土壤压实参数智能预测方法研究  

Research on Intelligent Prediction Method of Soil Compaction Parameters

作　　者：唐亮 张晓飞 李鸿钊 张文俊 郑军涛 刘广波 Tang Liang;Zhang Xiaofei;Li Hongzhao;Zhang Wenjun;Zheng Juntao;Liu Guangbo(Shandong Luqiao Group Co.Ltd.,Ji′nan,Shandong 250014;Shandong Transportation Institute,Ji′nan,Shandong 250031)

机构地区：[1]山东省路桥集团有限公司,山东济南250014 [2]山东省交通科学研究院,山东济南250031

出　　处：《江西建材》2024年第12期292-295,共4页

基　　金：山东省交通运输科技项目《高速公路路基韧性提升与智能压实综合评价技术研究》(项目编号:2023B45)

摘　　要：最大干密度和最佳含水率是路基压实的两个重要参数,影响路基的稳定性和承载力,从而影响公路和桥梁等基础设施的稳定性和耐久性。因此,对最大干密度和最佳含水率进行有效预测非常重要。文中采用皮尔森相关性分析,塑限与最大干密度、最佳含水率的相关性最高,其次为液限、细粒含量,而比重与最大干密度、最佳含水率无关。因此,文中采用随机森林(RF)、支持向量机(SVM)、K近邻算法三种机器学习算法建立以塑限、液限、细粒含量为输入,最大干密度和最佳含水率为输出的预测模型。经过训练和测试,文中比较了三种算法的最佳模型。结果表明,随机森林模型(RF)是预测最大干密度和最佳含水率的最佳模型,预测最大干密度时的相关系数R2为0.894 5,平均绝对误差(MAE)为0.162 5,最佳含水率R2为0.886 5,平均绝对误差(MAE)为0.001 2。利用机器学习模型对土壤压实参数进行预测,可以提高工程建设的效率,对施工具有一定的指导意义。Maximum dry density and optimal moisture content are two important parameters of subgrade compaction,which affect the stability and bearing capacity of subgrade,and thus affect the stability and durability of infrastructure such as highways and Bridges.Therefore,it is very important to effectively predict the maximum dry density and the best moisture content.Pearson correlation analysis was used in this paper.The correlation between plastic limit and maximum dry density and optimal water content was the highest,followed by liquid limit and fine particle content,while specific gravity had no correlation with maximum dry density and optimal water content.Therefore,in this paper,three machine learning algorithms,Random Forest(RF),support vector machine(SVM)and K-nearest neighbor algorithm,were used to establish a prediction model with plastic limit,liquid limit and fine particle content as inputs and maximum dry density and optimal moisture content as outputs.After training and testing,the best models of the three algorithms are compared in this paper.The results show that random forest model(RF)is the best model to predict the maximum dry density and the best water content.The correlation coefficient R2 is 0.8945,the mean absolute error(MAE)is 0.1625,the R2 of the best water content is 0.8865,the mean absolute error(MAE)is 0.0012.Using machine learning model to predict soil compaction parameters can improve the efficiency of engineering construction and has certain guiding significance for construction.

关 键 词：土壤压实参数 最大干密度 最佳含水率 机器学习 

分 类 号：U447[建筑科学—桥梁与隧道工程]