基于宏微观纹理特征融合的路面摩擦性能预测

Macro and micro texture based prediction of pavement surface friction

通过非接触式三维激光表面测试、机器学习,开展基于宏微观纹理特征融合的路面摩擦性能智能预测模型研究.来自俄克拉荷马州的45个测试站点被选取作现场测试平台.利用三维激光检测车和GripTester,分别获取行车道轮迹带路面摩擦数据、宏观纹理;利用LS-40三维激光表面分析仪获取集料表面三维微观纹理数据,测算4类微观纹理参数.利用机器学习算法,将路面摩擦与宏微观纹理特征建立联系.综合模型训练与测试,评价路面摩擦性能预测模型的准确率.模型的测试标准差为0.047,测试集R2为0.865.研究结果表明,86.5%的测试数据适用于所建立的机器学习预测模型,开发的评价指标及预测模型能够较好地预测路面摩擦性能.

The pavement skid resistance prediction model was analyzed based on macro and micro texture fusion using non-contact three-dimensional laser detection and machine learning.45 pavement sites in Oklahoma were identified as the testing beds.Pavement skid resistance and surface macro-texture data were collected in parallel at highway speeds using a grip tester and three-dimensional(3D)laser dection vehicle.Four types of 3D aggregate parameters were calculated to characterize the micro-texture of aggregate surface using LS-403D laser imaging scanner.Relationship between pavement surface friction and texture was analyzed using machine learning model.The accuracy of the developed model was verified by model training and testing.The standard deviation of the model was 0.047,and the R squared value of the model was 0.865.86.5%of the testing data fit the proposed friction model.Results show that the developed texture parameters and proposed friction prediction model can predict the pavement surface friction well.