Scholarlink

<p>This study explores the prediction of Marshall stability (MS) in asphalt mixes incorporating polyethylene terephthalate (PET), high-density polyethylene (HDPE), and polyvinyl chloride (PVC) plastic waste using a variety of machine learning models, including both traditional and innovative approaches. Emphasizing the significance of material selection for sustainable road construction, the evaluation employs metrics such as correlation coefficient (CC), mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), scatter index (SI), and comprehensive measure (COM). Results demonstrate superior performance of artificial neural networks over conventional models (CC: 0.957/0.956, MAE: 1.364/1.810, RMSE: 1.819/2.371, MAPE: 0.103%/0.111%, SI: 0.123/0.161, COM: 0.249). Additionally, novel Bagging and AdaBoosting SVM-based models achieve enhanced predictions (SI &lt; 5%, error margin: 0.1–0.2%). Sensitivity analysis identifies plastic size, asphalt content, and aggregate size as pivotal factors influencing MS, confirmed through SHapley Additive exPlanations (SHAP) analysis. Further analysis using partial dependence plots reveals that optimal mix designs for higher MS in modified asphalt concrete involve plastic sizes between 0.0 and 0.4&#xa0;mm and asphalt content ranging from 0.25 to 1%, contributing to a significant 40% increase in MS. This study offers insights crucial for refining asphalt concrete formulations, emphasizing meticulous control of plastic size, asphalt content, and aggregate size to enhance MS performance.</p> Graphical Abstract <p></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

Utilizing Conventional and State-of-the-Art Machine Learning Algorithms to Predict Marshall Stability of Modified Asphalt Mixes Incorporating PET, HDPE, and PVC Plastic Waste: Performance Evaluation and Mix Optimization

  • 期刊论文

  • 发表日期：

    2025年7月15日

查看全文
  • Bhupender Kumar,
  • Ahmad Alyaseen,
  • Navsal Kumar

摘要

This study explores the prediction of Marshall stability (MS) in asphalt mixes incorporating polyethylene terephthalate (PET), high-density polyethylene (HDPE), and polyvinyl chloride (PVC) plastic waste using a variety of machine learning models, including both traditional and innovative approaches. Emphasizing the significance of material selection for sustainable road construction, the evaluation employs metrics such as correlation coefficient (CC), mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), scatter index (SI), and comprehensive measure (COM). Results demonstrate superior performance of artificial neural networks over conventional models (CC: 0.957/0.956, MAE: 1.364/1.810, RMSE: 1.819/2.371, MAPE: 0.103%/0.111%, SI: 0.123/0.161, COM: 0.249). Additionally, novel Bagging and AdaBoosting SVM-based models achieve enhanced predictions (SI < 5%, error margin: 0.1–0.2%). Sensitivity analysis identifies plastic size, asphalt content, and aggregate size as pivotal factors influencing MS, confirmed through SHapley Additive exPlanations (SHAP) analysis. Further analysis using partial dependence plots reveals that optimal mix designs for higher MS in modified asphalt concrete involve plastic sizes between 0.0 and 0.4 mm and asphalt content ranging from 0.25 to 1%, contributing to a significant 40% increase in MS. This study offers insights crucial for refining asphalt concrete formulations, emphasizing meticulous control of plastic size, asphalt content, and aggregate size to enhance MS performance.

Graphical Abstract