects the rig-reported flow-in values, significantly improving delta-flow accuracy and minimizing VTT drift (Fig. 7, Fig. 8). This auto-calibration process eliminates the reliance on fixed efficiency factors, adapting to varying pump performance caused by changes in standpipe pressure and other operational conditions.
2. Modeled VTT for Displacement Tracking
A physics-based model calculates the expected theoretical displaced volume during tripping operations, considering factors like drill pipe displacement and fluid volume changes due to pressure variations. This modeled VTT serves as a reliable baseline for comparison with the corrected VTT signal (i.e., computing the delta VTT), validating the system's accuracy and providing a clear reference for volume tracking during the tripping operation (Fig. 9).
3. Tripping Displacement Fingerprinting and Automated Tally
The system generates a visual "fingerprint" of VTT and delta VTT trends for the current stand and overlays them with historical data from previous stands (Fig. 10). This enables real-time comparison, helping operators quickly identify deviations from expected patterns. Additionally, an automated tripping tally tracks key displacement and volume indicators per stand, replacing manual tracking and improving operational efficiency.
Demonstrated Performance and Field Deployment
The methodology has shown robust performance when evaluated on historical field data, with significant improvements in VTT accuracy. A strong improvement is observed on the error and drift of the VTT signal, improving the error by 85% (RMSE), reducing the drift slope by 97%, and finally improving the R-squared by 94%. The system's computational efficiency ensures compatibility with field computer resources, enabling real-time execution.
- The non-quantified measurement drifting is removed, improving accuracy.
- The calibration process is automated, making it consistent and repeatable, removing user bias.
An interpretable measurement of the tripping trend was also developed, using the Delta VTT, which indicates the deviation from the expected volumes. This methodology is aligned with the conventional tripping monitoring method, which should be easier to understand by experienced drilling crews.
Moreover, for accurate event detection, a method to monitor any abnormal conditions by fingerprinting the response of the last stands, also generating an automated tripping tally.
This represents a substantial improvement on tripping monitoring precision, reliability and safety, reducing false alarms and lost time, and promoting and improving the use of precise tools and methods to track potential well control risks.
Conclusion
This next-generation VTT auto-calibration system offers a comprehensive solution to the challenges of volume tracking and event detection in MPD tripping operations. By combining advanced rig-state detection and pattern recognition into an auto-calibration algorithm, and providing a reference measurement of the expected displacement with the modeled VTT, the system dramatically reduces false alarms, improves accuracy, and enhances overall tripping safety and efficiency. This innovative approach is slated for field deployment later this year, promising to transform well control practices in challenging environments.
https://doi.org/10.2118/SPE-228373-MS