A stacked ensemble model for rapid seismic damage assessment of masonry buildings using data from the February 2023, Kahramanmaraş, Türkiye earthquakes

Mostofi S., BAŞAĞA H. B., OKUR F. Y., Yilmaz Z., Taciroğlu E., Sezdirmez T., ...Daha Fazla

Engineering Failure Analysis, cilt.190, 2026 (SCI-Expanded, Scopus) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 190
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1016/j.engfailanal.2026.110754
  • Dergi Adı: Engineering Failure Analysis
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC
  • Anahtar Kelimeler: 2023Türkiye earthquakes, Artificial Intelligence, Decision-Support Tool, Machine Learning, Masonry structures, Rapid Damage Assessment, Stacked Ensemble Algorithm
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

Accurate damage assessment of buildings after earthquakes is vital for making rapid and informed decisions during rescue efforts and for ensuing rehabilitation and reconstruction efforts. The myriad of challenges observed during the February 2023 Kahramanmaraş Türkiye earthquakes highlighted the need for rapid and reliable methods. Masonry structures are particularly challenging due to their complex damage patterns, which can be both apparent and hidden. This study proposes a decision-support tool using a stacked ensemble algorithm, utilizing an extensive dataset gathered in the aftermath of the 2023 Türkiye events. The architecture consists of two integrated classifiers: the first differentiates intact from damaged buildings, while the second further classifies damaged structures into severely damaged or collapsed. Multiple benchmark algorithms including Random Forest, Balanced Random Forest, and XGBoost were finely tuned and evaluated for comparison. The stacked ensemble outperformed all individual models, achieving the highest balanced accuracy and G-Mean (∼82%). Most notably, it demonstrated the lowest false negative rate in identifying damaged structures (misclassifying damaged buildings as intact), which is essential in safety–critical scenarios. This performance highlights the potential of this model as a decision-support tool for emergency responders and urban planners in managing similar future events.