基于PSO-LSTM模型的沥青路面PCI值预测研究  

Research on Road PCI Value Prediction based on PSO-LSTM Model

作　　者：乔建刚 范颖蓉 王彦萍 李恩玉 QIAO Jian-gang;FAN Ying-rong;WANG Yan-ping;Ll En-yu(School of Civil Engineering and Transportation,Hebei University of Technology,Tianjin 300401,China;Shanxi Road&-Bridge Construction Group Company Limited Yangquan Branch,Yangquan 045000,China)

机构地区：[1]河北工业大学土木与交通学院,天津市300401 [2]山西路桥建设集团有限公司阳泉分公司,阳泉市045000

出　　处：《公路》2025年第4期52-59,共8页Highway

基　　金：山西省交通运输厅科技项目,项目编号2022-1-2。

摘　　要：为了及时掌握路面使用情况,保证国省道公路通行运输的舒适与安全,以实地调研山西省普通国省道沥青路面交通量、路面结构设计、路面养护等数据为基础,构建沥青路面损坏状况指数PCI指标预测模型。首先采用自变量相关分析法从16个影响路面使用性能因素中确定14个因素为自变量,通过随机森林模型进行影响因素重要度分析,以模型输出的重要度系数前60%的8个主要因素为路面PCI预测模型数据集,构建基于长短时记忆网络模型的路面PCI预测模型;之后,选用粒子群算法优化预测模型参数以提高模型精度。研究表明,使用PSO-LSTM预测模型对山西省普通国省道沥青路面PCI值的预测结果误差小、精度高,为山西省普通国省道的养护决策提供了一种研究思路。In order to timely grasp the service conditions of road surfaces and ensure the comfort and safety of transportation on national and provincial highways,a PCI index prediction model for asphalt pavement performance is constructed based on field investigations of the traffic volume,pavement structure design,and pavement maintenance data of ordinary national and provincial highways in Shanxi Province.Firstly,the independent variable correlation analysis method is used to determine 14 out of 16 factors that affect road performance as independent variables.The importance of influencing factors is analyzed using a random forest model,and the top 8 main factors with the top 60%of the model's output importance coefficients are used as the road PCI prediction model data set.A road PCI prediction model based on a long short-term memory network model is constructed;Afterwards,the particle swarm optimization algorithm is used to optimize the prediction model parameters so as to improve the accuracy of the model.Finally,the PSO-LSTM prediction model for the PCI value of asphalt pavement on ordinary national and provincial roads in Shanxi Province is determined,providing a research approach for the maintenance decision-making of ordinary national and provincial roads in Shanxi Province.

关 键 词：沥青路面 路面使用性能影响因素 随机森林 PCI预测 PSO-LSTM模型 

分 类 号：U416.217[交通运输工程—道路与铁道工程]