中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊

重庆交通大学学报(自然科学版) ›› 2018, Vol. 37 ›› Issue (02): 35-40.DOI: 10.3969/j.issn.1674-0696.2018.02.06

• 道路与铁道工程 • 上一篇    下一篇

基于BP神经网络的沥青混合料疲劳性能预测模型

谢春磊1,3,张勇1,3,耿红斌2,3,王学营1,2,3   

  1. (1.内蒙古自治区交通建设工程质量监督局,内蒙古 呼和浩特 010070;2.内蒙古高等级公路建设开发有限责任公司, 内蒙古 呼和浩特 010070;3.内蒙古自治区道路结构与材料重点实验室, 内蒙古 呼和浩特 010070)
  • 收稿日期:2017-02-15 修回日期:2017-06-21 出版日期:2018-02-20 发布日期:2018-03-02
  • 作者简介:谢春磊(1988—),男,内蒙古赤峰人,工程师,主要从事土木工程道路与桥梁方面的研究。E-mail: 20041510.xie@163.com。
  • 基金资助:
    交通运输部建设科技项目(2014318J21060)

Asphalt Mixture Fatigue Life Prediction Model Based on Neural Network

XIE Chunlei1,3,ZHANG Yong1,3,GENG Hongbin2,3,WANG Xueying1 ,2,3   

  1. (1.The Inner Mongolia Autonomous Region Traffic Construction Engineering Quality Supervision Bureau, Hohhot 010070,Inner Mongolia,P. R. China; 2.Inner Mongolia High Grade Highway Construction and Development Co.Ltd.,Hohhot 010070,Inner Mongolia,P. R. China; 3.The Inner Mongolia Autonomous Region Key Laboratory of Road Strucure and Mateiral,Hohhot 010070,Inner Mongolia,P. R. China)
  • Received:2017-02-15 Revised:2017-06-21 Online:2018-02-20 Published:2018-03-02

摘要: 以MATLAB为平台,搭建了一种单隐层结构的BP神经网络沥青混合料疲劳寿命预测模型。模型以沥青含量、空隙率和应变水平3个指标作为输入项, 疲劳寿命作为输出项。进行了不同试验条件下的四点弯曲疲劳试验,将试验数据归一化处理后得到27组数据用于训练。训练后的神经网络疲劳寿命预测模 型拟合优度R2可达到0.91。采用同一组数据对传统的疲劳方程模型进行拟合,并对比两者的预测效果,各项统计指标均显示神经网络模型预测结果明显优 于疲劳方程模型,且神经网络还具有疲劳方程所不具备的指标权重分析的功能。

关键词: 道路工程, 疲劳寿命预测, 神经网络, 四点弯曲小梁疲劳试验

Abstract: A neural network fatigue life prediction model with single hidden layer was established on MATLAB platform. The model took 3 indexes as the input term,such as the asphalt content,the void fraction and the strain level,and took the fatigue life as the output item.Four point bending fatigue test was carried out under different conditions,and a total of 27 sets of test data after normalization were obtained for training.After training,the goodness of fitness of the neural network fatigue life prediction model (R2) could reach 0.91.To make a comparsion,the same data was used to fit the traditional fatigue life regression equation.All the statistical indexes show that the prediction results of the neural network model are obviously superior to that of the fatigue equation model.Meanwhile,the neural network also has the function of the index weight analysis that the fatigue equation does not possess.

Key words: highway engineering, fatigue life prediction, neural network, four point bending fatigue test

中图分类号: