Generation of Pavement Maintenance Plans Considering Spatial Consistency: An Improved Artificial Neural Network Method

Abstract

The precise generation of project-level maintenance and rehabilitation (M&R) plans is pivotal for ensuring the operational feasibility and effectiveness of pavement interventions. Current M&R plans rarely account for practical on-site construction scenarios. This oversight leads to multiple M&R actions being mixed together on adjacent pavement sections, resulting in wastage of resources, manpower, and time. Furthermore, existing models for generating M&R plans rarely consider traffic safety factors and lack a systematic postmaintenance evaluation mechanism, resulting in the implementation of plans deviating from theoretical expectations. In response to these issues, this study proposes an improved M&R plan generation model based on an artificial neural network (ANN) called ANN_M. First, a clustering method is optimized to merge pavement sections to ensure spatial consistency. Next, this model integrates key traffic safety indicators. Finally, it introduces a postmaintenance evaluation feedback mechanism. Through a pavement performance prediction model, it evaluates the effectiveness of the M&R plans output by ANN_M and replaces plans with poor maintenance results. The M&R plan generation model is used to generate M&R plans with constructability for certain expressways in Shanxi Province. The results show that the addition of postmaintenance evaluation improved the model’s prediction accuracy by 5.57%. The introduction of pavement section spatial consistency is designed to eliminate execution fragmentation, thereby enhancing construction feasibility and potential operational efficiency. Compared with the traditional ANN M&R plan generation model that does not consider spatial consistency and postmaintenance evaluation, the prediction accuracy was improved by 8.66%, proving the obvious advantages of the ANN_M improved model in terms of accuracy. The ANN_M model offers a practical solution for consolidating disjointed M&R actions into spatially consistent and constructible M&R plans.