Abstract Investigations into rail corrugation within metro systems have traditionally focused on specific mechanisms, thereby limiting the generalizability of proposed theories. Understanding the commonalities in rail corrugation across diverse metro lines remains pivotal for elucidating its underlying mechanisms. The present study conducted extensive field surveys and tracking tests across 14 Chinese metro lines. By employing t-distributed stochastic neighbor embedding (t-SNE) for dimensional reduction and employing the unsupervised clustering algorithm DBSCAN, the research redefines the classification of metro rail corrugation based on characteristic information. The analysis encompassed spatial distribution and temporal evolution of this phenomenon. Findings revealed that floating slab tracks exhibited the highest proportion of rail corrugation at 47%. Notably, ordinary monolithic bed tracks employing damping fasteners were more prone to inducing rail corrugation. Corrugation primarily manifested in curve sections with radii between 300 and 500 m, featuring ordinary monolithic bed track and steel-spring floating slab track structures, with wavelengths typically between 30 and 120 mm. Stick–slip vibrations of the wheel–rail system maybe led to short-wavelength corrugations (40–60 mm), while longer wavelengths (200–300 mm) exhibited distinct fatigue damage characteristics, mainly observed in steel-spring floating slab tracks and small-radius curve sections of ordinary monolithic bed tracks and ladder sleeper tracks. A classification system comprising 57 correlated features categorized metro rail corrugation into four distinct types. These research outcomes serve as critical benchmarks for validating various theories pertaining to rail corrugation formation.

The formation, development and classification of rail corrugation: a survey on Chinese metro