Scholarlink

Ensuring road quality is crucial for the safety and comfort of all road users. The International Roughness Index (IRI) is a fundamental index for assessing the performance of flexible pavements. However, many conventional methods to gauge this index often fall short in accuracy and reliability. To address this, the present research harnessed ensemble machine learning methods, incorporating data from 55 asphalt pavement sections from the Long-Term Pavement Performance InfoPave database. The models were trained using pavement structure, age, traffic, and climatic factors to predict IRI values. Results indicate that ensemble learning methods effectively predict pavement performance, recording R2 values ranging from 0.82 to 0.87. Notably, the initial IRI emerged as the most significant predictor. Moreover, models showed heightened precision for pavements displaying IRI values under 2.0 m/km, indicating their aptness for early-stage flexible road pavements. These findings underscore the potential to refine these models further and the necessity for more diversified datasets.

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

Asphalt Pavement Performance Prediction Using Ensemble Learning Methods

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Tiago Tamagusko,
  • Adelino Ferreira

摘要

Ensuring road quality is crucial for the safety and comfort of all road users. The International Roughness Index (IRI) is a fundamental index for assessing the performance of flexible pavements. However, many conventional methods to gauge this index often fall short in accuracy and reliability. To address this, the present research harnessed ensemble machine learning methods, incorporating data from 55 asphalt pavement sections from the Long-Term Pavement Performance InfoPave database. The models were trained using pavement structure, age, traffic, and climatic factors to predict IRI values. Results indicate that ensemble learning methods effectively predict pavement performance, recording R2 values ranging from 0.82 to 0.87. Notably, the initial IRI emerged as the most significant predictor. Moreover, models showed heightened precision for pavements displaying IRI values under 2.0 m/km, indicating their aptness for early-stage flexible road pavements. These findings underscore the potential to refine these models further and the necessity for more diversified datasets.