LAUSR

Abstract Aggregate plays the role of skeleton filling in asphalt pavements. The shape of the aggregate affects the embedded structure between the aggregates, thus affecting the performance of asphalt concrete. In this study, extreme gradient boosting (XGBoost) classification is used to study the automatic shape classification of aggregates. The expression of main and microscopic features of aggregate was improved by transforming aggregate images into data, and a feature importance analysis method based on method fusion is proposed to select the feature parameters of aggregate morphology. Based on cross-validation, the XGBoost classification model was trained by optimizing the super parameter combination to complete the classification of aggregate shapes. Compared with the random forest model, the results show that the proposed method can effectively classify aggregate shapes. It is also proved that the two-dimensional images can reflect the three-dimensional features of the aggregate to some extent. This method provides a certain theoretical basis for the automatic classification of aggregate, and simultaneously it has important practical significance to promote the intelligent production of asphalt mixtures.