Automated Correction of Drillstring Data for Improved Reliability and Trust in Use of Physics-Based Models in Drilling Advisory Systems

A major challenge in utilizing physics-based models such as torque and drag (T&D), hydraulics, cutting transport, drillstring dynamics, etc., in real-time, and with high reliability and confidence has been the lack of accuracy in drillstring data available as input to these models. This paper describes a methodology to automatically cleanse drillstring data, ensuring that the model outputs are accurate and trustworthy.The data cleansing process involves comparing measured data with model predictions and iteratively resolving the anomalies. First, the unit weights and lengths of the bottom hole assembly (BHA) components are adjusted by minimizing a specially defined loss between the measured and predicted hook load from a T&D model. Subsequently, the outer diameter (OD) and inner diameter (ID) of the components are corrected by minimizing the discrepancy between the observed standpipe pressure and the modeled pump pressure using a hydraulics model. In addition, the method leverages a comprehensive database of manufacturer drillstring specifications to ensure that the corrected values are realistic.

The proposed approach was rigorously validated using field data, demonstrating its effectiveness across various BHA combinations and wellbore conditions. Additionally, by using a comprehensive drillstring specifications database, the process achieved greater robustness, particularly improving corrections in scenarios where field data was limited. After data cleansing, both the T&D and hydraulics models showed significant improvements in accuracy. The T&D model aligned more closely with field data, resulting in more accurate friction factor and downhole weight on bit (WOB) estimations. Furthermore, the hydraulics model provided better pressure estimates along the drill string, enhancing the safety of well construction. Other models, such as hole cleaning and drillstring dynamics, also stand to benefit significantly, leading to better downhole situational awareness and more reliable drilling operations.

This paper is the first to describe an automated drillstring data cleansing method that enhances the quality of critical input parameters for various drilling-related models. The proposed approach seamlessly integrates information from multiple models, real-time data, and a comprehensive database to ensure precise and consistent correction of drillstring specifications. This approach promises to significantly increase the use of physics-based models in real-time advisory and automation systems.

Contact

Get in Touch