基于GRA与BP神经网络的沥青路面模量参数反演分析  

A study on back-calculation of asphalt pavement moduli based on grey relational analysis and artificial neural networks

作　　者：邱欣[1] 刘锦俐 陶珏强 杨青[1] 胡永举[1] 

机构地区：[1]浙江师范大学工学院,浙江金华321004

出　　处：《浙江师范大学学报（自然科学版）》2018年第1期102-108,共7页Journal of Zhejiang Normal University:Natural Sciences

基　　金：国家自然科学基金资助项目(51408550);浙江省自然科学基金资助项目(LY18E080020);浙江省新苗人才计划项目(2017R404044)

摘　　要：为了提升沥青路面模量参数反演分析的可靠性,基于沥青路面动力有限元数值分析模型,在构建FWD路表弯沉与结构层模量参数对应数据库的基础上,将GRA和BP神经网络分析方法相结合,实现沥青路面模量参数的反演分析.结果表明:FWD径向距离为0,60,210 cm处的路表弯沉值分别与面层模量、基层模量和路基模量的关联度最大;相对于传统的BP神经网络方法,GRA方法的引入提升了沥青路面各结构层模量反演结果的精度和稳定性.研究成果为利用FWD测试技术诊断与评价沥青路面各结构层的性能状态提供了新视角.To enhance the reliability of back-calculation moduli of asphalt pavements,dynamic finite element analysis model was established to develop a database linking deflection basin values and pavement layer moduli,and then series of back-calculation models of asphalt pavement moduli were built by combing the grey relational analysis(GRA) with the back-propagation(BP) neural network. The results demonstrated that the three different kinds of offset distances(0,60,210 cm) were most related to surface modulus,base modulus and subgrade modulus,respectively. The combining analysis method could improve the accuracy and stability of the back-calculation modulus when compared to the traditional BP neural network model. The research would provide a new perspective for using FWD testing technology to evaluate the structural performance of asphalt pavements.

关 键 词：道路工程 沥青路面 模量反演 BP神经网络 灰色关联度 

分 类 号：U414[交通运输工程—道路与铁道工程]