Scholarlink

<p>The scale of tight gas pipelines is expanding, leading to frequent corrosion-induced perforations due to flow-accelerated corrosion, under-deposit corrosion, microbiological corrosion, and electrochemical corrosion. Existing models cannot accurately predict corrosion rates due to complex interactions among factors such as flow rates, pipeline inclination angles, corrosive fluid compositions, pH values, temperature, and pressure. This study proposes a novel QPSO-DE-BNN model that integrates Quantum-behaved Particle Swarm Optimization (QPSO) and Differential Evolution (DE) to optimize Bayesian neural network (BNN) for corrosion prediction. The model effectively addresses complex interactions among corrosion factors. It is established based on the long-distance natural gas pipeline in Region A and has been applied to the newly built natural gas pipeline in Region B. The model achieves a maximum relative error of − 9.27%, a mean absolute error (MAE) of 0.0048, and a mean absolute percentage error (MAPE) of 3.72%. It provides robust technical support for pipeline maintenance and safe operation, significantly reducing corrosion-induced failure risks.</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

A Novel Hybrid Algorithm Model for Predicting the CO2 Internal Corrosion Rate of Tight Gas Pipelines

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Guoxi He,
  • Jialin Li,
  • Kexi Liao

摘要

The scale of tight gas pipelines is expanding, leading to frequent corrosion-induced perforations due to flow-accelerated corrosion, under-deposit corrosion, microbiological corrosion, and electrochemical corrosion. Existing models cannot accurately predict corrosion rates due to complex interactions among factors such as flow rates, pipeline inclination angles, corrosive fluid compositions, pH values, temperature, and pressure. This study proposes a novel QPSO-DE-BNN model that integrates Quantum-behaved Particle Swarm Optimization (QPSO) and Differential Evolution (DE) to optimize Bayesian neural network (BNN) for corrosion prediction. The model effectively addresses complex interactions among corrosion factors. It is established based on the long-distance natural gas pipeline in Region A and has been applied to the newly built natural gas pipeline in Region B. The model achieves a maximum relative error of − 9.27%, a mean absolute error (MAE) of 0.0048, and a mean absolute percentage error (MAPE) of 3.72%. It provides robust technical support for pipeline maintenance and safe operation, significantly reducing corrosion-induced failure risks.