冬季桥梁沥青路面路表温度实时预估模型

A Real-time Estimation Model of Road Surface Temperature of Bridge Asphalt Pavement in Winter

桥梁段是冬季高速公路最易结冰的隐患路段之一。针对桥梁路面结冰预警急需解决的路表温度实时预估问题,以云南高原山区麻(柳湾)-昭(通)高速公路不同环境特征、结构形式桥梁段为研究路段,以自动交通气象站采集的2019—2021年路表温度和常规气象实测数据为基础。首先,利用统计方法对不同环境特征、结构形式桥梁段的路表温度与气象因素进行了相关性分析,确定了各气象因素与桥梁段路表温度之间的关系。其次,基于随机森林模型,构建了适用于各桥梁段的冬季路表温度实时预估模型。最后,对路表温度预估模型进行了交互性的迁移试验。结果表明:不同环境特征及类型桥梁的气象因素与路表温度间的相关系数有着相似规律,均与大气温度显著正相关,均与湿度显著负相关,但风速、风向和冰层厚度等气象因素取值差异仍较大,区域差异显著;所构建的路表温度预估模型有较好的回归预估性能,在不同桥梁段上的平均绝对误差分别为0.637,0.282,0.516℃,预估结果优于线性回归、支持向量机、BP神经网络模型;模型在1℃以下的易结冰温度段表现优异,适用于冬季路表温度预估;环境特征对模型影响较大,应引入表征环境的指标以提升模型的可迁移性。

The bridge section is one of the most vulnerable sections of highways to ice formation in winter.In view of the real-time estimation of road surface temperature for prewarning of pavement icing,taking the bridge sections with different environmental characteristics and structural forms in Maliuwan-Zhaotong expressway in the mountainous areas of Yunnan plateau as the research sections,based on the road surface temperature and conventional meteorological measurement data collected by the automatic transport meteorological station from 2019 to 2021,first,the correlations between road surface temperature and meteorological factors of the bridge sections with different environmental characteristics and different structural forms are analyzed,and the relationships between various meteorological factors and the surface temperature of the bridge sections are determined.Second,based on the random forest model,a real-time estimation model of winter road surface temperature for each bridge section is constructed.Finally,the interactive transfer experiment on the road surface temperature estimation model is conducted.The result shows that(1)The correlation coefficients between meteorological factors and road surface temperature of different environmental characteristics and of bridge types have similar patterns,with significant positive correlation with atmospheric temperature and significant negative correlation with humidity.However,there are still significant differences in the values of meteorological factors such as wind speed,direction and ice thickness,with significant regional differences.(2)The constructed road surface temperature estimation model has good regression estimation performance.The average absolute error on different bridge sections is 0.637,0.282,0.516℃respectively.The estimation result is better than those of linear regression,support vector machine and BP neural network models.(3)The model performs well in the ice prone temperature range below 1℃and is suitable for estimating road surface temperature in winter.Environmental characteristics have great influence on the model,and the indicators representing the environment should be introduced to improve the transferability of the model.