基于机器学习的海相软土压缩特性参数预测研究  

Study on Prediction of Compression Parameters of Marine Soft Soil Based on Machine Learning

作　　者：斯宇航 林泓语 SI Yuhang;LIN Hongyu(Key Laboratory of Road and Traffic Engineering of the Ministry of Education,Tongji University,Shanghai 201800,China)

机构地区：[1]同济大学道路与交通工程教育部重点实验室,上海201800

出　　处：《交通科技》2025年第2期27-32,共6页Transportation Science & Technology

摘　　要：为实现对于海相软土压缩特性参数的智能高效预测,文中采用机器学习方法,首先建立包含932组样本的海相软土土体参数数据库,其中主要包括取样深度D、含水率W_(0)、湿密度ρ、干密度ρ_(d)等变量,以及代表压缩特性参数的压缩系数a_(v)和压缩模量E_(s)。通过Spearman相关性分析获取了各变量之间的相关性关系,然后采用支持向量机(SVM)、BP神经网络(BPNN)、随机森林(RF)以及极限提升树(XGBoost)结合贝叶斯超参数优化对海相软土压缩特性参数展开预测研究。结果表明,机器学习方法能够有效预测土体压缩特性参数,其中XGBoost在训练集与测试集上均取得了最优的表现,测试集上对2种参数预测的相关系数(R^(2))均大于0.8,分别为0.860和0.801;同时模型在预测实测值较小的压缩特性参数的性能优于较大的压缩特性参数。To achieve intelligent and efficient prediction of compression parameters for marine soft soil,machine learning methodsare adopted.Firstly,a marine soft soil parameter database containing 932 sets of samples is established,which mainly includes variables such as sampling depth D,moisture content W_0,wet density ρ,dry density ρ_d,as well as compression coefficient a_(v) and compression modulus E_(s) representing compression parameters.Spearman correlation analysis is used to obtain the correlation relationship between variables.Then,support vector machine(SVM),BP neural Network(BPNN),random forest(RF),and extreme boosting tree(XGBoost) combined with Bayesian hyperparameter optimizationare used to study the prediction of compression parameters in marine soft soil.The research results indicate that machine learning methods can effectively predict soil compression parameters,with XGBoost achieving the best performance on both the training and testing sets.The correlation coefficients(R~2) for predicting the two parameters on the testing set reach over 0.8,which are 0.860 and 0.801,respectively.At the same time,the model performs better in predicting compression parameters with smaller measured values than larger ones.

关 键 词：海相软土 压缩特性参数 机器学习 相关性分析 

分 类 号：U419[交通运输工程—道路与铁道工程]