Data-driven methods for data quality evaluation, maintenance assessment and decision model fusion in asphalt pavement management system

Abstract

Asphalt pavement management systems (PMSs) often face limitations in fulfilling all functions during maintenance and rehabilitation (M&R) projects. This study addresses key challenges by introducing a data-driven approach to enhance PMS intelligence. First, an artificial neural network (ANN)-based method is proposed for automated quality assessment of historical M&R data, evaluating data validity, accuracy, and timeliness to support manual reviews. Second, a data-driven theoretical benefit assessment model for M&R plans is developed, complemented by an ANN-based pavement deterioration prediction model, which reduces uncertainty in long-term performance forecasting and enables standardized, rapid estimation of M&R benefits. Lastly, a plug-and-play framework is proposed to integrate and manage decision-making models, enhancing coordination between new and existing technologies within PMS. These innovations collectively improve decision quality and adaptability. A case study in Shanxi Province, China, validates the effectiveness of the proposed methods, highlighting their practical applicability in large-scale highway maintenance management systems.