Perpustakaan Digital ITB

Sand production poses a critical operational and economic challenge in the oil and gas industry, leading to equipment failure, production loss, and wellbore integrity issues. Conventional geomechanical models are often data-intensive, and their predictive indices can yield contradictory results, limiting their practical application in mature fields. This study introduces a novel, data-driven framework that synergizes machine learning (ML) with empirical geomechanics to provide an actionable sand production risk assessment for wells produced with Electrical Submersible Pumps (ESPs). The core of the methodology involves using an Extra Trees Regressor, trained on limited core data and conventional well logs, to accurately predict the cementation exponent (m) as a robust proxy for rock integrity (R² = 0.9896 on unseen data). A field-specific failure envelope was then empirically derived by correlating historical sanding events with predicted 'm' values and operational normalized drawdown, defining a quantitative boundary between stable and unstable production regimes. The principal finding reveals a counter-intuitive risk dynamic: a high-risk 'catastrophic failure' regime exists at low operational drawdowns, while a more stable, 'managed-risk' regime is achievable at higher drawdowns. Validation on a blind test set of wells demonstrated the framework's superior predictive power, correctly identifying wells that failed prematurely (<35 days) in the high-risk zone and those achieving long run lives (>1000 days) in the managed-risk zone, outperforming conventional indices. These insights are consolidated into a novel quadrant-based risk map, an intuitive decision-support tool that visualizes a well's risk profile based on its rock integrity and operational stress. Ultimately, this framework provides a strategic sand management philosophy, enabling engineers to move beyond simple prediction to proactively design production and artificial lift strategies that either avoid the catastrophic failure envelope or consciously operate within a manageable risk regime to maximize asset value.