基于机器学习的不锈钢管混凝土构件轴压承载力预测方法研究

Prediction of Axial Compression Capacity of Concrete-Filled Stainless Steel Tubular Members by Machine Learning

基于收集的189个不锈钢管混凝土构件的轴压试验数据建立了机器学习数据库,采用6种机器学习模型(随机森林、决策树、支持向量机、多层感知器、极端梯度提升和自适应提升)分别对圆形和方形截面不锈钢管混凝土构件进行了轴压承载力预测研究,同时分析了不同参数对机器学习模型预测精度的影响。研究结果表明,上述6种机器学习预测模型中预测精度最高的是极端梯度提升模型,对于圆形和方形截面构件所有参数的预测,该模型的均方根误差分别为0.0435和0.0188,且预测值与试验值之比的变异系数分别为0.127和0.166,相较于现有标准,该模型表现出更高的预测精度和更广泛的适用范围,可为工程应用提供数据支撑和理论指导。

In this paper,the axial compression experimental data of 189 concrete-filled stainless steel tubular members were collected to build a machine learning dataset.Six machine learning models(Random Forest,Decision Tree,SVM,MLP,XGboost and Adaboost)were used to predict the axial bearing capacity of circular and square concrete-filled stainless steel tubular members,and the influence of different parameters on the prediction accuracy of machine learning models was analyzed based on parameter analysis.In the six machine learning prediction models mentioned above,the XGBoost model exhibits the highest prediction accuracy.For the prediction of circular and square section members with all parameters,this model achieves root-meansquare errors of 0.0435 and 0.0188,respectively.Additionally,it demonstrates lower coefficient of variation values for the predicted value-to-experimental value ratio for circular(0.127)and square(0.166)section members.This model outperforms the current standards,offering improved prediction accuracy and a broader range of applicability,thereby providing valuable data support and theoretical guidance for engineering applications.