PSI - Issue 41

P.M.D. Carvalho et al. / Procedia Structural Integrity 41 (2022) 24–35 Carvalho et al. / Structural Integrity Procedia 00 (2019) 000 – 000

28

5

adhesive was applied. The upper adherends were then placed on top of the lower adherend and properly aligned. The value of t A was ensured by placing calibrated shims at the ends of each bonded section. The adherends were kept aligned with each other using spring clamps. Finally, the joints were left at room temperature for a week to allow the adhesive to cure. After the curing process, the excess of adhesive was removed from the joints by mechanical means using hand and power tools accordingly. The joints were tested until failure using a Universal Testing Machine (UTM) Shimadzu AGX-100 with a 100 kN load cell. The upper adherends (L- shaped) were clamped by the UTM’s upper grips, the lower adhe rend was clamped at both ends on a plate, which was gripped on the lower UTM’s grips (Fig. 2).

Fig. 2. Experimental testing of the joints, a) overall view of the UTM, b) close- up of the joint’s testing setup.

P m that each tested joint reached was recorded, and later all of them were averaged to calculate the experimental P m per upper adherend thickness. Similarly, the load-displacement ( P -  ) curves for each specimen were recorded, which are necessary for validating the numerical models.

2.4. Numerical modelling

The overall geometry of the joint corresponds to that shown in Fig. 1, with the four cases of t P2 . All cases were created numerically within the ABAQUS ® environment (ABAQUS 2017, Dassault Systèmes. RI, USA) as 2D models under the assumption of plane strain because of the adhesive layer dimensions (Adams and Peppiatt 1974). In addition, all the models have symmetry along the YZ plane, which was used to reduce the computational cost. Each model was simulated using the material properties defined in Table 1. These eight cases are considered as a benchmark for the following stage. In this work, the effect of dual-adhesive bond-line is studied, and so the combination of both adhesives was studied with the combination SikaForce ® 7752-Araldite ® 2015-SikaForce ® 7752, in a similar way to reference (Akpinar et al. 2013). The studied proportions of L O were 12.5%-75%-12.5% and 33%-34%-33%. This process was achieved by dividing the adhesive layer of each model by the desired proportion, resulting in eight models with dual adhesive bond-lines. The numerical modelling also followed two paths, one using CZM to estimate the numerical P m of each model, and the other using continuum elements throughout the model to calculate stresses at the mid-thickness of the adhesive layer. Nevertheless, for both cases, the adherends were meshed using four-node plane strain continuum elements (CPE4R). The mesh seed contained a biased seed to ensure a finer discretization at the proximities of the adhesive layer (Campilho et al. 2013), as shown in Fig. 3 . The adherends’ mesh had an approximate element size varying from 0.2 mm to 2 mm.