【摘要】 为准确预估沥青路面使用性能衰变规律,提出了基于K最邻近非参数回归的预测方法,探索大数据挖掘技术在路面结构使用性能预测的应用。利用自然分区、交通量等级、面层类型及厚度、基层类型及厚度、路龄作为特征向量,将PQI值及其评价等级作为输出向量,构建了沥青路面结构使用性能KNN预测模型,并将模型应用于广东省普通国省道典型沥青路面结构使用性能预测中。结果表明,K=3时的PQI值预测精度优于K=5时的预测精度;K=3和K=5时的PQI值预测结果的平均绝对百分误差分别为0.737%和0.793%,均小于1%,说明K最近邻算法预测沥青路面使用性能的准确度较高。更多还原

【Abstract】 In order to predict accurately the attenuation of asphalt pavement performance, a prediction method of asphalt pavement performance based on K nearest neighbor(KNN) nonparametric regression is developed to explore the application of big data mining technology in pavement performance prediction. The KNN model for performance prediction of asphalt pavement is constructed by taking natural zoning, traffic volume grade, surface type and thickness, base type and thickness, road age as eigenvectors, pavement quality index(PQI)value and its evaluation grade as output vectors. Furthermore, the KNN model is applied to the performance prediction of typical asphalt pavement structure in common national and provincial trunk highways in Guangdong province. The results show that the prediction accuracy of the PQI value at K=3 is better than those of K=5; the average absolute percentage errors(MAPE) of the prediction results of the PQI values at K=3 and K=5 are 0.737% and 0.793% respectively, which are less than 1%. It shows that the KNN algorithm has a high accuracy in predicting the performance of asphalt pavement.更多还原