88 5 Discussion Despite having a small sample size, several useful observations were made from the comparisons between DIC and the KGR-1 regarding the influence of instrumentation on the degree of data scatter and mechanical properties derived from the measured adhesive response. Generally speaking, the shear stress–strain curves obtained with the KGR-1 system and DIC, shown in Fig. 9.5, are in excellent agreement and have low scatter. With this experimental test, some scatter can always be expected due to the inherent variability in the epoxy adhesive, degree of cure within the bond, bond line thickness between specimens, residual stresses within the specimens, and defects within the bond line [23]. However, any biases imposed by these variations were mitigated through random assignment of the specimens and repeated measurements of the shear modulus with different measurement techniques on each specimen. DIC measurements have lower variance compared the KGR-1 measurements for all mechanical properties, with exception to knee stress and knee strain. In the case of the shear modulus, the aggregate statistics show that the variance associated with the KGR-1 is up to 300 % greater than DIC. This may be caused by progressive pin slippage or rotation, which was observed in almost all post-test inspections. The significant difference detected for measured knee strain, and proportional limit stress and strain between all techniques suggests that the shape of the stress–strain curve may have an appreciable dependence on strain measurement technique. A closer examination of the elastic region (Fig. 9.6) shows that, in addition to measuring lower strains, the DIC-area extractions had higher noise levels compared to the point extractions of adherend displacement. This is a result of reduced contrast in the bond line, strain resolution limits imposed by system noise (100 microstrain) and spatial resolution within the bond line. The issue of spatial resolution, within and along the bond line, is an inherent compromise when using DIC on this specimen due to the large aspect ratio of the gauge region. The resolution chosen for this work (5.30 μm/pixel) was such that one half of the overlap region was captured, from which strain extractions were likely to be more representative of the average response of the whole overlap, compared to a setup with higher resolution across the bond line. Further studies are required to investigate the sensitivity to spatial resolution within the bond line. A separate test with a high contrast, fine speckle pattern applied to the surface resulted in a slight reduction in measurement noise, but had no appreciable effect on measured mechanical properties. The strains resolved using the DIC point extraction technique were not affected by the speckle pattern. The minimal improvements from the speckle pattern do not justify the time required to paint each specimen, especially in the context of adhesive characterization for certification due to the extensive amount of testing required. However, this requires further study before definitive claims regarding the effects of surface contrast can be made. With increasing shear strain, all techniques show larger scatter, likely due to inter-specimen variations (varying bond line thickness, bond line imperfections, etc.). The KGR-1 measured slightly higher ultimate strains compared to either DIC technique, however the validity of these ultimate strains was brought into question as post-test inspection showed substantial pin slippage on four of the nine specimens tested to failure. Similar issues were encountered when collecting data from the dummy specimen, as 16 tests were required to capture 10 quality data sets. The ultimate shear strains measured within the bond line were consistently lower due to large relative displacements within the bond line, which caused problems for the correlation algorithm. This problem was first observed just beyond the onset of plasticity, and incremental correlation of images was required to capture the response. However, near failure the surface features changed so drastically that correlation was not possible. The omission of these regions of large deformation from the average shear strain measurement, causes the strain attenuation. This issue was not resolved using a speckled pattern as the paint did not adhere well to the surface under such large plastic deformations. Mech. property Shear modulus, MPa Prop. limit stress, MPa Prop. limit strain Knee stress, MPa Knee strain Ultimate strain Aggregate KGR-1 x s( ) 690.8 (68.5) 12.6 (2.6) 0.016 (0.005) 35.5 (0.4) 0.144 (0.011) 1.008 (0.067) DIC-P x s( ) 661.1 (22.9) 17.2 (0.8) 0.024 (0.001) 35.4 (0.4) 0.126 (0.011) 0.911 (0.055) DIC-A x s( ) 716.7 (29.8) 16.7 (0.8) 0.023 (0.002) 35.3 (0.8) 0.112 (0.007) 0.842 (0.050) KGR-1 vs. DIC-P NSD SD SD NSD SD SD KGR-1 vs. DIC-A NSD SD SD NSD SD SD DIC-P vs. DIC-A SD SD NSD NSD SD SD Table 9.1 (continued) J. Van Blitterswyk et al.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==