Scholarlink

This research investigates the improvement of artificial neural network (ANN) models in forecasting the maximum strength of concrete-filled steel tube (CFST) columns using mathematical transformations. We have successfully enhanced the distribution of features by applying several transformations such as logarithmic, reciprocal, square, square root, custom, Box-Cox, and Yeo-Johnson on a dataset of 663 CFST columns. This enhancement aims to increase the quality of input data for a five-layer perceptron ANN model. The square transformation demonstrates exceptional performance by significantly reducing training and validation losses during 1000 epochs, instilling confidence in its potential to enhance the model's generalization. In future research, we plan to extend the evaluation period to more accurately assess the consistency and effectiveness of the transformation, potentially improving the use of artificial neural networks in making predictions in structural engineering.

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

Predicting Limit Strength of CFST Columns: Impacts of Mathematical Transformers in ANN Models

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tran-Trung Nguyen,
  • Phu-Cuong Nguyen

摘要

This research investigates the improvement of artificial neural network (ANN) models in forecasting the maximum strength of concrete-filled steel tube (CFST) columns using mathematical transformations. We have successfully enhanced the distribution of features by applying several transformations such as logarithmic, reciprocal, square, square root, custom, Box-Cox, and Yeo-Johnson on a dataset of 663 CFST columns. This enhancement aims to increase the quality of input data for a five-layer perceptron ANN model. The square transformation demonstrates exceptional performance by significantly reducing training and validation losses during 1000 epochs, instilling confidence in its potential to enhance the model's generalization. In future research, we plan to extend the evaluation period to more accurately assess the consistency and effectiveness of the transformation, potentially improving the use of artificial neural networks in making predictions in structural engineering.