PSI - Issue 17

Formiga J. et al. / Procedia Structural Integrity 17 (2019) 886–893 "Formiga J, Sousa L., Infante V." / Structural Integrity Procedia 00 (2019) 000 – 000

892

7

As expected, the numerical curve is linear and since the insert configuration has a higher slope (stiffer) it will withstand bigger loads with less displacement. The experimental curves are mainly nonlinear. The first nonlinear region results from the contact adjustments between the rig and the specimen. After that, a linear trend starts, and this is the real reinforcement ’s capability. Af ter the linear trend some components of the structures start to get some damage and their resistance to loading starts decreasing until failure happens. 4.1. Inserts configuration According to the results presented in Fig. 7, experimental curves show a similar shape, but they have different trends. The result of insert 2 was ignored because during the test, the nuts broke down due to small bolt thread that could had influenced results. The inserts 3 and 4 have similar trends but the first one has an abrupt decrease of stiffness that can be a result of some internal defect. Comparing these two plots with number one, it is observed that the damages to the specimen started earlier on inserts 3 and 4 than on insert 1 that can keep their linear trend until almost a load of 14000N. In general, all specimens held up their designing load of 8025N and the damages only started to be observed after 10000N. Considering only the linear part and then analysing the slopes of different curves, apart from the insert 2, experimental results are similar to the numerical linear trend. This shows exact but not precise results since they are approximate from the numerical one but sufficiently far away from each other. This imprecision on results could be derived from the manufacture process since it’s difficult to put the carbon reinforcements all in the same place. Also, the hole for the insert is handmade so it could be tighter for those compared to the others, leading to some differences between specimens. Looking into Fig. 8b, with the cut section of the insert tested plate, the failure modes could be analysed. First, honeycomb core crush is observed, resulting from the plate bending, that induces core shear and leads into cells collapse. Second, in the upper face of the insert the adhesive broke down, this rupture happened due to the face sheet bending in the middle of the holes inducing shear in the adhesive. Third, delamination and ply failure are observable in the upper ply, especially in the top corner of the insert. The insert forced the upper laminate to break matching the place where, according to numerical analysis, the composite stresses and composite failure are higher.

Fig. 8. (a) Chamfer test plate cut section; (b) insert test plate cut section.

4.2. Chamfers configuration In this case, the initial offset between the numerical and experimental curves are even more noticeable in Fig. 7b. However, the results of the four specimens are closer than the ones on inserts which is quite unexpected due to the complex shape of the chamfer and the difficulty associated with its production. The graph ’ s general trend presented in Fig. 7b shows that damages start to express in higher loads, the linear area of the chamfers is larger, but the failure happened immediately after, reflecting a small "plastic zone". Considering the linear part, results are concise once all specimens have relatively similar slopes. A small difference is observable between the first and the last two results that could result from small differences on the distance between the bolts and the chamfer edges. In terms of numerical model, slope values reflect that experimental results are far away from numerical ones. This is expected due to the adaptation needed for chamfers manufacture that is mentioned above and different to what was modelled. Also, despite the tremendous effort, the produced geometry cannot be completely