期刊
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING
卷 23, 期 11, 页码 3716-3728出版社
TAYLOR & FRANCIS LTD
DOI: 10.1080/10298436.2021.1915493
关键词
Weigh-in motion; nonlinear regression; support vector regression; surface condition index; axle load spectra
类别
资金
- Mountain-Plains Consortium
- U.S. Department of Transportation
This study aims to analyze the impact of traffic loading on pavement performance using traditional regression models and machine learning methods. The use of support vector regression (SVR) method significantly increases prediction accuracy, and incorporating characteristics such as axle number and fitted Gaussian distribution of axle load spectra further improves model accuracy.
This study aims to use traditional regression model and machine learning method to analyse the impact of traffic loading on pavement performance. Pavement condition data were obtained from pavement management systems (PMS) and axle loads of truck traffic were collected at weigh-in-motion (WIM) stations. Support vector regression (SVR) method was selected for modelling pavement performance since it provides the flexibility to find the appropriate hyperplane in higher dimensions to fit the data and customise control errors in an acceptable range. Compared to traditional nonlinear regression model, the accuracy of pavement performance prediction was significantly increased by utilising the SVR method. The model accuracy was further improved by considering the number of axles and fitted Gaussian distribution of axle load spectra in the performance model. The derived SVR models were further used to investigate the impact of overweight truck on pavement life reduction considering characteristics of axle load distributions. The proposed pavement performance model can be further used in determining pavement damage caused by overweight trucks for pavement rehabilitation strategy and fee analysis is permitted.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据