Scholarlink

In contemporary structural engineering, concrete-filled steel tube (CFST) columns are highly regarded for their superior strength, ductility, and fire resistance. Evaluating the ultimate axial compressive load capacity (UACLC) of these columns is essential but challenging due to the complex interactions between material and geometric variables. This investigation explored the application of decision tree (DT) models to predict the UACLC of CFST columns. The research utilized a comprehensive dataset of 932 experimental samples for model training. Key hyperparameters, including tree depth, minimum leaf size, and learning rate, were systematically optimized to enhance model performance. The optimal model was achieved with a learning rate of 0.2, a tree depth of 10, and a minimum leaf size of 1. This optimized model demonstrated an R2 value of 0.999 during training and 0.982 in testing. The model's accuracy was further corroborated by low RMSE values (0.179 for training and 0.561 for testing). The study also examines the integration of the model's outcomes into practical design guidelines, emphasizing its interpretability and potential for improving structural design and safety. This research contributes to the expanding field of machine learning applications in structural engineering, demonstrating that optimized DT models provide a reliable and interpretable tool for predicting CFST column behavior.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Optimizing Decision Tree Models for Accurate Prediction of Ultimate Axial Compressive Load Capacity of CFST Columns

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Megha Gupta,
  • Satya Prakash,
  • Sufyan Ghani

摘要

In contemporary structural engineering, concrete-filled steel tube (CFST) columns are highly regarded for their superior strength, ductility, and fire resistance. Evaluating the ultimate axial compressive load capacity (UACLC) of these columns is essential but challenging due to the complex interactions between material and geometric variables. This investigation explored the application of decision tree (DT) models to predict the UACLC of CFST columns. The research utilized a comprehensive dataset of 932 experimental samples for model training. Key hyperparameters, including tree depth, minimum leaf size, and learning rate, were systematically optimized to enhance model performance. The optimal model was achieved with a learning rate of 0.2, a tree depth of 10, and a minimum leaf size of 1. This optimized model demonstrated an R2 value of 0.999 during training and 0.982 in testing. The model's accuracy was further corroborated by low RMSE values (0.179 for training and 0.561 for testing). The study also examines the integration of the model's outcomes into practical design guidelines, emphasizing its interpretability and potential for improving structural design and safety. This research contributes to the expanding field of machine learning applications in structural engineering, demonstrating that optimized DT models provide a reliable and interpretable tool for predicting CFST column behavior.