Scholarlink

Purpose <p>This study investigates the desorption behavior of petroleum hydrocarbons (PHC) from soils with different particle sizes by combining experimental analysis, mathematical modeling, and machine learning (ML). It addresses gaps by exploring the relationships among the targeted soil properties and their impact on desorption removal efficiency (RE), providing a foundation for advanced, data-driven, site-specific soil remediation strategies.</p> Methods <p>Desorption kinetics and isotherm models were employed to describe the desorption behavior of PHC from soils. ML models including support vector regression (SVR), artificial neural networks (ANN), random forest (RF) and extreme gradient boosting (XGB) were used for further insights and predictions of PHC desorption.</p> Results <p>Scanning electron microscopy (SEM) and Brunauer-Emmett-Teller (BET) surface area analysis revealed an increase in surface area and pore volume with decreasing particle size, accompanied by a reduction in pore diameter. The pseudo-second order (PSO) kinetic model (R<sup>2</sup> = 0.991–0.998) and the Freundlich isotherm (R<sup>2</sup> = 0.988–0.996) demonstrated the best fit to the experimental data. Approximately 2700 data points were analyzed in ML and explored the soil properties relations and RE predictions. Extreme gradient boosting (XGB) shows the best performance based on an RMSE of 0.367 and an R² of 0.999.</p> Conclusion <p>Desorption techniques are widely reported as active soil remediation technologies. The outcome of this research provides a solid foundation for developing efficient data driven and a reference contaminated sites remediation strategy for PHC contaminated soils particularly of varying soil textures and compositions.</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

Integrated experimental and machine learning analysis for modeling desorption attenuation of petroleum hydrocarbon in soils with varying particle sizes

  • 期刊论文

  • 发表日期：

    2026年3月27日

查看全文
  • Ali Ammar,
  • Waseem Akram,
  • Yanmeng Zhang,
  • Zhangjiu Liao,
  • Yanfeng Zhang,
  • Lingyan Zhu

摘要

Purpose

This study investigates the desorption behavior of petroleum hydrocarbons (PHC) from soils with different particle sizes by combining experimental analysis, mathematical modeling, and machine learning (ML). It addresses gaps by exploring the relationships among the targeted soil properties and their impact on desorption removal efficiency (RE), providing a foundation for advanced, data-driven, site-specific soil remediation strategies.

Methods

Desorption kinetics and isotherm models were employed to describe the desorption behavior of PHC from soils. ML models including support vector regression (SVR), artificial neural networks (ANN), random forest (RF) and extreme gradient boosting (XGB) were used for further insights and predictions of PHC desorption.

Results

Scanning electron microscopy (SEM) and Brunauer-Emmett-Teller (BET) surface area analysis revealed an increase in surface area and pore volume with decreasing particle size, accompanied by a reduction in pore diameter. The pseudo-second order (PSO) kinetic model (R2 = 0.991–0.998) and the Freundlich isotherm (R2 = 0.988–0.996) demonstrated the best fit to the experimental data. Approximately 2700 data points were analyzed in ML and explored the soil properties relations and RE predictions. Extreme gradient boosting (XGB) shows the best performance based on an RMSE of 0.367 and an R² of 0.999.

Conclusion

Desorption techniques are widely reported as active soil remediation technologies. The outcome of this research provides a solid foundation for developing efficient data driven and a reference contaminated sites remediation strategy for PHC contaminated soils particularly of varying soil textures and compositions.