PSI - Issue 64

Quentin Sourisseau et al. / Procedia Structural Integrity 64 (2024) 893–900 Quentin SOURISSEAU/ Structural Integrity Procedia 00 (2019) 000 – 000

899 7

guarantee the absence of yielding before debonding. The produced large specimen (full size) was then cut with a water jet to obtain the final samples: 2 m long and 50 mm wide. 4.2. Obtained results and comparison with numerical expectations The samples were tested in tension (at 2 mm/min) or in bending (at 10 mm/min). For all the tested specimens, there was a competition between the GFRP/GFRP and the GFRP/steel interfaces, regardless the mode of loading of the specimen. This mode of failure is closed to the one obtained during the experimental campaign conducted on equivalent interface samples, proving the adequacy of the method. The obtained ultimate capacities at FRP debonding are given in Table 3. It can be observed that a really low dispersion is obtained indicating a repeatable process. It is interesting to note that, in tension, the applied force level corresponds to tensile stresses in the steel adherend of 787 MPa, proving the interest of the studied technique for the reinforcement of steel elements.

Table 3: Ultimate capacities at FRP debonding in tension and bending for the real size samples. Sample (Tension) Tension (kN) Sample (Bending) Bending (kN) T1 840 B1 32 T2 790 B2 29 T3 795 B3 30 T4 740 B4 29 T5 769 Average 787 30 Standard deviation/average 3% 3%

4.3. Comparison with numerical expectations A finite element model of the real size samples was developed adopting the same methodology than for the modeling of the fracture mechanics investigations. Mesh sensitivity was studied considering mesh sizes ranging from 0.5 mm to 2 mm. The failure loads obtained via the FE model for the large specimens tested in tension and in bending are compared to those obtained experimentally respectively in Table 4 and Table 5. The results show that the implemented numerical model is capable of correctly predicting the behavior of the bonded FRP patch stressed in tension (difference between the numerical and experimental values smaller than 6%). In the case of bending loading, the results present a difference of up to 24%. More investigations would be needed on this topic.

Table 4: Failure loads obtained numerically and experimentally for large specimens tested in tension. Mesh size (mm) Numerical failure load (kN)

Experimental/numerical comparison

2 1

856.1 828.5 832.4

+8.6% +5.1% +5.6%

0.5

Table 5: Failure loads obtained numerically and experimentally for large specimens tested in flexure Mesh size (mm) Numerical failure load (kN)

Experimental/numerical comparison

2 1

36.7 37.3 37.2

+22.2% +24.1% +23.9%

0.5

5. Conclusions The presented methodology aimed at assessing the use of an equivalent interface approach for the characterization and the design of an adhesively bonded FRP patch for steel structures. The approach is based on the use of fracture mechanics and the equivalent interface sample is valid for the studied FRP patch. The equivalent interface samples allowed obtaining critical energy toughnesses in different modes of loading. Those were then used to determine cohesive zone model parameters adopting a bilinear shape hypothesis. The developed models were then implemented