Logo image
Back
Investigation of Bond-Slip Behaviour and Development of ANN-Integrated Finite Element Analysis Framework for the Optimisation of Hybrid GFRP-Steel Reinforced Concrete Beam Designs
Dissertation   Open access

Investigation of Bond-Slip Behaviour and Development of ANN-Integrated Finite Element Analysis Framework for the Optimisation of Hybrid GFRP-Steel Reinforced Concrete Beam Designs

Rajeev Devaraj
University of the Sunshine Coast, Queensland
Doctor of Philosophy, University of the Sunshine Coast, Queensland
2026
DOI:
https://doi.org/10.25907/01042
pdf
R.Devaraj_Thesis4.91 MBDownloadView
Thesis Open Access CC BY V4.0

Abstract

Structural engineering Modelling and simulation Hybrid GFRP-Steel Reinforcement Bond-Slip Behaviour Artificial Neural Networks (ANN) Finite Element Analysis (FEA) Structural Performance
Hybrid GFRP-steel reinforced concrete (RC) beams have gained significant attention due to the combined benefits of the superior durability of glass fibre-reinforced polymers (GFRP) and the ductility and energy dissipation properties of steel reinforcement. However, accurately simulating the structural behaviour of these hybrid systems remains challenging due to the complexity of the bond interface between the GFRP bars and concrete. Conventional finite element (FE) models often assume perfect bond conditions, which can lead to inaccurate predictions of load distribution, ductility, and failure mechanisms. This research addresses these limitations by investigating the bond-slip behaviour of GFRP bars in concrete through extensive laboratory pull-out tests and developing a data-driven prediction model to enhance FE simulations. Detailed parametric studies were conducted to investigate the factors influencing bond strength, such as surface morphology, embedment length, and diameter of GFRP bars. Ribbed GFRP bars exhibited the highest bond energy (89.4 J) and average bond strength (11.9 MPa). Based on experimental data, a Back Propagation Artificial Neural Network (BPANN)-based prediction model was developed, demonstrating high predictive capability with correlation coefficients ranging from 0.88 to 0.89. This model was then used to predict the force-slip relationship, which was integrated into FE simulations in ANSYS, achieving an excellent match with experimental data (R² > 0.98). The enhanced FE simulations also improved the prediction of crack propagation in concrete elements during bond failure. The laboratory tests on RC beams further explored the effects of two GFRP surface treatments, sand-coated and ribbed, on the performance of hybrid GFRP-steel reinforced beams through three-point bending tests. The results showed the superiority of ribbed bars in terms of higher load and moment-carrying capacity (11%) over sand-coated bars. Additionally, the increased slippage observed for the sand-coated bars also caused the steel rebar to become active and ductile earlier. Finally, the ANN-based bond model was integrated into FE simulations using a hybrid GFRP-steel beam with sand-coated bars, and the results were validated against experimental data. Validation of this model to the test data shows its ability to accurately predict the flexural behaviour up to ultimate loading, internal strain distribution in the composite reinforcement system, and concrete crack development, supporting the feasibility of using ANN-based bond prediction models to optimize the design of hybrid reinforcement systems. This research provides a robust framework that integrates ANN-based bond model trained on experimental data with the efficiency of FE analysis, thereby improving the design, analysis, and optimisation of hybrid GFRP-steel RC beams.

Details

Metrics

2 File views/ downloads
2 Record Views
Logo image