Evaluation of soft computing techniques for predicting Marshall Stability of waste plastic-reinforced asphalt concrete

Pavement engineering has long prioritized enhancing the quality and durability of asphalt concrete, a foundational material in road construction. This study employs advanced soft computing techniques to predict the Marshall Stability (MS) of asphalt concrete reinforced with waste plastic. Techniques such as Artificial Neural Network (ANN), Random Forest (RF), Random Tree (RT), Support Vector Machine (SVM), and Bagging RT are utilized. Evaluation of model effectiveness is conducted using seven statistical metrics: coefficient of correlation (CC), root mean square error (RMSE), mean absolute error (MAE), relative absolute error (RAE), root relative squared error (RRSE), mean absolute percentage error (MAPE), scatter index (SI), and comprehensive measure (COM). Among the applied models, the Bagging RT model emerges as the top performer, exhibiting superior performance across multiple metrics. Specifically, the Bagging RT model achieves impressive CC values of 0.921 and 0.834, indicating strong correlations between predicted and actual MS values. Additionally, it demonstrates low error metrics, with RMSE values of 1.632 and 2.869, and MAE values of 1.207 and 2.081, respectively. A sensitivity analysis conducted for the Bagging RT-based model underscores the significant influence of aggregate size on MS prediction, highlighting the model’s capability to elucidate critical factors shaping material stability.

路面工程领域长期以来始终将提升沥青混凝土——道路建设的核心基础材料——的质量与耐久性作为核心研究目标。本研究采用先进的软计算技术，对废弃塑料增强型沥青混凝土的马歇尔稳定度（Marshall Stability, MS）开展预测工作。研究所选用的模型包括人工神经网络（Artificial Neural Network, ANN）、随机森林（Random Forest, RF）、随机树（Random Tree, RT）、支持向量机（Support Vector Machine, SVM）以及装袋随机树（Bagging RT）。模型效能评估采用7项统计指标：相关系数（coefficient of correlation, CC）、均方根误差（root mean square error, RMSE）、平均绝对误差（mean absolute error, MAE）、相对绝对误差（relative absolute error, RAE）、相对均方根误差（root relative squared error, RRSE）、平均绝对百分比误差（mean absolute percentage error, MAPE）、散布指数（scatter index, SI）与综合评价指标（comprehensive measure, COM）。在所采用的模型中，装袋随机树模型表现最优，在多项评估指标上均展现出卓越性能。具体而言，该模型的相关系数（CC）分别达到0.921与0.834，表明预测值与实际马歇尔稳定度值之间存在极强的相关性。此外，其误差指标表现优异，均方根误差（RMSE）分别为1.632与2.869，平均绝对误差（MAE）分别为1.207与2.081。针对该装袋随机树模型开展的敏感性分析表明，集料粒径对马歇尔稳定度的预测结果具有显著影响，同时凸显了该模型能够阐明影响材料稳定性的关键因素的能力。