Development of a novel ensemble learning model for predicting asphalt volumetric properties using experimental data...

Area of Science:

• Civil Engineering
• Materials Science
• Data Science

Background:

• Asphalt mixtures are crucial for road infrastructure.
• Predicting asphalt performance is key for durable and cost-effective pavement design.
• Traditional methods often lack the predictive power of advanced algorithms.

Purpose of the Study:

• To predict key volumetric properties of asphalt mixtures: Aggregate Void Percentage (AVP), Percentage of Voids Filled with Bitumen (PVFB), and Percentage of Voids in the Marshall Sample (PVMS).
• To evaluate the efficacy of machine learning algorithms, specifically XGBoost and LightGBM, in forecasting these properties.
• To enhance prediction accuracy and stability using ensemble techniques and advanced hyperparameter optimization.

Main Methods:

• Collected and characterized ~200 asphalt samples from Ardabil, Iran, using 11 influential features.
• Employed XGBoost and LightGBM regression algorithms for volumetric property prediction.
• Utilized ensemble methods (Voting, Stacking) and optimization algorithms (APO, GGO) for model improvement and hyperparameter tuning.
• Performed feature selection and sensitivity analysis for dimensionality reduction and interpretability.

Main Results:

• XGBoost demonstrated excellent prediction accuracy, indicated by high R-squared and low RMSE values for all output variables.
• Ensemble techniques and hyperparameter optimization further enhanced model performance and prediction stability.
• Feature selection and sensitivity analysis aided in understanding parameter influence and model interpretability.

Conclusions:

• Machine learning, particularly XGBoost, effectively predicts asphalt mixture performance using experimental data.
• Combining traditional pavement data with advanced ML techniques offers a pathway to more cost-effective pavement design and management.
• The study highlights the potential of data-driven approaches in optimizing construction materials and infrastructure longevity.