摘要
为探索沥青混合料材料组成设计新思路,通过构建材料性能数据库,提出了一种数据驱动下的沥青混合料材料组成设计方法。首先以沥青种类、空隙率等沥青混合料材料组成作为模型输入,以动态模量、动稳定度等混合料性能作为模型输出,分别构建了反向传播(Back propagation, BP)神经网络、极限梯度提升(Extreme gradient boosting, XGBoost)和随机森林(Random forest, RF)三种机器学习预测模型,并建立了MySQL材料性能数据库;然后运用数据库结构化SQL查询语句,以沥青混合料性能为搜索条件,通过反向匹配获得沥青混合料材料组成参考值,提出了数据驱动下的沥青混合料材料组成设计方法;最后通过实例分析验证了设计方法的可行性。研究结果表明:三种机器学习模型中XGBoost较BP神经网络决定系数R^(2)提升了0.03~0.40,较RF提升了0.01~0.08,适用于沥青混合料性能预测;并通过实例分析与室内试验相结合的方法,验证了基于数据库反向匹配实现沥青混合料材料组成设计的可行性。研究成果可为揭示沥青混合料与其材料组成之间的内在关联、指导沥青混合料材料组成设计提供重要参考。
To explore the new ideas of asphalt mixture material composition design,the work proposes a data-driven asphalt mixture material composition design method based on the material performance database.Firstly,three machine learning prediction models,namely,BP neural network,XGBoost and Random forest,were constructed with asphalt type,air void ratio and other asphalt mixture material composition as model inputs,and dynamic modulus,dynamic stability and other mixture properties as model outputs,and a MySQL material performance database was established with the help of them.Then,using the structured SQL query statement of the database,the asphalt mixture performance is used as the query condition,and the database is reverse-matched to obtain the reference value of the material composition of asphalt mixtures,to put forward the data-driven design method of asphalt mixture material composition.Finally,the feasibility of the design method is verified by example analysis.The results show that XGBoost has the best prediction ability among the three machine learning models,and its coefficient of determination R^(2) is improved by 0.03—0.40 compared with BP neural network,and 0.01—0.08 compared with RF,which can be better used for predicting the performance of asphalt mixtures;and the feasibility of realizing the design of asphalt mixture composition based on reverse database matching is verified through the combination of case study and indoor experiments.The results of the study provide important references for revealing the performance correlation between asphalt mixtures and their material composition and guiding the design of asphalt mixture material composition.
作者
刘朝晖
盛佳豪
柳力
LIU Zhaohui;SHENG Jiahao;LIU Li(School of Traffic and Transportation Engineering,Changsha University of Science and Technology,Changsha 410114,China)
出处
《材料导报》
北大核心
2025年第4期71-79,共9页
Materials Reports
基金
国家重点研发计划(2021YFB2601000)
长沙市杰出创新青年培养计划项目(kq2306009)
国家自然科学基金(52208423,52278437)
湖南省教育厅科学研究重点项目(24A0224)
长沙理工大学研究生科研创新项目(CSLGCX23019)。
关键词
道路工程
沥青混合料
材料组成设计
机器学习
数据库
road engineering
asphalt mixture
material composition design
machine learning
database
