A smart computational framework with metaheuristic optimization and an explicit analytical equation for rebar–concrete interfacial bond strength prediction using lap-spliced beams
Advances in Engineering Software, cilt.223, ss.1-22, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 223
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.advengsoft.2026.104325
- Dergi Adı: Advances in Engineering Software
- Derginin Tarandığı İndeksler: Applied Science & Technology Source, Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Scopus, Science Citation Index Expanded (SCI-EXPANDED), Compendex, INSPEC, zbMATH
- Sayfa Sayıları: ss.1-22
- İnönü Üniversitesi Adresli: Evet
Özet
This study investigates the practical applicability of an explainable artificial intelligence (AI)-assisted framework, enhanced by metaheuristic optimization algorithms, for predicting the bond strength between reinforcing bars and concrete in reinforced concrete members. A database of 401 beam specimens collected from the literature was filtered using the Isolation Forest method, resulting in 381 valid specimens characterized by nine input variables. Ten AI models were developed by combining two advanced machine learning algorithms with five metaheuristic optimization techniques. These models were systematically evaluated and compared with five conventional empirical equations. To improve interpretability, Shapley Additive Explanations (SHAP) and Individual Conditional Expectation (ICE) analyses were performed. In addition, a graphical user interface (GUI) was developed to enable rapid and reliable bond strength estimation for practical applications. The results show that all models achieved high predictive accuracy, with average coefficients of determination (R²) of 0.961 and 0.944 in the training and testing phases, respectively. The Salp Swarm Algorithm–Gradient Boosting Machine (SSA–GBM) model exhibited the best performance, with R² = 0.949, root mean square error (RMSE) = 0.748, mean absolute error (MAE) = 0.594, mean absolute percentage error (MAPE) = 12.01%, a total score of 40, and a 95% uncertainty index (U 95 ) of 1.058 in the testing phase. SHAP and ICE analyses identified specimen height, splice length, concrete compressive strength, concrete cover, and the cover-to-splice length ratio as key variables. A simplified analytical equation was also proposed, achieving R² = 0.8204, RMSE = 1.1248, MAE = 0.8623, and MAPE = 17.05%.