Assessment of Marshall Stability of asphalt concrete with plastic waste using soft computing techniques

Asphalt concrete has long been the major material used in pavement engineering. As a result, improving the quality of asphalt concrete for better performance and longer durability has remained a research priority. The purpose of this research is to evaluate the potential of soft computing-based models for predicting the Marshall Stability of plastic waste asphalt mix, such as Random Forest (RF), Random Tree (RT), Bagging RF, Bagging RT, and artificial neural networks (ANN)-based models. Three statistical indices are used to assess each model's performance with various input variables, such as bitumen content (BC), plastic content (PC), bitumen grade (VG), and plastic size (PS), to derive the Marshall Stability (MS). The efficacy of the used models was evaluated using three statistical indices: the coefficient of correlation (CC), the mean absolute error (MAE), and root-mean-square error (RMSE). With CC values of 0.942 and 0.8957, MAE values of 1.0591 and 1.4736, and RMSE values of 1.5121 and 2.2225 for both the training and testing stages, the performance evaluation results showed that the RF-based model outperformed all other models for predicting the Marshall Stability (MS) of asphalt concrete using plastic waste. The MS of the asphalt mix is a key outcome of the sensitivity analysis pointing to the size of the plastic as an important parameter in the case of plastic waste.