基于机器学习的横向裂缝修复性养护方案后评价

Post-Evaluation of Repairable Lateral Crack Maintenance Scheme Based on Machine Learning

摘要:

针对江苏省高速公路沥青路面横向裂缝快速增长的问题，采用机器学习的方法对2018—2021年间徐淮高速、京台高速、汾灌高速和新扬高速的大修工程数据进行分析。研究比较了单层铣刨预处理、双层铣刨预处理、冷再生预处理和注浆预处理4种不同的裂缝修复方案，采用“一缝一档”对裂缝的修复效果进行长期跟踪观测，并提出二次反射率作为方案的评价指标。结果表明，注浆预处理方案在观测期内未出现裂缝再次反射，表现出最佳的修复效果；而冷再生预处理方案的效果最差。通过CART分类决策树模型进一步验证了双层预处理方案对于贯穿单车道的横向裂缝具有较好的抑制效果，为高速公路沥青路面横向裂缝大修施工方案的选择提供了科学的指导意义。

Abstract:

Aiming at the problem of rapid growth of lateral cracks on the asphalt pavement of expressways in Jiangsu Province, the machine learning methods are used to analyze the overhaul engineering data of Xuhuai Expressway, Jingtai Expressway, Fengguan Expressway and Xinyang Expressway from 2018 to 2021. Four different crack repairing schemes of single-layer milling pretreatment, double-layer milling pretreatment, cold recycling pretreatment and grouting pretreatment are studied and compared. The “one crack and one record” method is used to track the effects of crack repair for the long time. And the secondary reflectance is proposed an evaluation index of the schemes. The results show that in the pretreatment scheme of grouting, no rereflection of cracks occur during the observation period with the best repair effect. The cold recycling pretreatment scheme has the worst effect. The CART classification decision tree model is used to further verify that the double-layer pretreatment scheme has a good inhibition effect on the lateral crack running throughout the single lane, which provides the scientific guidance for the selection of overhaul construction schemes for lateral cracks on the expressway asphalt pavement.