Analytical Approach to Enhancing Efficiency of Silt Loading Collection in EPA Vacuum Sweep Method Using K-Means Clustering

Abstract

This study explores the application of K-means clustering to optimize the selection of sampling locations for suspended silt loading (sL) on asphalt pavements, addressing the limitations of traditional random sampling methods in the EPA method. The objective was to identify reliable sampling points for road dust concentration measurement, with a focus on improving the accuracy of data collection using the vacuum sweep method. The elbow method was used to determine the optimal number of clusters, revealing that three clusters were ideal for 25 m intervals and five for 100 m intervals. The clustering analysis identified specific sampling locations within the 25 m and 100 m road sections, such as 1.5–4.5 m and 12–18 m, and 15–18 m, 39–42 m, 57 m, 69 m, and 87 m, respectively, which adequately captured sL characteristics. The silhouette score of 0.6247 confirmed the effectiveness of the clustering method in distinguishing distinct groups with similar sL characteristics. The comparison of clustered versus non-clustered sections across 15 pavement segments showed an error rate of approximately 6%. Properly selecting sampling points ensures more accurate dust concentration data, which is crucial for effective road maintenance and environmental management. The findings highlight that optimizing the sampling process can significantly enhance the precision of dust monitoring, especially in areas with varying sL characteristics.