Automated Real-Time Torque-and-Drag Analysis Improves Drilling Performance
You have access to this full article to experience the outstanding content available to SPE members and JPT subscribers.
Significant progress has been made on physics-based torque-and-drag (T&D) models that can run either offline or in real time. Despite its numerous benefits, real-time T&D analysis is not prevalent because it requires merging real-time and contextual data of dissimilar frequency and quality, along with repeated calibration, the results of which are not easily accessible to the user. In this paper, the application of a real-time T&D model is demonstrated. The process of T&D analysis was automated, and the time and cost required to run physical models offline was reduced or, in some cases, eliminated.
Traditional electronic drilling recorders (EDRs) are third-party systems that collect rig-sensor data. Major limitations in the operator’s ability to fully leverage the potential of this data exist, including issues with rig-sensor-measurement quality and rigsite data-aggregation methods, relatively slow data-sampling rates, and limited interoperability. To this end, the operator initiated a project to use drilling data better, focusing on improving data quality by building on previous work validating rig-sensor data.
Rig-Based T&D Advisory System
T&D Modeling. Onshore US operators are investing heavily in unconventional horizontal plays. In these wells, excessive T&D is a critical limiting factor in exposure to productive formations.
Predictive T&D computer models were developed as early as 1984. An initial model assumed that T&D are both caused by sliding friction on the wellbore, using the product of the normal force and friction coefficient to yield values for T&D. Later, the model was put in standard differential form and made to include the effects of mud pressure. These models (now termed soft-string) treat tubulars like a rope, ignoring bending moments and assuming continuous contact with the wellbore.
Conversely, the stiff-string model was designed to account for borehole-assembly stiffness and radial clearances but requires more-complex numerical techniques such as finite-element analysis. Many researchers have focused efforts on increasing accuracy of the stiff-string model and decreasing the required time and computing power. Other researchers have focused on obtaining analytical solutions for T&D calculations.
A major weakness of the standard T&D model is the use of the minimum-curvature method in calculating wellbore trajectory, because the bending moment is discontinuous at survey points and a smooth curved wellbore is assumed between these points.
The soft-string model is used currently as the industry standard for routine T&D analysis. Accordingly, a 3D soft-string T&D model has been developed by the authors.
Model Verification and Input Calibration. A good match between the model and the industry-standard package was found for various openhole friction factors. Now that a fully verified 3D soft-string model has been obtained, the next step is to implement it for real-time application in an automated manner.
The inputs required to create a T&D model include wellbore surveys; block weight; mud weight; mud type; drillpipe weights, grades, and connections; and casing depths and friction factors (for both cased and openhole sections).
A natural application of real-time T&D calculations is the automated detection of excessive drag conditions while tripping out (overpull) or while tripping in (underpull). It is essential to recognize such events and their warning signs automatically in order to avoid excessive wear and potential damage to drilling equipment and hazardous stuck-pipe incidents.
For the calibration routines, and for real-time visualization of T&D results, hookload and torque measurements had to be selected from pertinent operations. The goal in selecting those values is to pull sufficient data points to identify deviations from modeled trends adequately without creating so much noise that the models are unusable. Furthermore, consistency in the manner in which data points are pulled is essential. The authors’ initial data-selection criteria for pick-up weight while tripping out of hole, slackoff weight while tripping in hole, rotating off-bottom weight/free rotating weight, and free rotating torque are detailed as follows:
- Tripping out: Rotary speed must be zero, the bit must be off bottom by at least 10 ft, the blocks must be moving upward, and the hookload must be greater than a reference hookload selected by the engine after the slips are removed for the current tripping interval.
- Tripping in: Rotary speed must be zero, the bit must be off bottom by at least 10 ft, the blocks must be moving downward, and the hookload must be less than a reference hookload selected by the engine after the slips are removed for the current tripping interval.
- Rotating off bottom: Bit depth must be constant over the last 30 seconds, the bit must be off bottom by at least 3 ft, rotary speed must be greater than 10 rev/min, pump rate must be greater than 20 strokes/min, and the hookload must be greater than the block weight.
Furthermore, hookload and torque values were initially averaged if multiple values were recorded within an increment of 30 ft and the results were stored at the last recorded depth. In an effort to analyze the effectiveness of the real-time data selection, the resulting data points were compared with unfiltered 1-Hz data from the third-party EDR provider for a test well. The company had recorded overpulls of 40,000 lbf at depths of 12,350, 12,250, and 12,050 ft measured depth (MD) for this well. While the rig was able to pull through each of the trouble spots, the decision was made to make a dedicated cleanout run with the subsequent bottomhole assembly (BHA) because of concerns with the well condition.
While the unfiltered 1-Hz data group displayed quite a bit of noise, significant overpull could be seen at depths of 12,000 ft and below (as recorded in time summaries) as well as at 9,000 ft. The filtered data lacked the same resolution in both tripping-out and tripping-in data from 6,000 to 9,000 ft. In a second iteration of the hookload plot, the reference hookload was removed from the required-data list.
After removing the reference hookload, the model was able to detect data points for the entirety of the trip in the hole. However, the unfiltered data still revealed that tripping-out points were not capturing the full overpull values. The depth range for averaging was then decreased to 3 ft for the third iteration.
Decreasing the depth range for the averaging method increased the number of overall data points recorded but still did not capture the full value of overpulls. The averaging method was removed altogether for a fourth iteration.
After removing the averaging altogether, the full value of the overpulls could be seen on the trip out, but irrelevant data such as block weight were still filtered out. A larger amount and variance of hookload data points in a depth range are valuable because these are often an indication of difficult tripping conditions, as was the case for this test well. More iterations are being run with different filtering techniques to further refine the data-selection process. Field personnel reported excessive overpull at approximately 12,000 ft MD.
Beliefs for both overpull and underpull were high, near approximately 9,000 and 12,000 ft MD, where jumps in the recorded hookload were seen for both tripping in and out. The drilling engineer and superintendent determined that a cleanout run was needed to improve the well’s condition before returning to drilling.
The last BHA run for this well still exhibited an openhole-friction factor of 0.2 for the trip out according to the automated T&D model. A drilling engineer can then take this friction factor and model the production casing run to estimate if the casing will experience lockup or helical buckling on the trip in. Automated overpull- and underpull-event detection can also make use of the information from this BHA run to alert the drilling engineer of potentially troublesome areas. In addition to the BHA models, the automated T&D engine also can be used for casing runs if all necessary contextual data are available.
The underpull-event belief for the production-casing run (Fig. 1) showed a high probability near the bottom of the production-casing run. This is a common troublesome zone in such operations, often requiring repeated circulation and rotation to force the last joints of casing to the bottom of the hole.
Making these analyses readily available and visible to the rig crew is essential for real-time application.
The authors have created a system that automatically generates T&D models in real time using contextual and real-time data. The rig-state engine then classifies hookload values as either pickup, slackoff, or rotating off-bottom weights to generate a traditional broomstick plot in real time. Cased-hole friction factors dependent on mud type, and openhole friction factors can be calibrated automatically to identify troublesome zones in the wellbore. In addition, an event-detection engine is in place to recognize overpull/underpull events autonomously while tripping. This tool has demonstrated value for assisting in real-time decision making on difficult BHA trips and planning casing runs to avoid costly equipment failure and downtime.
For a limited time, the complete paper SPE 191426 is free to SPE members.
Automated Real-Time Torque-and-Drag Analysis Improves Drilling Performance
01 February 2019
Equinor, Transocean Ink Deal for Five Automated Rigs
The contract is helping to solidify Europe’s offshore sector as the focal point for the rise of automated drilling technology.
Automated Operations and Wired Drillpipe Benefit Arctic Drilling
This paper presents a case history of drilling automation system pilot deployment, including the use of wired drillpipe, on an Arctic drilling operation.
Automation-Adoption Approach Maps Human/System Interaction
This paper describes the progress of directional-drilling-automation systems along the cognitive functions and levels of automation as defined by the Levels of Automation Taxonomy (LOAT) hierarchy introduced by the Drilling Systems Automation Roadmap Industry Initiative.
Don't miss out on the latest technology delivered to your email weekly. Sign up for the JPT newsletter. If you are not logged in, you will receive a confirmation email that you will need to click on to confirm you want to receive the newsletter.
12 February 2019
05 February 2019
06 February 2019
08 February 2019