Concrete-Filled Steel Tubular Specimen Tensile Performance with GAN-Augmented Data

Six experimental specimens and ten parametric models form the basis for analyzing square concrete-filled steel tubular members under axial tension. A Generative Adversarial Network (GAN) was used to augment the data, improving predictive model accuracy on an augmented test set. The dataset, created by Hongtao Zhang and last updated in June 2026, includes finite element simulation results, parametric analyses, and machine learning predictions.

Use Cases

• Training machine learning models to predict ultimate load based on confinement coefficient and slenderness ratio.
• Comparing finite element simulation results against experimental data and code-based predictions.
• Assessing the effectiveness of WGAN-GP data augmentation for improving model performance with limited physical test data.

Strengths

• Includes results from six physical specimens and ten parametric models, providing a basis for validation.
• Machine learning models trained on GAN-augmented data achieved high predictive accuracy, with R² scores of 0.997 and 0.9855.
• Finite element simulations showed good agreement with test results, with simulation-to-test ratios all below 0.95.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 5.5 KB, indicating limited scope.

Provenance

Source

Hongtao Zhang via figshare

Collection Method

Based on experimental results of six specimens, with further parametric finite element modeling and machine learning predictions.

Freshness

Last updated 2026-06-01 17:33:25; freshness should be verified.