Automated Preprocessing Techniques for High Frequency Downhole Sensor Data

Baumgartner, Theresa (Shell) | Ashok, Pradeepkumar (University of Texas) | van Oort, Eric (University of Texas)



Good quality downhole drilling data can provide valuable insights into the downhole environment, allowing for improvement of the drilling process. It can also greatly facilitate drilling automation. Despite these benefits, there are still high barriers to using downhole data in a timely and efficient manner. The work presented here aims to improve its usability for engineers and analysts by introducing a variety of strategies to automatically correct and draw insights from downhole measurements without human inputs.

Downhole dynamics measurements show errors that are currently inevitable, particularly because downhole sensors are affected by factors such as high pressures, high temperatures and a lack of appropriate calibration procedures. Methodologies for automatic corrections of such errors are presented and described in detail. All approaches are tested on medium to high frequency downhole data from a variety of field data sets.

Commonly observed sensor errors include drifts in accelerometer and strain gauge data. An algorithm described here corrects vibration data for offsets and enables a comparison of vibration levels throughout runs which otherwise would be impacted by such drifts. Strain gauge sensor drift affects weight/torque on bit measurements, which are generally corrected manually. The algorithms proposed here make better corrections than manual procedures by finding the exact instance of neutrality. This can potentially make time-consuming taring procedures in rig operations obsolete. Time alignment of downhole and surface data is another barrier for a comprehensive analysis and is often a source of many errors. Simple but effective methodologies are described that auto-align time-based data sets, even considering latencies due to travel times. In addition, novel data reduction techniques that help to effectively process, display, and analyze high-frequency data are also discussed.

Analysis of downhole data currently requires skills and experience that must be developed in a labor-intensive way from scratch in many organizations. This work summarizes practical experiences and novel scientific insights that can help any engineer to kick-start downhole data analysis. The paper aims to increase transparency and share ideas amongst the drilling community, with the goal of improving drilling performance through downhole data analysis.