A data-and-knowledge-driven framework for automated generation of asphalt pavement maintenance strategies

Abstract

Listen

Translate article

ABSTRACT Due to the inefficiency of manual decision-making, automation of pavement maintenance strategy-making (PMSM) has become a focal point of attention in pavement engineering. The current automated PMSM processes lack a framework capable of integrating engineering prior knowledge into decision. In this paper, a data-and-knowledge-driven (DKD) framework was proposed to achieve the automated generation of asphalt pavement maintenance strategies. A pavement performance prediction model Convolutional Neural Network (CNN) was trained considering multiple influencing factors to analyse pavement deterioration, and maintenance thresholds were set according to predicted performance data distribution. The model incorporating Bidirectional Encoder Representations from Transformers (BERT), Bidirectional Long-Short-Term Memory (Bi-LSTM) and Conditional Random Field (CRF) was employed to extract the name entities from pavement maintenance domain knowledge, constructing the knowledge graph to achieve maintenance treatment determination. A freeway in Eastern China was selected for case study. The result demonstrated that 87 one-hundred-meter pavement sections on the freeway need preventive maintenance totally, utilising micro-surfacing in 2023. Furthermore, a level of automation assessment method and a reasonableness of decision process assessment method were proposed to assess this framework. The results indicate that this framework significantly enhances the level of automation compared to traditional PMSM methods while maintaining decision reasonableness.