Unsupervised Clustering of Asphalt Pavement Conditions Using Fuzzy C-Means Algorithm with Principal Component Analysis Aided Dimensionality Reduction

This paper presents an innovative approach for categorizing pavement distress parameters by applying fuzzy C-means (FCM) clustering. Principal component analysis (PCA) visualizes clusters based on significant variability-explaining components, providing in-depth insights. The methodology is executed on an extensive dataset encompassing 1308 km of India’s road network, specifically focusing on flexible pavements surveyed via a network survey vehicle (NSV). The dataset comprises nine essential attributes: international roughness index (IRI), cracking, wide structural crack, raveling, pothole, loss of surface material, rutting, edge breaking, and texture depth, collectively reflecting pavement condition. Clustering analysis outcomes reveal a clear segregation of pavements into two distinct clusters. The initial cluster encompasses pavements in relatively good condition, while the latter encapsulates pavements in poor condition. Principal component analysis, driven by PC1 and PC2, distinctly captures group patterns, explaining over 90% of data variability. PC1 scores facilitate the two-cluster classification of pavements. The fuzzy C-means algorithm assigns membership degrees to each data instance within the two groups. Detailed examination of data instances within each cluster further enhances our understanding of these groups. Group characteristics are established and documented based on cluster centers and concurrent statistical values. This study’s key findings include accurate damage assessment, real-field location identification, prioritization of severely distressed pavement stretches, and recommendations for remedial measures.