PSI - Issue 47

L.A.R. Gomes et al. / Procedia Structural Integrity 47 (2023) 94–101 Gomes et al. / Structural Integrity Procedia 00 (2019) 000 – 000

98

5

3. Results 3.1. Validation with SLJ geometry

Validation with experiments was carried out on a SLJ geometry bonded with the AV138 and high-strength steel adherends (DIN 55 Si7). Numerical modelling was performed identically to the description of section 2.3, namely with 2D models composed by CPE4 solid elements (adherends) and COH2D4 cohesive elements (adhesive). The validation study, and experimental and numerical details, are found in reference (Valente et al. 2019). Comparing the maximum load ( P m ) and the P m displacement (  P m ) with the experimental data, it is possible to validate the technique.

16

16

3.6

13.61

11.75

12

12

2.7

8

8

1.8

P [kN]

 P m [mm]

P m [kN]

1.56

4

0.9

4

0.27

0

0

0

Experimental

Numerical

0

0.5

1

1.5

2

2.5

 [mm]

Pm dPm P m  P m

a)

b)

Experimental

Numerical

Fig. 2. Experimental and numerical comparison of P -  curves (a) and P m and  P m values (b).

It is possible to evaluate the experimental/CZM data of the P -  curves, and both P m and  P m , in Fig. 2. The results presented in Fig. 2 (a) highlight a relevant difference in the P -  curves. However, it should be noted that  measurement and precision depend on the equipment and extraction method. In this work,  was estimated by integrating the accelerometer data, which proved to be inaccurate. On the other hand, comparing the experimental and numerical P m , the results are considered acceptable, although the numerical P m slightly overpredicts the experiments. This difference is justified by the theoretical consideration of perfect fabrication and bonding, and uniform adhesive thickness. It is possible to notice, in Fig. 2 (b), that the CZM P m prediction exceeds the experiments by 15.8%, showing partial inability to replicate the theoretical perfect joint conditions implicit in the numerical analysis. Added to the aforementioned issues, the load cell precision and specimens’ misalignments in the grips are other possible causes for these differences. Comparing the experimental and numerical  P m results, the numerical values are lower than the experimental values by 82.7%. One possible explanation for this deviation is the error induced by integration of the accelerometer´s data, as previously mentioned. Concluding, the numerical technique has good capabilities in the P m prediction, making it possible to undertake a fully numerical study on DLJ strength. 3.2. Elastic stress distributions  y and  xy stresses are the relevant stress components within the context of in-plane loaded bonded joints. In the plots that follow, the length layer length in the x -axis is normalized by L O , in such a way that 0≤ x / L O ≤1. On the other hand, both  y and  xy stresses appearing in the y -axis are divided by the average  xy in the adhesive layer for each L O (  avg ). Fig. 3 presents σ y / τ avg stresses in the adhesive layer as a function of L O for both adhesives: AV138 (a) and 7752 (b). The maximum σ y stresses are reached at the overlap ends with a pronounced gradient, in while σ y stresses are almost nil along most of the overlap zone. This behaviour is related to the load asymmetry, which causes bending moments that lead to peak stresses (Nunes et al. 2016). Nonetheless, this effect is less significant than that observed in SLJ (Campilho et al. 2011). L O has a major effect on stress distributions within the adhesive layer, in the sense that smaller L O markedly lead to reduced peak σ y stresses than higher L O . Peak σ y / τ avg values for the AV138 attained 3.93 and 8.88 for L O =12.5 and 50 mm. respectively, while the equivalent values for the 7752 were 1.72 and 3.91. With these results, it can be concluded that the joint load transfer efficiency of joints bonded with the 7752 is higher than