A Numerical Investigation of Prestress-Induced Tensile-Damage Control in GFRP-Reinforced Concrete Beams Using an Equivalent Cooling Method

Prestressing may delay tensile-damage initiation in glass-fibre reinforced polymer (GFRP)-reinforced concrete beams, but the relationship between transferred reinforcement stress and the subsequent tensile-damage development requires consistent numerical evaluation. This study develops and calibrates a three-dimensional finite-element model based on the concrete damaged plasticity (CDP) framework against published four-point bending experimental data. For the Goldston benchmark beam, the FE-predicted first-cracking load, peak load and deflection at peak load differ from the corresponding experimental values by 4.3%, 10.2% and 6.5%, respectively. Five parametric beam configurations are investigated: non-prestressed steel-reinforced B40, non-prestressed GFRP-reinforced C40, and prestressed GFRP-reinforced beams P4, P8 and P12. An equivalent cooling method is employed to introduce prestress, and the mean post-transfer GFRP axial stress is extracted from the numerical results. Using a consistent numerical damage-onset criterion of DAMAGET ≥ 0.001 within a predefined bottom midspan region, the numerical tensile-damage onset loads are 20.54, 16.39, 26.51, 35.70 and 45.51 kN for B40, C40, P4, P8 and P12, respectively. The mean post-transfer GFRP axial stresses in P4, P8 and P12 are 57.27, 114.54 and 171.79 MPa, respectively. Compared with the non-prestressed GFRP reference C40, the corresponding numerical tensile-damage onset loads increase by 61.7%, 117.8% and 177.7%, respectively. At a common loading-point displacement of 5 mm, the element-volume fraction of the predefined midspan region with DAMAGET > 0.05 decreases from 28.39% for C40 to 16.54% for P12, although the local maximum damage does not decrease monotonically. The results demonstrate that prestress can delay numerical tensile-damage initiation and reduce the spatial extent of appreciable midspan tensile damage under the adopted embedded-reinforcement model, without directly predicting physical crack width.