Scholarlink

<p>Asphalt infrastructure is critical to modern transportation networks, yet maintaining it has become increasingly challenging due to rising traffic volumes and aging conditions. This study investigated the application of advanced machine learning techniques, specifically XGBoost and Logistic Regression, for predicting failures in asphalt pavement. Utilizing a comprehensive dataset spanning from 2014 to 2023, we found that XGBoost significantly outperformed Logistic Regression across key performance metrics, achieving a recall of 0.763, an F1 Score of 0.730, a Matthews Correlation Coefficient (MCC) of 0.701, and an Area Under the Curve (AUC) of 0.871. In contrast, Logistic Regression recorded a recall of 0.679, an F1 Score of 0.665, an MCC of 0.640, and an AUC of 0.845. These findings underscore the importance of accurate pavement failure detection in mitigating safety risks and reducing maintenance costs. Moreover, this research provides a versatile, cost-effective solution applicable to large metropolitan areas, mid-sized cities, and smaller municipalities, all of which face significant road maintenance challenges. By optimizing machine learning methodologies to function efficiently across diverse computing systems, this study empowers municipalities with limited resources to implement predictive maintenance strategies. Rather than assuming ideal conditions, this study embraces real-world limitations—such as sparse sensor data and fragmented records—and demonstrates that effective predictions are still achievable. However, the dataset reflects temperate urban environments and may require further adaptation for broader climatic and infrastructural contexts. Future research could explore additional algorithms and the integration of advanced surveillance techniques, such as drone monitoring and IoT sensor networks, to enhance predictive accuracy and enable real-time maintenance tracking. Additionally, expanding the analysis to include more diverse geographic regions and pavement types will help improve generalizability and scalability. Ultimately, this study promotes the development of resilient, sustainable road networks that align with broader goals of reducing resource waste and environmental impact.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Asphalt Management: Machine Learning for Predictive Maintenance and Sustainability in Urban Areas

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Yasin Asadi

摘要

Asphalt infrastructure is critical to modern transportation networks, yet maintaining it has become increasingly challenging due to rising traffic volumes and aging conditions. This study investigated the application of advanced machine learning techniques, specifically XGBoost and Logistic Regression, for predicting failures in asphalt pavement. Utilizing a comprehensive dataset spanning from 2014 to 2023, we found that XGBoost significantly outperformed Logistic Regression across key performance metrics, achieving a recall of 0.763, an F1 Score of 0.730, a Matthews Correlation Coefficient (MCC) of 0.701, and an Area Under the Curve (AUC) of 0.871. In contrast, Logistic Regression recorded a recall of 0.679, an F1 Score of 0.665, an MCC of 0.640, and an AUC of 0.845. These findings underscore the importance of accurate pavement failure detection in mitigating safety risks and reducing maintenance costs. Moreover, this research provides a versatile, cost-effective solution applicable to large metropolitan areas, mid-sized cities, and smaller municipalities, all of which face significant road maintenance challenges. By optimizing machine learning methodologies to function efficiently across diverse computing systems, this study empowers municipalities with limited resources to implement predictive maintenance strategies. Rather than assuming ideal conditions, this study embraces real-world limitations—such as sparse sensor data and fragmented records—and demonstrates that effective predictions are still achievable. However, the dataset reflects temperate urban environments and may require further adaptation for broader climatic and infrastructural contexts. Future research could explore additional algorithms and the integration of advanced surveillance techniques, such as drone monitoring and IoT sensor networks, to enhance predictive accuracy and enable real-time maintenance tracking. Additionally, expanding the analysis to include more diverse geographic regions and pavement types will help improve generalizability and scalability. Ultimately, this study promotes the development of resilient, sustainable road networks that align with broader goals of reducing resource waste and environmental impact.