A rapid multi-level framework for seismic damage assessment of reinforced concrete structures
Structures, cilt.92, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 92
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.istruc.2026.112798
- Dergi Adı: Structures
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus
- Anahtar Kelimeler: AI-based decision support system, Computer vision, Post-earthquake rapid assessment, RC buildings, YOLOv11 segmentation
- Erciyes Üniversitesi Adresli: Evet
Özet
Post-earthquake damage assessment of reinforced concrete (RC) buildings presents significant challenges in practice. The number of damaged buildings often exceeds the available capacity of expert teams, while psychological stress in the aftermath of the disaster and limited experience among inspection personnel further complicate the process. Decisions made immediately after an earthquake are critical for determining the structural safety of buildings and have substantial implications for both national economy and recovery process. Motivated by this issue, this study proposes an artificial intelligence (AI)-based decision support model designed to rapidly and accurately determine the damage levels of RC structures following major earthquakes in Türkiye. The model aims to minimize the subjectivity, reduce time consumption, and eliminate inconsistencies associated with traditional visual inspection methods. In this context, computer vision (CV) and deep learning techniques were employed to automatically classify both the type and severity of damage. The model was trained using field images collected from five major earthquakes between 2020 and 2023 and structured in accordance with the latest damage assessment codes. Damage types (crack-type and compression-type) were identified from structural elements, and damage levels were classified according to a five-tier scale (O–D). A dual-layer classification approach was adopted to improve the model’s ability to distinguish morphologically similar damage patterns. The model, developed using the YOLOv11 architecture, was tested against 132 classification and segmentation algorithms. Results demonstrate that AI-supported systems have strong potential to provide rapid, consistent, and engineering-compliant damage evaluations in the field, contributing significiantly to disaster management and structural engineering practices.