【摘要】 型钢混凝土构件在高层建筑中广泛应用，其出色的变形能力对结构抗震性能有着至关重要作用，而箍筋设计与型钢混凝土构件变形能力息息相关。当前国内外规范中箍筋设计多采用经验模型，准确性较低，难以准确捕捉配箍率和不同影响变量之间的非线性关系，因此提出了1种基于机器学习的型钢混凝土柱配箍预测模型。基于从国内外文献中收集到的626组型钢混凝土柱的试验数据，通过数据清洗和特征工程确定了5个输入特征，从12种算法中选出性能最强算法——极限梯度提升算法，采用贝叶斯优化进行超参数调优，得到最终的配箍预测模型(R~2=0.878)。与经验模型相比，机器学习模型具有更高的精度和更低的离散性。进行安全概率分析以获得用于实际工程的参考。为型钢混凝土柱箍筋设计提供了1种更准确的工具，并为既有建筑的型钢混凝土柱构造措施评估和改造提供了参考。更多还原

【Abstract】 Steel-reinforced concrete(SRC) members are widely used in high-rise buildings, and their excellent ductility deformation capacity plays a vital role in seismic performance. Stirrup design is closely related to the ductility of SRC members. Current design codes mostly rely on empirical models, which are less accurate and struggle to capture the nonlinear relationship between stirrup ratio and various influencing variables. Therefore, this paper proposes a machine learning-based prediction model for stirrup configuration in SRC columns. Based on 626 experimental datasets collected from domestic and international literature, five input features are determined through data cleaning and feature engineering. The extreme gradient boosting algorithm, selected as the best performer among 12 candidates, is optimized using Bayesian optimization for hyper-parameter tuning, resulting in a final prediction model with R~2=0.878. Compared with empirical models, the machine learning model demonstrates higher accuracy and lower dispersion. Probabilistic safety analysis is conducted to provide references for practical engineering applications. Overall, this study offers a more accurate tool for stirrup design in SRC columns and provides a reference for evaluating and retrofitting existing structures.更多还原