基于组合模型的沥青路面车辙预测
Prediction of Rutting of Asphalt Pavement Based on Combined Model
张琛 1汪海年 1王宠惠1
作者信息
- 1. 长安大学道路结构与材料交通行业重点实验室，西安 710064
- 折叠
摘要
为了对沥青路面车辙进行预测研究,结合陕西某高速公路沥青路面的车辙病害现状,以美国AASHTO新力学经验法中的Pavement-ME预测模型( PME模型)和灰色预测理论为基础,对PME预测模型的参数进行了本地化处理,提出了考虑数据时效的动态灰色预测模型DGM (1,1),并采用数理统计方法对车辙的发育趋势进行基于组合模型的预测研究.研究结果表明：PME模型和DGM灰色模型的拟合精度分别为0.963和0.941,可以作为组合模型的子模型；DGM-PME组合模型的预测精度要高于单一子模型的预测精度,其中采用误差平方和倒数法定权得到的组合模型的预测精度相对最高,其子模型权重分配为L12=0.601和L22=0.399,预测精度最高为0.0045；DGM-PME组合模型在个别时期的预测效果不如单一子模型,属于正常的误差现象；用DGM-PME组合模型来替代单一子模型进行沥青路面车辙的预测研究是可行的,本文所建立的组合模型适用于项目级沥青路面的车辙发育趋势预测.
Abstract
To predict the rutting of asphalt pavement, Pavement-Me prediction model ( PME) with the localized input parameter and a dynamic grey prediction model (DGM (1,1)) with data timeliness were established based on the PME and grey prediction theory, in the current rutting situation in a highway of Shaanxi Province. The mathematical statistic method was used to conduct the prediction research of rutting based on a combined model. The results reveal that the fitting precision of PME model is 0. 963, and the DGM (1,1) model is 0. 941, which can be used as a submodel of the combined model. The prediction precision of DGM-PME combined model is higher than that of single submodel, in which the combined model which adopts error square and reciprocal weighting has the highest prediction precision, and its rank allocation is L12 =0. 601 and L22 =0. 399, the maximum of prediction precision can be 0. 0045 . The prediction result of the combined model at some periods is inferior to that of some single prediction models, which belongs to normal error range. It is feasible to replace the single model with the DGM-PME combined model to predict the rutting of asphalt pavement. And it is more suitable to use the combined model established in this paper to predict rutting development tendency for the project-level asphalt pavement.
关键词
沥青路面/车辙预测/Pavement-ME/灰色预测理论/组合模型Key words
asphalt pavement/rutting prediction/Pavement-ME/grey prediction theory/combined model引用本文复制引用
基金项目
国家科技支撑计划项目(2014BAG05B04)
交通运输部应用基础研究项目(2014319812180)
长安大学优秀博士学位论文培育项目(310821150010)
出版年
2016
NETL
NSTL
维普
万方数据