Scholarlink

Buckling restrained braces (BRBs) are widely used as seismic protection devices in structures to enhance lateral strength, stiffness and dissipate seismic energy. The estimation of cumulative plastic deformation capacity (CPD) is considered vital to investigate their plastic deformation performance under cyclic loading. This study adopted a data-driven approach to predict the CPD capacity of BRBs using synthetic data generated via generative adversarial networks (GAN) technique to overcome the limited literature data (only 193 samples). Four machine learning (ML) algorithms, such as Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), Gradient Boosting (GB), and Random Forest (RF) were trained using the synthetic data (772 samples) and tested utilizing 193 samples of experimental data based on ‘Train on Synthesized - Test on Real data’ approach. Four evaluation metrics, the coefficient of determination (R2), mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE) were utilized to evaluate the performance of models. The results indicated that the developed models achieved prominent predictive performance, with CPD predictions corresponding well to the testing samples. The CatBoost model provided better, and more reliable results, achieving R2 of 0.98 compared to GB, RF, and the XGBoost. Finally, the SHapley Additive exPlanations (SHAPs) algorithm was used to estimate the relative importance of the variables affecting BRB’s CPD capacity.

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

Data-Driven Prediction of Cumulative Plastic Deformation for Buckling Restrained Braces Using Generative Adversarial Networks (Gan) Synthetic Data Augmentation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hubdar Hussain,
  • John Mark Go Payawal,
  • Dong-keon Kim

摘要

Buckling restrained braces (BRBs) are widely used as seismic protection devices in structures to enhance lateral strength, stiffness and dissipate seismic energy. The estimation of cumulative plastic deformation capacity (CPD) is considered vital to investigate their plastic deformation performance under cyclic loading. This study adopted a data-driven approach to predict the CPD capacity of BRBs using synthetic data generated via generative adversarial networks (GAN) technique to overcome the limited literature data (only 193 samples). Four machine learning (ML) algorithms, such as Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), Gradient Boosting (GB), and Random Forest (RF) were trained using the synthetic data (772 samples) and tested utilizing 193 samples of experimental data based on ‘Train on Synthesized - Test on Real data’ approach. Four evaluation metrics, the coefficient of determination (R2), mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE) were utilized to evaluate the performance of models. The results indicated that the developed models achieved prominent predictive performance, with CPD predictions corresponding well to the testing samples. The CatBoost model provided better, and more reliable results, achieving R2 of 0.98 compared to GB, RF, and the XGBoost. Finally, the SHapley Additive exPlanations (SHAPs) algorithm was used to estimate the relative importance of the variables affecting BRB’s CPD capacity.