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) ›› 2023, Vol. 42 ›› Issue (3): 71-77.DOI: 10.3969/j.issn.1674-0696.2023.03.10

• 交通基础设施工程 • 上一篇    

基于神经网络算法的沥青路面使用性能组合预测模型研究

曹雪娟1,卢治琳2,吴博文2,黄莹2,王民3   

  1. (1. 重庆交通大学 材料科学与工程学院,重庆 400074; 2. 重庆交通大学 土木工程学院,重庆 400074; 3. 重庆市智翔铺道技术工程有限公司,重庆 400067)
  • 收稿日期:2021-08-23 修回日期:2023-02-25 发布日期:2023-05-11
  • 作者简介:曹雪娟(1979—),女,四川邻水人,教授,博士,主要从事新型路面材料方面的研究。E-mail:caoxj@cqjtu.edu.cn 通信作者:卢治琳(1997—),女,重庆人,硕士研究生,主要从事道路工程方面的研究。E-mail:1257120770@qq.com
  • 基金资助:
    重庆市教育委员会科学技术研究项目(KJZD-M201900701);材料工程重庆市研究生联合培养基地项目(201907);重庆交通大学研究生科研创新项目(2022S0015)

Combined Prediction Model of Asphalt Pavement Performance Based on Neural Network Algorithm

CAO Xuejuan1, LU Zhilin2, WU Bowen2, HUANG Ying2, WANG Min3   

  1. (1. School of Material Science and Engineering, Chongqing Jiaotong University, Chongqing 400074, China; 2. School of Civil Engineering, Chongqing Jiaotong University, Chongqing 400074, China; 3. Chongqing Zhixiang Paving Technology Engineering Co., Ltd., Chongqing 400067, China)
  • Received:2021-08-23 Revised:2023-02-25 Published:2023-05-11

摘要: 为了准确预测沥青路面使用性能变化规律,提出了基于径向基神经网络算法的路面使用性能组合预测模型PCA-GA-RBF;针对神经网络收敛速度慢、模型参数容易陷入局部最优的问题,采用主成分分析算法对路面使用性能影响因素进行了降维处理,利用遗传算法对神经网络结构进行了优化;通过路面行驶质量的预测分析对组合预测模型进行了验证。研究表明:组合预测模型PCA-GA-RBF的拟合优度R2=0.820,均方根误差S=2.645,比单一RBF神经网络预测模型误差降低了11.4%,平均预测准确率为84.13%;组合预测模型计算速率快、预测精度高、预测效果好。

关键词: 道路工程;RBF神经网络模型;主成分分析算法;遗传算法

Abstract: In order to accurately predict the change law of asphalt pavement performance, a combined prediction model PCA-GA-RBF of asphalt pavement performance based on radial basis function neural network algorithm was proposed. In view of the slow convergence speed of the neural network and easiness falling into local optimization of model parameters, the principal component analysis algorithm was used to reduce the dimension of the factors affecting the pavement performance, and the genetic algorithm was used to optimize the structure of the neural network. The combined prediction model was validated through predictive analysis of road driving quality. The research shows that the goodness of fit R2 of PCA-GA-RBF is 0.820, and the root mean square error S is 2.645, which is 11.4% lower than that of the single RBF neural network prediction model, and the average prediction accuracy of PCA-GA-RBF is 84.13%. The combined prediction model has fast calculation speed, high prediction accuracy and good prediction effect.

Key words: highway engineering; RBF neural network model; principal component analysis algorithm; genetic algorithm

中图分类号: