Machine learning based prediction of joint shear strength for fiber reinforced concrete beam-column connections
COMPOSITE STRUCTURES, cilt.395, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 395
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.compstruct.2026.120772
- Dergi Adı: COMPOSITE STRUCTURES
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Chimica, Compendex, INSPEC, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO)
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
Beam-column joints are among the most susceptible regions of reinforced concrete (RC) moment resisting frames, prone to high shear forces. Fiber reinforced concrete (FRC) mitigates this susceptibility by preventing brittle joint shear failure and promoting beam plastic hinging. Therefore, accurately predicting FRC beam-column joint shear strength is essential for reliable seismic performance assessment. However, existing equations were calibrated on limited datasets mostly containing only one fiber and joint type, evaluated on the same data used for development, overlooking nonlinear feature interactions. Machine learning (ML) offers a data-driven alternative capturing these complex dependencies, handling categorical inputs, and objectively evaluating accuracy on unseen data. For RC joints, ML models are proved to outperform analytical formulations, confirming their suitability for shear strength prediction. This study extends that capability to FRC joints by developing a comprehensive framework with ten ML models, built on a database of 220 specimens. XGB and other ensemble methods achieve R 2 above 0.90 and MAPE below 9 % for test set, validated across 100 repeated random splits. SHAP identifies compressive strength as the most influential feature. A user-friendly GUI is developed to enable practical application of models, which provides a robust foundation for ML assisted seismic design of FRC structures.