沥青路面结冰条件判别标准及SVM预测分析

Discriminative Criterion of Asphalt Pavement Icing Condition and Analysis on SVM Prediction

为了探讨沥青路面结冰与道路气象环境间的相关关系,实现路面结冰状态的准确预报,在分析影响沥青路面结冰的主要气象环境因素的基础上,探讨了不同降水类型与气象环境参数的关联性,建立了沥青路面路表温度预估分析模型。结合Norrman路面结冰打滑判别准则,提出了浙中地区冬季沥青路面结冰状态判别标准。通过支持向量机方法(SVM),以RBF函数作为模型核函数,构建了不同降水类型条件下以地表温度与气温为输入量的沥青路面结冰预测模型。结果表明:不同降水条件下沥青路面结冰时的日均气温、日均测点路表温度、日均风速、日均降水量以及日均相对湿度存在差异,其中,日均气温与日均测点路表温度变化差异较大,日均风速与日均降水量次之,日均相对湿度变化差异较小;通过气象环境监测数据可实现降水类型的间接判别和沥青路面路表温度的预估分析;不同结冰类型对车辆行驶安全的影响从高到低依次为雪降到严寒路面、雨降到严寒路面和雪降到温暖路面;使用SVM构建的路面结冰预测模型对路面结冰状态的预测效果较好,错报率低于6.5%,未出现结冰天气的漏报,同时具有良好的泛化能力,充分展现了SVM在道路气象预报领域的应用前景。

In order to discuss the correlation between icing of asphalt pavement and meteorological environment and implement an accurate prediction of icing condition of pavement,the correlation between precipitation types and meteorological factors is surveyed based on analyzing the main meteorological parameters effect on asphalt pavement icing, and the prediction analysis model of asphalt pavement temperature is established. Considering the road slipperiness discriminative rule presented by Norrman,the discriminative criterion of icing condition of asphalt pavement in central area of Zhejiang Province in winter is proposed. The prediction model of pavement surface icing prediction,which taking asphalt pavement surface temperature and air temperature as the input under the condition of different precipitation types,is built by the method of RBF based on SVM. The result demonstrates that（1） the daily average air temperature,daily average pavement surface temperature at measuring points,daily average wind speed,daily average rainfall and daily average relative humidity have differences under the condition of different precipitation types,ofwhich the variation features of daily average air temperature and daily average pavement surface temperature are quite obvious,those of daily average wind speed and daily average precipitation come next,and those of daily average relative humidity is similar;（ 2） the indirect discrimination of precipitation type and the prediction of pavement surface temperature could be accomplished by meteorological monitoring data;（3） the influence of pavement surface icing conditions on driving safety is in the descending order of snowing at cold surface,rainwater freezing at cold surface,and snowing at warm surface;（ 4） the SVM based prediction model of pavement surface icing has precise prediction result and the mis-reporting rate is below 6. 5% with no omissions,and its generalization ability is good,which fully demonstrates the application prospects of SVM in pavement weather prediction field.