Scholarlink

<p>Leak detection in subsea petroleum production pipelines represents a critical challenge for offshore operations, especially under increasingly stringent environmental and regulatory requirements. This work presents a complete framework for detecting oil leaks and seawater influxes through partial ruptures in deepwater production systems. The methodology integrates the generation of synthetic time-series data using multiphase transient flow simulation and the training of supervised machine learning models, validated through real and simulated datasets of operational events. Using pressure and temperature measurements from four standard field sensors, the proposed system detects anomalies associated with both leakage and influx phenomena with high sensitivity and low false-alarm rates. A LightGBM-based classifier demonstrated superior performance, achieving detection rates above 95% for moderate or severe ruptures after statistical smoothing and operational parameter tuning. The system was validated through a large-scale testing campaign and has already been deployed for real-time monitoring of multiple producing wells. Results confirm that this software-based detection approach enhances operational safety, reduces environmental risk, and can complement existing hardware-based monitoring systems in offshore fields.</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

Detection of oil leak and seawater influx through partial ruptures in subsea petroleum production pipelines

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Daniel Centurion Barrionuevo,
  • Galileu Henke de Oliveira,
  • Anderson Carlos Faller,
  • Saon Crispim Vieira,
  • Benno Waldemar Assmann,
  • Jean Carlos Dias de Araújo,
  • Alexey Pavlov

摘要

Leak detection in subsea petroleum production pipelines represents a critical challenge for offshore operations, especially under increasingly stringent environmental and regulatory requirements. This work presents a complete framework for detecting oil leaks and seawater influxes through partial ruptures in deepwater production systems. The methodology integrates the generation of synthetic time-series data using multiphase transient flow simulation and the training of supervised machine learning models, validated through real and simulated datasets of operational events. Using pressure and temperature measurements from four standard field sensors, the proposed system detects anomalies associated with both leakage and influx phenomena with high sensitivity and low false-alarm rates. A LightGBM-based classifier demonstrated superior performance, achieving detection rates above 95% for moderate or severe ruptures after statistical smoothing and operational parameter tuning. The system was validated through a large-scale testing campaign and has already been deployed for real-time monitoring of multiple producing wells. Results confirm that this software-based detection approach enhances operational safety, reduces environmental risk, and can complement existing hardware-based monitoring systems in offshore fields.