Speaker
Description
This paper investigates the application of machine learning (ML) to model and predict drilling mud loss in subsurface formations with conductive natural fractures. Mud loss during drilling is a complex and costly issue that disrupts operations and increases non-productive time. The goal of this study is to develop an ML-based tool that leverages type-curves and physics-informed models to predict key parameters such as cumulative mud loss volume, maximum mud loss duration, and equivalent hydraulic fracture aperture.
The study integrates a physics-informed approach to model mud loss using the Herschel-Bulkley fluid model, which accounts for non-Newtonian fluid behavior. A Latin Hypercube Sampling (LHC) method systematically varies uncertain parameters, such as yield stress, consistency factor, and hydraulic fracture aperture, to generate a robust training dataset. We introduce a novel concept of terminal mud loss volume (TMLV) and terminal mud loss time (TMLT) to measure and predict mud loss dynamics. An artificial neural network (ANN) is employed to predict mud loss behavior, using cumulative mud loss data as input. The model was trained and validated using both synthetic and field data to ensure accuracy and adaptability. Early mud loss trends are incorporated to improve predictions and refine estimates of fracture conductivity.
The developed ML-based model demonstrated high accuracy in predicting cumulative mud loss, maximum loss duration, and equivalent hydraulic fracture aperture under a range of conditions. It effectively captured the complex, nonlinear relationships governing mud loss behavior in fractured formations. The ANN model successfully integrated physics-informed equations, yielding predictions that are closely aligned with field observations. This streamlined approach reduces computational demands while maintaining reliability, offering practical solutions for real-time decision-making in lost circulation scenarios.
This study introduces a novel machine-learning framework for modeling and mitigating mud loss in naturally fractured formations. By combining physics-based models with ML techniques, the proposed tool enhances the predictive capabilities of traditional methods and provides actionable insights for managing lost circulation. The approach is adaptable to diverse scenarios, making it an accurate and efficient solution for addressing one of the oil and gas industry’s most persistent challenges.
| Country | Saudi Arabia |
|---|---|
| Student Awards | I would like to submit this presentation into both awards |
| Acceptance of the Terms & Conditions | Click here to agree |
