As first step an optimal pre-treatment method is developed by TUD and an optimal coupon design is calculated by PATRAS. Using those, four types of possible titanium to composite joint are manufactured:
Union 1 Thermoset / Secondary bonding
Union 2 Thermoset / Co-bonding with adhesive
Union 3 Thermoset / Co-bonding without adhesive
Union 4 Thermoplastic / Secondary bonding
Static double cantilever beam (DCB) and End-Notched Flexure (ENF) tests are performed at NLR and UPATRAS on the titanium/composite samples.
The results are evaluated and show that during the DCB tests, the coupons with thermoset (CFRP) material show an unstable damage growth, starting at the material interface and growing to inside the composite laminate. Although this indicates a strong interface resulting from excellent pre-treatment, this behaviour limits the calculation of only the initiation fracture toughness values.
From the overall analysis of results of the static tests, union 2 is selected for further tests.
For fatigue testing, the experimental campaign consists of two different R-ratios, under the DCB and ENF loading configurations, and testing of aged and non-aged specimens (the aged specimens did gain 0.1% weight).
As in the static test, after the initiation of DCB experiments, the crack tended to escape the adhesive interface and create a path inside the composite lamination.
Looking at the results in terms of energy release rate, the following observations can be made:
-In the DCB case a repetitive behaviour was observed in the fatigue crack growth rate curves for each R-ratio case separately via G_max. Also G_max seemed to be affected by the R-ratio, with higher R-ratio leading to higher SERR values (as seen in literature). In the ENF case G_max grouped data from different R-ratios experiments into a more compact manner. However, no clear threshold value can be determined.
-Plotting the data using ∆√G collapsed the data from all DCB experiments (with different R-ratios) into a single group. On the other hand, for the ENF case, ∆√G clearly separated the data based on the different R-ratios values that were used for the experiments and reproduced a behaviour observed in literature. Experiments with higher R-ratio moved to left part of the graph.
- When plotting the total energy dissipation per cycle versus the fatigue crack growth, the data collapse into one curve for all cases.
- A comparison between the non-aged and aged specimens shows that there is a need of higher energy dissipation for fatigue crack growth of the non-aged specimens
For the dynamic test, the results of several fracture toughness tests under high rate testing on a Split Hopkinson pressure bar (SHPB) shows that the values of mode I SERR decrease as the opening displacement rate is increased. The kinetic energy does not affect the results on this range of velocities, as values are close to 1% of the strain energy; therefore the reduction of SERR is due to the reduction of the load. The mode II experiments present values of SERR larger than the quasi-static results. A physical explanation for this phenomenon may be attributed to different type of failure as observed after fractography analysis. The studied fracture surfaces show mainly adhesive failure on the DCB specimens and cohesive on the ENF specimen and this can explain the experimental evidence suggesting that mode I SERR decreases whilst mode II SERR increases.
Based on the test campaign on coupons, two panels with a small artificial debonding between the titanium and the CFRP stringer are fabricated. The first panel was subjected to the quasi-static monotonic compression load until the final fracture. Based on the maximum load obtained from the first test (250 kN), the second panel was subjected to a step-by-step increasing load cycle with the load step of 50 kN, including the loading, unloading and reloading to a higher load level. The AE signals were recorded and showed that the first localized AE event occurred close to the artificial debonding at the end of the 3rd load cycle and it propagated during the 4th load cycle. In the 5th load cycle, new damage occurred at the right side of the panel which is far from the artificial damage. By comparing the AE results with the DIC results, it was found that although both techniques detected the damage initiation. Finally, the AE signals of different damage mechanisms were classified into adhesive failure, cohesive failure and titanium yielding based on the AE frequency. These damages were then localized by AE. The localized damages were consistent with the images of the damage surface of the panel. Finally, the FE simulation was employed to detect and follow the damage initiation and propagation in the panel in detail. The maximum load of the panel predicted by FE was consistent with the experimental one with an error of 4%. The FE model predicted the damage propagation at the load of 190 kN which was similar to the prediction of AE. In addition, the detected location of the initial damage growth for both AE and FE was close to the artificial debonding. This work showed the potential of AE technique for structural health monitoring of aeronautical hybrid structures.
