Data modeling analysis of GFRP tubular filled concrete column based on small sample deep meta learning method

Area of Science:

• Engineering Data Analysis
• Material Science
• Machine Learning

Background:

• Small-sample regression is a challenge in engineering data analysis.
• Traditional regression models struggle with limited datasets.
• Meta-learning offers a promising avenue for improving model performance with scarce data.

Purpose of the Study:

• To develop a meta-learning method for small-sample regression in engineering applications.
• To enhance deep neural networks through optimization-based data augmentation.
• To optimize the performance of glass fiber reinforced plastic (GFRP) for wrapping concrete short columns.

Main Methods:

• Integration of traditional regression models with meta-learning.
• Application of optimization-based data augmentation techniques.
• Development of a deep neural network architecture.

Main Results:

• The proposed meta-learning method demonstrated superior performance in small-sample regression tasks.
• Outperformed traditional models like Support Vector Regression (SVR), Gaussian Process Regression (GPR), and Radial Basis Function Neural Networks (RBFNN).
• Successfully optimized glass fiber reinforced plastic (GFRP) for concrete short column wrapping.

Conclusions:

• Deep learning, particularly meta-learning, shows significant potential for handling limited data in material analysis.
• The developed method offers a new approach for engineering data analysis with small sample sizes.
• This research opens new opportunities in material data analysis and engineering applications.