|
Title |
Prediction of Backbone Curves of Reinforced Concrete Columns Using a Conditional Generative Model
|
|
Authors |
송영민(Youngmin Song) ; 신지욱(Jiuk Shin) |
|
DOI |
https://doi.org/10.4334/JKCI.2026.38.1.051 |
|
Keywords |
철근콘크리트 기둥; 포락선; 머신러닝; 조건부 생성 모델; GAN reinforced concrete column; backbone curve; machine learning; conditional generative model; GAN |
|
Abstract |
Recent earthquakes have led to severe column damage in many existing reinforced concrete (RC) buildings with seismically deficient detailing. To mitigate such damage, a rapid seismic performance evaluation method is needed. However, conventional numerical simulations and experiment-based approaches are often limited due to their complexity. This study proposes a cGAN model in which an improved VAEGAN structure was implemented to rapidly predict lateral resisting capacities of RC rectangular section columns in the form of backbone curves. The effects of structural parameters on lateral resistance were considered, and eight input variables were selected, including material properties, geometric conditions, reinforcement details, and axial load. A total of 171 data samples generated from previous experimental studies were utilized to train and test the proposed model. The generated backbone curves from the proposed model have good correlation with the actual experimental data within slight variations (average error=8.15 %, MSE=0.13 %, R2=0.92) in the test dataset. The well-validated prediction model can serve as an efficient tool to rapidly estimate the lateral resisting capacities of RC columns and establish appropriate retrofit strategies using simplified structural information.
|