面板数据马尔可夫预测沥青路面性能衰变方法

Method for Asphalt Pavement Deterioration Forecasting Based on Panel-data Markov

针对辽宁中北部地区公路沥青路面使用性能的检测数据,将路面结构、交通等级相同的归为同一子集,根据面板数据原理,用E-View软件计算某子集衰变函数的截距与系数,用MATLAB计算路面性能衰变的马尔可夫状态转移矩阵,进而对某路段所处路况的概率进行预测.结果表明:同一子内集的各路段样本路面性能衰变规律具有相似性.利用同一子集内数据的样本数量(本文以34组为例)的增多,可以弥补数据历时年份少(本文以3为例)的不足.为检验预测精确度,将测试样本的2014年预测数据与2014年实测数据相比较,误差均在±5之内,可实现在数据累积年份不足时沥青路面使用性能的预测.

Markov models of Panel-data were proposed to resolve performance forecasting on asphalt pavements. The data was detected from north and middle region of Liaoning province. Intercept and coefficient of the disintegration function were calculated by a particular software named E-View which was based on the theory of panel- data. Markov state-transition matrix of calculating performance forecasting of asphalt pavement was conducted by MATLAB,in which the probability of the status of a specific road segment was predicted. The result shows that deterioration laws of each road segments in a same subset are similar. The problem of lack of data of accumulative years can be solved by utilizing the increasement of the sample data in a same subset. In order to test the prediction accuracy,the forecast data of the test samples were compared with the detection data of the same samples which were detected in 2014. The trial sample error was within ± 5 ,which means the performance forecasting of asphalt pavement can be predicted by this method when the data of accumulative years is insufficient.