Optimizing Seismic Performance Assessment: A Web-Based Enhanced Visual Screening Method Integrated with Machine Learning for Reinforced Concrete Structures

Seismic vulnerability assessment of reinforced concrete (RC) structures is crucial in earthquake-prone regions to mitigate risks to life and property. This study proposes a systematic three-phase framework for enhanced seismic risk assessment: (1) Automation, (2) Evaluation, and (3) Predictive Modeling. For the Automation Phase, a web-based tool was developed to digitize and streamline the Turkish Rapid Visual Screening (RVS) procedure, eliminating manual calculation errors while improving efficiency. During the Evaluation Phase, we applied this tool to assess 600 buildings, classifying them into four distinct risk categories (no, low, moderate, and high risk) through standardized scoring. Finally, in the Predictive Modeling Phase we conducted correlation analysis to identify key seismic risk factors (e.g., building height showing a strong negative correlation, while soft-story mechanisms and short columns emerged as critical vulnerabilities) and implemented three machine learning models (XGBoost, Random Forest, and AdaBoost) for risk prediction, with XGBoost achieving superior accuracy. The framework’s validation confirmed the web tool’s reliability relative to conventional methods while revealing most buildings as low-risk, demonstrating how this integrated approach—combining automated screening, large-scale assessment, and data-driven prediction—provides a scalable solution for seismic risk mitigation in vulnerable regions.