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) ›› 2019, Vol. 38 ›› Issue (02): 44-50.DOI: 10.3969/j.issn.1674-0696.2019.02.07

• 道路与铁道工程 • 上一篇    下一篇

基于GA灰色神经网络的沥青路面使用性能预测

陈仕周1,2, 李山1, 熊峰2, 李冠男1   

  1. (1. 重庆交通大学 土木工程学院,重庆400074; 2. 重庆鹏方路面工程技术研究院,重庆400054)
  • 收稿日期:2017-11-30 修回日期:2018-03-14 出版日期:2019-02-22 发布日期:2019-02-22
  • 作者简介:陈仕周(1965—),男,四川南充人,研究员,主要从事桥面铺装和道路建材方面的研究。E-mail: 2093136401@qq.com。 通信作者:李山(1993—),男,重庆人,硕士研究生,主要从事道路工程方面的研究。E-mail: 872644669@qq.com。

Forecasting of Asphalt Pavement Performance Based on GA-Gray Neural Network

CHEN Shizhou1, 2, LI Shan1, XIONG Feng2, LI Guannan1   

  1. (1.School of Civil Engineering, Chongqing Jiaotong University, Chongqing 400074, P. R. China; 2. Chongqing Pengfang Pavement Engineering Technology Research Institute, Chongqing 400054, P. R. China)
  • Received:2017-11-30 Revised:2018-03-14 Online:2019-02-22 Published:2019-02-22

摘要: 为准确预估我国沥青路面使用性能的变化趋势,在传统灰色预测模型GM(1,1)的基础之上,提出了无偏GM(1,1)模型和滑动GM(1,1)模型,并通过遗 传算法(GA)优化后的BP神经网络对传统、无偏与滑动GM(1,1)模型进行了组合,得到了兼顾灰色理论、遗传算法和BP神经网络优点的GA-灰色神经网络组合 预测模型,并以具体实例验证了该模型的有效性。结果表明:传统GM(1,1)模型的平均相对误差为4.67%,无偏GM(1,1)模型的平均相对误差为4.64%,滑动 GM(1,1)模型的平均相对误差为4.63%,灰色神经网络组合模型的平均相对误差为2.41%,而GA-灰色神经网络组合模型平均相对误差仅为0.54%,证明所提 出的组合模型预测精度较高,误差较小,可作为制定路面养护计划的依据。

关键词: 道路工程, 沥青路面使用性能, 灰色预测模型, BP神经网络, 遗传算法

Abstract: In order to accurately predict the variation trend of asphalt pavement performance in our country, the unbiased GM(1,1) model and sliding GM(1,1) model were proposed on the basis of the traditional gray forecasting model GM(1,1) model.The traditional model, unbiased model and sliding GM (1, 1) model were combined with BP neural network optimized by genetic algorithm (GA), and a combined forecasting model of GA-grey neural network was obtained, which took into account the advantages of grey theory, genetic algorithm and BP neural network.Finally, the effectiveness of the proposed combination model was verified with specific examples.The results show that: the average relative error of traditional GM(1,1) model is 4.67%; the average relative error of unbiased GM(1,1) model is 4.64%; the average relative error of sliding GM(1,1) model is 4.63%; the average relative error of gray neural network combination model is 2.41%,while the average relative error of GA-gray neural network combination model is only 0.54%. Therefore, it is proved that the proposed combination model has higher prediction accuracy and smaller error, which can be used as a basis for pavement maintenance planning.

Key words: highway engineering, asphalt pavement performance, gray prediction model, BP neural network, genetic algorithm

中图分类号: