基于梯度提升模型的圆钢管混凝土轴压强度预测  

Prediction of Axial Compressive Strength of Circular Concrete-Filled Steel Tubes Using Gradient Boosting-Based Machine Learning Models

作　　者：林宽宏 LIN Kuanhong

机构地区：[1]福建省工业设备安装有限公司,福建福州350001

出　　处：《福建建设科技》2025年第2期20-23,共4页Fujian Construction Science & Technology

摘　　要：钢管混凝土受力性能复杂,而力学分析模型简化了部分关键因素,导致轴压强度预测准确性不足。机器学习技术通过深度挖掘数据输入与输出之间的非线性关系,为复杂工程问题的数据处理与分析提供了全新的解决方案。本研究基于2553个圆形钢管混凝土构件的轴心受压试验数据,采用梯度提升决策树(GBDT)和轻量梯度提升机(LightGBM)两种机器学习算法,对圆钢管混凝土的轴压强度进行了预测分析。结果表明,两种机器学习模型的大多数预测数据点与实际值的偏差均在20%以内,且训练集和测试集的决定系数均超过0.94,表明这两种模型能够满足圆钢管混凝土轴压强度预测的精度要求。此外,GBDT模型的预测精度略优于LightGBM,因此推荐GBDT模型用于预测圆钢管混凝土轴压承载力。The mechanical behavior of concrete-filled steel tubes(CFST)is highly complex,and traditional mechanical models often simplify critical factors,resulting in limited accuracy in predicting axial compressive strength.Machine learning(ML)techniques offer a novel approach by uncovering nonlinear relationships between input and output data,providing new insights into data processing and analysis for complex engineering problems.This study utilizes a dataset comprising 2,553 axial compression test results of circular CFST specimens and employs two gradient boosting-based ML algorithms,Gradient Boosting Decision Trees(GBDT)and Light Gradient Boosting Machine(LightGBM),to predict their axial compressive strength.The results demonstrate that the majority of the predicted data points from both models fall within a 20%deviation from the actual values,with coefficients of determination(R 2)exceeding 0.94 on both the training and testing datasets.These findings indicate that both models achieve satisfactory prediction accuracy for the axial compressive strength of circular CFSTs.Moreover,the GBDT model exhibits slightly higher prediction accuracy than LightGBM,making it the better model for predicting the axial compressive capacity of circular CFSTs.

关 键 词：钢管混凝土 轴压强度 机器学习 梯度提升模型 

分 类 号：TU398[建筑科学—结构工程]