Data-Driven Assessment of Asphalt Pavement Rutting and Cracking Using Multisource Road Data for Preventive Maintenance

Rutting and cracking caused by traffic, environmental exposure, and existing damage threaten asphalt pavement serviceability. This study develops an interpretable, data-driven framework for deterioration-risk assessment and preventive maintenance using the Xinxiang section of National Highway G107, China. Multisource data on pavement condition, historical distress and maintenance, traffic operations, road geometry, and environmental conditions were integrated. Rutting severity and crack counts were modeled as the response variables for permanent deformation and cracking, respectively. Rutting severity was modeled using binary logit regression, while crack counts were analyzed using standard negative binomial and negative binomial conditional autoregressive models. Peak-hour traffic volume and pavement distress ratio were positively associated with higher rutting severity. For cracking, pavement distress ratio was positively associated with crack counts in the standard model, while grade became more influential after spatial dependence was considered. The standard negative binomial model achieved a mean absolute deviation (MAD) of 20.192 and a mean squared prediction error (MSPE) of 922.203, slightly outperforming the spatial model. Predicted crack counts identified five candidate sections requiring priority field inspection. The framework converts routine road data into section-level evidence for targeted preventive maintenance.