基于BP神经网络和Logistic回归分析的沥青路面平整度模型预测  

Predictive Modelling of Asphalt Pavement Roughness Using BP Neural Network and Logistic Regression Analysis

作　　者：赵飞 施发垦 陈利 ZHAO Fei;SHI Fa-ken;CHEN Li(CCCC Xi’an Road Construction Machinery Co.,Ltd.,Xi’an 710020,Shaanxi,China)

机构地区：[1]中交西安筑路机械有限公司

出　　处：《筑路机械与施工机械化》2019年第7期110-114,共5页Road Machinery & Construction Mechanization

摘　　要：为了解决国际平整度指数IRI预测模型准确性不高等问题,以路面长期性能(LTPP)数据库实测数据为基础,采用机器学习中BP神经网络建模方法对提取出的数据进行预测分析,并与传统的Logistic回归分析结果做评估对比。分析结果表明,利用传统的Logistic方法和BP神经网络得出的复判定系数分别为0.731、0.876,说明采用的BP神经网络平整度预测模型具有较高效率和预测精度以及较低的复杂度,能够有效评价路面使用性能。In order to solve the problem of low accuracy of the international roughness index(IRI)prediction model,based on measured data of the long-term pavement performance(LTPP),the BP neural network modelling method in machine learning was used to predict and analyze the extracted data,which would be compared with the results of the traditional Logistic regression analysis.The results show that the complex decision coefficients obtained by the traditional Logistic method and BP neural network are 0.731 and 0.876,respectively,which indicates that the roughness prediction model using BP neural network has the advantages of high efficiency,prediction accuracy and low complexity.The prediction model is capable of evaluating the pavement performance effectively.

关 键 词：交通运输工程 路面长期性能 LOGISTIC模型 国际平整度指数 

分 类 号：U416.06[交通运输工程—道路与铁道工程]