Scholarlink

<p>Unintentional wellbore deviation presents persistent challenges in drilling operations, often leading to cumulative trajectory irregularities that can compromise downhole equipment reliability and production performance. This study presents an integrated framework that combines machine learning (ML)-based prediction of deviation risk with a tortuosity-based engineering diagnostic to interpret the geometric and operational consequences of deviation. Using well-log-derived formation and geomechanical properties, multiple supervised ML classifiers are developed and evaluated, with the Random Forest (RF) model demonstrating robust performance in identifying deviation-prone intervals. The RF model demonstrated strong predictive performance, achieving a test accuracy of 96.67%, cross-validation accuracy of 98.31%, precision of 97%, recall of 99%, F1-score of 98%, ROC-AUC of 0.9975, PR-AUC of 0.9946, MCC of 0.963, and a log-loss of 0.647. A comprehensive feature importance and sensitivity analysis revealed features such as formation density, shale volume, and Poisson’s ratio as the top classifiers, accounting for 46.4%, 28.5%, and 19.2%, respectively, of the model’s decision-making criteria. To translate the deviation risk into physically interpretable trajectory complexity, a three-dimensional tortuosity index is incorporated, capturing cumulative curvature arising from frequent inclination and azimuth changes along the wellbore. Application to field wells from a major onshore unconventional basin in the United States with documented artificial lift system failures indicated that elevated vertical-section tortuosity is associated with specific failure mechanisms, including sucker rod and tubing-related failures, highlighting the mechanical relevance of trajectory quality beyond directional control. In contrast, analysis of lateral-section tortuosity showed no observable association with first-year production normalized by lateral length, consistent with prior findings that tortuosity in the lateral section does not directly impair production performance. By linking formation-driven deviation risk, trajectory complexity, and mechanical implications, the proposed framework provides a practical methodology for planning-stage deviation risk characterization and post-drill assessment of wellbore quality, supporting informed well design and artificial lift decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From deviation risk to trajectory complexity: machine learning-based deviation characterization and tortuosity analysis for wellbore quality assessment

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月4日

查看全文
  • Rupom Bhattacherjee,
  • Nayem Ahmed,
  • Sree Sujon Sutradhor

摘要

Unintentional wellbore deviation presents persistent challenges in drilling operations, often leading to cumulative trajectory irregularities that can compromise downhole equipment reliability and production performance. This study presents an integrated framework that combines machine learning (ML)-based prediction of deviation risk with a tortuosity-based engineering diagnostic to interpret the geometric and operational consequences of deviation. Using well-log-derived formation and geomechanical properties, multiple supervised ML classifiers are developed and evaluated, with the Random Forest (RF) model demonstrating robust performance in identifying deviation-prone intervals. The RF model demonstrated strong predictive performance, achieving a test accuracy of 96.67%, cross-validation accuracy of 98.31%, precision of 97%, recall of 99%, F1-score of 98%, ROC-AUC of 0.9975, PR-AUC of 0.9946, MCC of 0.963, and a log-loss of 0.647. A comprehensive feature importance and sensitivity analysis revealed features such as formation density, shale volume, and Poisson’s ratio as the top classifiers, accounting for 46.4%, 28.5%, and 19.2%, respectively, of the model’s decision-making criteria. To translate the deviation risk into physically interpretable trajectory complexity, a three-dimensional tortuosity index is incorporated, capturing cumulative curvature arising from frequent inclination and azimuth changes along the wellbore. Application to field wells from a major onshore unconventional basin in the United States with documented artificial lift system failures indicated that elevated vertical-section tortuosity is associated with specific failure mechanisms, including sucker rod and tubing-related failures, highlighting the mechanical relevance of trajectory quality beyond directional control. In contrast, analysis of lateral-section tortuosity showed no observable association with first-year production normalized by lateral length, consistent with prior findings that tortuosity in the lateral section does not directly impair production performance. By linking formation-driven deviation risk, trajectory complexity, and mechanical implications, the proposed framework provides a practical methodology for planning-stage deviation risk characterization and post-drill assessment of wellbore quality, supporting informed well design and artificial lift decision-making.