- 129 views
- 65 downloads
Machine Learning Applied to Prediction of Pavement Performance under Cold Conditions
-
- Author / Creator
- Huang, Yunyan
-
Pavements in cold regions face challenging environmental conditions, including prolonged periods of extreme cold, frost heave, and freeze-thaw cycles during late winter and early spring. These factors contribute to premature damage, reduced service life, and increased maintenance costs for cold region pavements. Previous studies have shown that these issues are primarily influenced by environmental factors rather than traffic loads, necessitating specific research for cold region pavements. To address these challenges, a full-scale Integrated Road Research Facility (IRRF) test road was constructed in Edmonton, Alberta, Canada to investigate the impact of the environment on pavement structures and the long-term performance of insulation materials.
Accurately predicting pavement temperature and moisture content within the base and subgrade layers is crucial for assessing the load-bearing capacity and overall performance of cold region pavements. Traditional approaches, including numerical and statistical models, often lack suitability for cold regions or provide limited predictions within the asphalt layer. In this research, artificial intelligence techniques, specifically machine learning models, were employed to predict pavement temperature and moisture content under cold conditions. Environmental data collected from the IRRF test road, along with weather information from a local weather station, were utilized to develop and train the machine learning models. These models exhibited higher accuracy compared to existing models in the literature, showcasing their potential for improved pavement performance prediction.
To investigate the long-term effects of different embankment and insulation materials, Falling Weight Deflectometer (FWD) tests were conducted on the IRRF test road. The FWD test results were analyzed to evaluate structural capacity changes in various sections after five years of operation. Test sections with embankments backfilled with a mixture of tire-derived aggregate and soil, as well as sections insulated with bottom ash, demonstrated performance comparable to conventional sections, exhibiting lower loss in load-bearing capacity.
Additionally, machine learning models were employed to enhance road management practices. Specifically, they were utilized to estimate the start and end dates of spring load restrictions and winter weight premium. By replacing fixed dates with dynamic predictions derived from frost and thaw depths, road management can become more adaptive and responsive.
In conclusion, artificial intelligence, specifically machine learning, presents a practical and robust method for predicting pavement temperature and moisture content at various depths, thereby enhancing pavement performance assessment. Furthermore, these models can be utilized to improve road management by providing accurate predictions for the timing of seasonal policies. The development of pavement temperature and moisture content prediction models requires a comprehensive dataset incorporating pavement measurements, moisture data, and relevant weather information. -
- Graduation date
- Fall 2023
-
- Type of Item
- Thesis
-
- Degree
- Doctor of Philosophy
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.