摘要
为提高城镇沥青路面结构面层最大剪应力计算的效率和准确性,本文基于弹性层状体系相关程序计算了多组各种路面结构的剪应力值,作为神经网络模型的训练数据.然后,采用MATLAB软件构建BP(Back Propagation)神经网络拟合模型,并使用Levenberg-Marquardt(LM)算法进行优化.随后,通过Sobol法进行全局敏感性分析,评估影响面层最大剪应力的关键参数.研究结果表明,本文拟合获得的BP神经网络模型在预测最大剪应力时,与基于弹性层状体系相关程序计算的结果表现出高度相关性,相关系数R超过0.9999.从敏感性分析结果可得,在水平力系数较低时,面层厚度对最大剪应力的影响最大,而在水平力系数较高时,面层剪切模量的影响更显著.故基于本文提出的剪应力计算模型,可以显著提升计算效率和准确性,为优化城镇沥青路面结构设计提供了参考.
To enhance the computational efficiency and accuracy of determining the maximum shear stress in the surface layer of urban asphalt pavement structures,this study calculated shear stress values for multiple pavement configurations using an elastic layered system analysis program.These results served as training data for a neural network model.A BP(Back Propagation)neural network fitting model was subsequently constructed in MATLAB,with optimization via the Levenberg-Marquardt(LM)algorithm.Global sensitivity analysis was then performed using the Sobol method to identify key parameters influencing the maximum shear stress in the surface layer.The results indicate that the proposed BP neural network model exhibits a strong correlation with calculations from the elastic layered system program,achieving a correlation coefficient R exceeding 0.9999.Sensitivity analysis revealed that the surface layer thickness predominantly affects the maximum shear stress under low horizontal force coefficients,whereas the shear modulus of the surface layer becomes more influential at higher coefficients.The developed shear stress calculation model significantly improves both efficiency and accuracy,offering a valuable reference for optimizing urban asphalt pavement structural design.
作者
朱凌志
宋云连
ZHU Lingzhi;SONG Yunlian(School of Civil Engineering,Inner Mongolia University of Technology,Hohhot 010051,Inner Mongolia,China;Inner Mongolia Key Laboratory of Green Construction and Intelligent Operation and Maintenance of Civil Engineering,Hohhot 010051,Inner Mongolia,China)
出处
《力学季刊》
北大核心
2025年第2期471-484,共14页
Chinese Quarterly of Mechanics
基金
内蒙古自治区自然科学基金(2023LHMS05036)
内蒙古自治区直属高校基本科研业务费(JY20240004)。
关键词
沥青路面
面层剪应力
神经网络
敏感性分析
asphalt pavement
surface layer shear stress
neural network
sensitivity analysis
