Scholarlink

<p>Heavy crude oils present persistent challenges in production, transportation, and upgrading due to their inherently high viscosity. This study develops an artificial neural network (ANN)–based framework to predict the viscosity reduction percentage of East Baghdad heavy crude oil subjected to ultrasonic treatment using different additive systems, including alumina (Al₂O₃) nanoparticles and nanofluid formulations. A comprehensive experimental dataset was generated over a wide operating range, covering treatment temperatures of 20–75&#xa0;°C, ultrasonic exposure times of 15–60&#xa0;min, and varying additive types and concentrations. In addition to operational parameters, key physicochemical properties of the crude oil, including density, sulfur content, and trace metal concentrations (vanadium and nickel), were incorporated to account for compositional effects. A multilayer perceptron ANN was trained and validated using normalized inputs, achieving high predictive accuracy on the test dataset, with a coefficient of determination of 0.957 and low error metrics. The model maintained robust performance with reduced input sets, while the inclusion of vanadium and nickel contents resulted in measurable improvement in prediction accuracy, reflecting their influence on asphaltene stability and ultrasonic response. Permutation-based variable importance analysis identified treatment temperature and ultrasonic exposure time as the dominant factors governing viscosity reduction, followed by initial viscosity and compositional parameters. The results demonstrate that the proposed ANN framework provides an accurate and interpretable tool for predicting viscosity reduction under laboratory-scale conditions and supports the optimization of ultrasonic–additive upgrading strategies for heavy crude oils, subject to further validation across different crude oil systems.</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

Ultrasonic upgrading of east Baghdad heavy crude: explainable ANN prediction of viscosity reduction across additive systems

  • 期刊论文

  • 发表日期：

    2026年6月3日

查看全文
  • Luay Ahmed Khamees,
  • Ghassan H. Abdul-Majeed,
  • Ayad A. Alhaleem

摘要

Heavy crude oils present persistent challenges in production, transportation, and upgrading due to their inherently high viscosity. This study develops an artificial neural network (ANN)–based framework to predict the viscosity reduction percentage of East Baghdad heavy crude oil subjected to ultrasonic treatment using different additive systems, including alumina (Al₂O₃) nanoparticles and nanofluid formulations. A comprehensive experimental dataset was generated over a wide operating range, covering treatment temperatures of 20–75 °C, ultrasonic exposure times of 15–60 min, and varying additive types and concentrations. In addition to operational parameters, key physicochemical properties of the crude oil, including density, sulfur content, and trace metal concentrations (vanadium and nickel), were incorporated to account for compositional effects. A multilayer perceptron ANN was trained and validated using normalized inputs, achieving high predictive accuracy on the test dataset, with a coefficient of determination of 0.957 and low error metrics. The model maintained robust performance with reduced input sets, while the inclusion of vanadium and nickel contents resulted in measurable improvement in prediction accuracy, reflecting their influence on asphaltene stability and ultrasonic response. Permutation-based variable importance analysis identified treatment temperature and ultrasonic exposure time as the dominant factors governing viscosity reduction, followed by initial viscosity and compositional parameters. The results demonstrate that the proposed ANN framework provides an accurate and interpretable tool for predicting viscosity reduction under laboratory-scale conditions and supports the optimization of ultrasonic–additive upgrading strategies for heavy crude oils, subject to further validation across different crude oil systems.