An intelligent framework for identifying void clogging in porous asphalt concrete based on tire-pavement interaction noise

• A non-destructive acoustic framework was proposed for clogging identification in porous asphalt concrete (PAC). • Tire–pavement noise increased with service age in field OBSI tests, with lane differences corresponding to traffic load. • Root mean square (RMS) and sound intensity level (SIL) were the most sensitive features for clogging identification. • RMS achieved the highest classification accuracy and robustness, further improved with 1-s sampling. • Level 2 clogging (3000≤C<3600) was identified as the stage where drainage and noise performance can still be effectively restored. This study proposes an intelligent acoustic-based framework for identifying void clogging in porous asphalt concrete (PAC) and informing maintenance timing. Tire–pavement interaction noise was collected from field sections with different pavement structures, service ages, and lane positions, followed by multi-domain feature extraction and clogging classification using a deep residual fully connected network (DRFCN). The results indicate that: (1) Tire–pavement noise increases with service age and consistently follows the order slow lane > fast lane > emergency lane due to traffic load distribution; (2) Root mean square (RMS) and sound intensity level (SIL) serve as the most sensitive acoustic features for clogging identification, with RMS achieving the highest accuracy and robustness—further enhanced under 1-s sampling; (3) Field cleaning tests suggest that Level 2, defined by a permeability coefficient 3000 ≤ C < 3600, allows more effective restoration of drainage and acoustic performance, whereas advanced clogging shows limited recoverability. The proposed method provides a non-destructive and practical approach for condition assessment, offering technical support for performance-oriented maintenance decision-making in PAC systems.