al., 2006]. This sensor proved to have a linear symmetric strain response under static and dynamic loading. However, the CNTs were only included within the sensor itself. A similar study showed that multidirectional strains could be measured using an isotropic film of CNTs placed on a four point probe [Dharap, et al., 2004]. This probe then could be used at different locations sensing a linear strain response in all locations. Another more recent study investigated the use of CNTs as a replacement for strain gauges [Nofar, et al., 2009]. This study placed semi-conductive multiwall CNT-fiberglass-epoxy polymer composites under both tensile and cyclic loading to detect failure. It was shown that the multiwall CNTs were able to outperform regular strain gauges in sensing different types of failures. This outcome was due to their ability to be interspersed within the composite and, as a result, react more sensitively to the changing stress fields around them.
Although much work has been done using CNTs as strain gages, more limited research has been focused on the inclusion of CNTs to monitor crack propagation. In one study, CNTs were first dispersed into a polymer matrix and then infiltrated into layers and bundles of conventional fibers [Thostenson & Chou, 2006]. This technique created a percolating network which was then used as a sensor to detect the onset, nature, and evolution of damage in advanced-polymer-based composites. A similar study demonstrated that a network of CNTs throughout the composite material was an effective way to monitor fatigue-induced damage, as well as opportunities for damage repair [Zhang, 2007]. Yet another study showed that if a high aspect ratio could be maintained throughout the entire network of CNTs, they could be highly conductive within the structure allowing for damage detection [Chou & Thostenson, 2008].
Each of these studies, however, used a network of CNTs dispersed throughout the composite base material to enable damage detection. These methods, although successful in the detection of damage, may not isolate the interfacial damage and may be impractical for large composite sections. In order to detect the interface damage, this study focuses on a layer of CNTs percolated along the interface layers. [Bily, et al., 2010].
Because CNTs have high electrical conductivity, CNTs at the joint interface are used for potential crack detection. As a crack propagates through the interface containing CNTs, the electric conducting paths through CNTs are disrupted. As a result, the electric conductivity through the interface is lowed. By measuring the electric conductivity or resistivity, the crack growth is monitored.
Figure 25 shows that two metal sheets are used at both ends of the specimens so that they can be used as the leads for measuring electrical conductivity or resistivity through the interface. A Teflon film is used to represent the initial crack. CNTs are spread to connect between the two metal sheets. The upper layers are laid on the bottom layers using the VARTM process and the final plate is cut into strips for the Mode II fracture testing. Both carbon fiber composites and fiberglass composites are studied for crack monitoring.
Mode II testing of ten carbon composite coupons containing CNTs at the interface layers are tested. Based on initial test results, the ratio of initial crack length (a) to span length (2L) needs to be greater than 0.2 to ensure failure by Mode II crack propagation. Hence, the initial crack length is set at 40 mm for a full span length of 160 mm and coupon width of 24 mm. Test speed is 1 mm/min displacement at the point of load application.
Prior to the start of testing, each coupon’s resistance through the interface is measured for baseline comparisons. Starting resistance readings shows a high degree of scatter (26.5 to 1081.0 ohms) across coupons. This scatter is due to the non-uniform distribution of CNTs, directly resulting from the dispersion technique and VARTM process. However, each
particular coupon’s resistance is essentially constant (within a tenth of an ohm) based on multiple measurements.
Steel sheet (electrical contact)
Teflon film (initial crack)
CNTs dispersed on previously cured bottom layer
Steel sheet (electrical contact)
Fig. 25. Bottom layer of two-step cure sample covered with CNTs, stainless steel sheets, and Teflon film
During the three-point bending tests, resistance across each coupon is recorded at 30 second intervals. These values generally vary little from the initial readings throughout the test. Once additional cracking takes place, the sample is left in the bent position under load. The readings taken in the bent position are again constant, only fluctuating to the tenth of an ohm, and within ±4% of the initial resistance values for 8 of the 10 samples. However, when the coupons are released from this bent position, the resistances increase an average of 16%. The variance in the percentage increase can be attributed to the non-uniform CNT dispersion. It is important to note that the resistance changes due to crack growth are generally observed only after the cracked specimens are unloaded.
After unloading, the CNT-reinforced carbon fiber samples are subsequently tested with loading and unloading cycles using a force of just 100 N – a small load enough not to cause any further interface crack growth. During each loading cycle, the resistance is measured before loading and after unloading. The measured resistances show consistency in the reading, varying 0.3% to 2.2% from sample to sample. Thus, despite continued load cycles, the resistances do not significantly increase as the crack does not grow any further.
With resistance change shown to be dependent on crack growth, the CNT-reinforced carbon fiber coupons are then further cracked under additional loading. Once the crack propagates (determined by both sight and sound), the load is removed and the new crack length and the corresponding resistance are both measured. This process is repeated until the crack tip eventually reaches the point of load application, at which point it is no longer possible to further crack the coupons under three-point bending load. The resulting data are plotted to determine a relationship between change of crack length and change in resistance.
Figure 26 shows the plots for three different samples, which have low (5.85 ohms/ mm), medium (34.1 ohms/mm), or high (164.7 ohms/mm) values of resistance change versus crack
growth length. Because each sample has different (non-uniform) dispersion of CNTs at the interface, the resistance readings are different. However, the resistance change is very linear in terms of crack growth for any individual sample, albeit with different slope for each sample. This trend can be useful to predict the crack length for a given sample if the slope is determined from a couple of initial measurements. This increase in resistance is related to the fact that the cracks for the carbon fiber composites with CNTs propagate through the layer of CNTs. Thus, as the crack continues to propagate, the CNTs are separated from each other, and their ability to conduct electricity along the interface is decreased.
E-glass fiber composite coupons containing CNTs are also tested. Fiberglass samples have an initial crack length 40 mm, a span length 160 mm, and a width 24 mm. These geometric parameters along with a Mode II test speed of 1 mm deflection per minute result in coupon failure through crack propagation.
Prior to loading, resistance is measured for ten coupons containing CNTs along the interface for baseline comparisons. An advantage to using fiberglass for testing is that the CNTs inside the composite can easily be seen. Some coupons do not conduct as areas within the coupons are observed to be void of CNTs. Each of the coupons that conduct electricity has a dark visual path of CNTs that are continuous throughout the length of the coupon. This result shows that for CNTs to be effective, the network must be contiguous, as expected. It also shows that CNTs can be effective even in non-conductive composite materials. In order to ensure that CNTs are contiguous in non-conductive media, reliable means for more uniform dispersion during the VARTM process should be developed and employed.
Even though only four of the coupons constructed are conductive, all coupons containing CNTs are put through Mode II testing and values of the resistance readings are recorded at 30 second intervals. As expected, the six coupons that initially do not conduct registered no readings throughout the test. The resistance values for the four conducting fiberglass coupons, although much higher than those obtained for the carbon fiber composite coupons, show the same trends. During the loading, the resistance readings vary little from the initial readings (within 6%). After the sample cracks and continues to crack, the resistance readings are steady (varying only a few ohms at a time), again consistent with carbon fiber composite coupons with CNTs. When the initial loading is completed, the sample is left in the bent position as was done previously. The readings in the bent position are constant, but all readings have increased from the initial values, with an average increase of 24%. When the coupons are released from this bent position, the resistances increase further. Although each coupon shows an increase in resistance, there is a scatter due to the non-uniform distribution of CNTs. The four conductive samples have resistance increases of 17, 27, 28, and 100% (for an average of 43%) compared to the initial values. This increase is qualitatively consistent with carbon fiber composites.
As before, subsequent additional loads are applied such that additional crack growth would not occur. Corresponding resistance readings are taken both before loading and after unloading. This cycling is done at least three times for each sample. The readings are consistent for each sample, only varying by at most 6.4%. Again the difference can be attributed to the uneven distribution of CNTs across the coupons. Thus, the resistance readings are not significantly altered by loads which do not produce crack growth. After taking the consistency readings, the fiberglass coupons are again loaded for crack growth using the Instron machine. Unfortunately, no useful information is gathered from this step. Upon further crack propagation, resistance readings jump to over 1 MQ. These high readings are indications that the CNTs are no longer touching and the samples are acting as open circuits.
When testing the carbon fiber composites, the way in which they failed is expected based on previous research. Fiberglass, however, is a bit surprising in its behavior both with and without CNTs. During testing of fiberglass coupons with CNTs, a loud cracking sound is heard upon failure followed by a quick decrease in the loading. This loud cracking sound is not observed during testing of fiberglass composites without CNTs. Instead, a soft crackling sound is heard. Furthemore with the fiberglass coupons without CNTs, after the crack is visually and audibly verified, additional loading is still possible
Differences in both the sound of failure, and crack propagation can be attributed to the CNTs. In the non-reinforced samples, crack propagation begins at the tip of the initial crack, and continues to propagate through the interface. This crack without CNT-reinforcement occurs
early in the loading process and slowly propagates while still maintaining an increasing load. For the fiberglass composites reinforced with CNTs, the crack also initially propagates from the initial tip through the interface. However, at a certain point the crack takes a path of least resistance outside the layer of CNTs, as shown in Fig. 27. This result is observed in the CNT – reinforced fiberglass samples and is the source for the loud cracking sound.
Crack propagation path
Initial crack CNTs at interface
Fig. 27. Path of crack propagation for fiberglass composites with CNT application:
(a) schematic sketch, and (b) picture of actual specimen.
This chapter studied the strength of composite scarf joints. First of all, a modelling technique was developed to accurately predict the joint strength. The technique is based on fracture mechanics with a very small initial crack at the most critical interface location. The crack is embedded in a resin layer with an orientation equal to the taper ratio of the scarf joint. For the mixed mode fracture, the interactive quadratic criterion is selected. This technique resulted in predicted joint strength very comparable to experimental data under different loading conditions and different taper ratios of the scarf joints.
Furthermore, CNTs were used to enhance the joint strength as well as to monitor crack growth along the interface. The introduction of CNTs along the interface especially improved the Mode II fracture strength much more than the Mode I fracture strength. The study also selected an optimal surface density of CNTs as well as the type of CNTs. The surface density of 7.5 g/m2 was optimal and the MWCNTs with a larger diameter produced a greater strength. When the joint strength was improved enough with CNTs, the failure occurred under compression along the undulated section rather than the joint interface. That is, the joint interface was not the weakest joint any more.
High electric conductivity of contiguous CNTs at the interface yields a low resistivity. However, crack growth disrupts the conductivity thereby increasing the resistivity. Such a significant change in resistivity is only observed when the specimen is unloaded. The change of resistivity is very linearly proportional to the crack growth length. However, the proportional constant is different from sample to sample because the CNT distributions are non-uniform. The knowledge in the chapter aids to better design and analysis of scarf joints with integrity.