Synergistic hybrid deep learning approaches for high-fidelity simulation of FRP-confined CFST columns Sarra Sendjasni, Mohammed Berradia, Bilel Zerouali, Ali Raza, Riad Benzaid5d, Klaus Holschemacher
Abstract; Full Text (5153K) .	pages 001-37.	DOI: 10.12989/smm.2026.13.2.01

Abstract
The use of fiber-reinforced polymer (FRP) materials for the retrofitting and strengthening of structural elements has become a well-established and widely adopted practice. In recent years, FRP-confined Concrete-Filled Steel Tube (FRP-CFST) columns have been increasingly utilized in structural engineering due to their superior loadbearing capacity and construction efficiency. Accurately predicting their axial load-bearing capacity is essential to ensure structural safety and reliability. Although numerous studies have been conducted to estimate the axial capacity of CFSTs under concentric loading using regression analysis and conventional machine learning techniques, only a limited number have explored hybrid deep learning approaches, which have demonstrated markedly higher predictive accuracy. In the present study, advanced deep learning and hybrid deep learning models namely, Bidirectional Long Short-Term Memory (BiLSTM) and Convolutional Neural Network–Bidirectional Long Short- Term Memory (CNN-BiLSTM) were used to predict the axial load-bearing capacity of FRP-CFST columns under concentric loading. A comprehensive database comprising 250 experimental specimens collected from the literature was employed for model training and testing. The comparative analysis revealed that the CNN–BiLSTM model significantly outperformed the BiLSTM model, achieving a lower RMSE of 204.555 compared to 387.940 for BiLSTM, and reducing the MAE from 238.561 to 150.589. While empirical models provided useful insights, they exhibited greater deviations in predictive accuracy. The results highlight the CNN-BiLSTM enhanced capability in simulating complex structural behaviors, positioning it as a valuable tool for future applications in structural engineering and material science.

Key Words
axial load-bearing capacity; concrete-filled steel tubes; deep learning; fiber reinforced polymer; hybrid deep learning

Address
Sarra Sendjasni, Mohammed Berradia: Department of Civil Engineering, Laboratory of Structures, Geotechnics and Risks (LSGR), Hassiba Benbouali University of Chlef, B.P 78C, Ouled Fares Chlef 02180, Algeria
Bilel Zerouali: Laboratory of Architecture, Cities and Environment, Faculty of Civil Engineering and Architecture,
Department of Hydraulic, Hassiba Benbouali University of Chlef, B.P. 78C, Ouled Fares, Chlef 02180, Algeria;
New Era and Development in Civil Engineering Research Group, Scientific Research Center, Al-Ayen Iraqi University (AUIQ), Nasiriyah, Thi-Qar, 64001, Iraq
Ali Raza: Department of Civil Engineering, University of Engineering and Technology Taxila, 47080, Pakistan
Riad Benzaid: Department of Civil Engineering, L.G.G Research laboratory, Jijel University, BP.96 Ouled Issa,
Jijel-18000, Algeria
Klaus Holschemacher: Structural Concrete Institute, HTWK Leipzig, Leipzig, Germany