Abstract
The skin/stiffener interface debonding has been a longstanding problem for composite stiffened panels. Proper crack-arresting reinforcements become a necessity for the wide application on large-scale framed structures. Z-pinning was employed to strengthen composite skin/stiffener bond in this study. Herein, static and fatigue tensile tests were conducted on a generic configuration to characterize the improvement of Z-pinning on skin/stiffener debonding resistance to skin stretching. Results show that the improvements on ultimate debonding strength and fatigue life are significant, even though the effect on crack onset is marginal under either monotonic or cyclic loading. Z-pinning changes the unstable continuous crack growth into a propagation-suspension-propagation evolution pattern. The crack growth rate is decreased by up to three orders of magnitude due to Z-pinning. Effects of pin distribution were experimentally studied. A locally densified distribution is found to be more effective than the traditional uniform distribution.
Introduction
The ceaseless striving for light-weight aircrafts resorts to usage of composites in large amounts. Thin wall structures are the most common composite parts, but have a problem with the structural stability. An economical solution to the sensitivity to buckle and impact is to attach stiffeners. Nowadays, co-cure, co-bond, and secondary bond are methods to assemble stiffeners instead of mechanical fasteners to avoid holes creation and unwanted weight. Even though co-cured bond is weaker than high quality adhesives, it is widely used in large framed composite panels due to the growth of highly integrated manufacturing techniques. To that end, crack-stopping reinforcement should be exploited. Stitching is a popular through-thickness reinforcement method, especially for composites made with dry fiber fabrics. 1 But stitching leads to the poor operability for prepreg layups. Z-pinning technology was especially developed for the through-thickness reinforcement of prepreg forming laminates, which was shown to work with dry fiber composites as well. 2 Small diameter fiber-reinforced polymer rods (pins) are inserted into staked prepregs in through-thickness direction and combined with composite laminates via co-cure to increase the interlaminar strength, impact resistance, and fracture toughness.3–6 The efficacy of Z-pinning has been tested on a few simple joints. Z-pinning enabled up to 40% increase of static strength and 40% increase of fatigue strength (at 106 cycles) to lap joints in Chang et al.’s work. 7 Koh et al.8–10 found out that without altering the stiffness and fracture initiation, Z-pinning achieved a 75% improvement on the ultimate load and 600% on the total energy absorption for T-shaped joints at 4% pin volume fraction. Also work has been done on the influence of some design parameters including pin diameter, shape, density, and materials, as well as laminate materials and layup sequences.3,7,9,11
The remarkable improvements of Z-pinning on the properties of simple joints boost the utilization on skin-to-stiffener joints. Edge cracks tend to initiate at skin/stiffener interface when large panels undergo cyclic stretching or sudden impact. Research demonstrated that an immediate collapse of the thin skin usually follows the initiation of skin/stiffener interfacial cracks.
12
Thus, the knowledge and the control of crack growth at skin/stiffener interface are essential for the exploitation of composite stiffened thin-wall structures. Since testing and modeling of large stiffened panels are time consuming and expensive, NASA
13
developed a generic specimen configuration, so called “skin/flange configuration,” to investigate the debonding behavior of stiffener, stringer, or frame-reinforced thin-wall composite panels by simplifying stiffeners into a single flange attached to the skin as shown in Figure 1. Over the last two decades, this configuration has been used in the studies on both static and fatigue behavior of adhesive-bonded thermoset joints and welded thermoplastic joints.14–17
Simplification of the skin/flange configuration.
The objective of this study is to investigate the effect of Z-pinning on crack evolution at skin/stiffener interface with the skin panel under cyclic tension load, based on the simplified skin/flange configuration. Besides, the influence of pin distribution will be explored to achieve higher debonding resistance. These results could push forward the application of Z-pinning on performance improvement of large composite stiffened panels.
Materials and testing methods
Materials and specimens fabrication
Figure 2 depicts the configuration and dimension of the generic skin/flange specimen as developed by NASA.13,18 The specimen consists of an approximately 3.6 mm ([0/90/±45]3s) thick, 200 mm long skin laminate and an approximately 2.4 mm ([0/90/±45]2s) thick, 42 mm long flange laminate. The skin ends are strengthened by 55 mm long glass fiber composite tabs to guarantee a uniform distribution of load. The free length of the specimen is equal to 90 mm.
Dimension details of the skin/flange configuration specimen.
Both skin and flange laminates were made from NCT 301 (Newport) TR50S carbon fiber reinforced epoxy resin prepregs. The key intention for this co-cured skin/flange specimen is to eliminate distortion in the skin laminate under the flange. A special manufacture process was developed. As shown in Figure 3, two sublaminates were laid up with one used as skin laminate while the other was cut into the flange laminate and two filler laminates. These four laminates were assembled as depicted with a 0.05 mm thick release film preventing the bonding between filler laminates and the specimen assembly. The use of filler laminates made from the same material and layup with the flange laminate guaranteed equal thickness reduction during consolidation, which maintained the flatness of the skin laminate top surface.
Schematic of specimen fabrication process.
Pins made from T300/epoxy composites were inserted into the stacked assembly from the flange laminate (top surface) to the skin laminate (bottom surface). The pin diameter is 0.5 mm. A chamfered edge was created at the pin leading end to facilitate the insertion while the other end was chopped flat to minimize pin tilting under pressure. Pins were planted into a medium density foam block within a grid pattern. The foam block acted as a lateral supporter preventing pins from buckling collapse. The Z-pinned foam and a metal plate were placed on the top. The assembly was heated to 60℃ in an oven to reduce the friction resistance from prepregs during insertion. Pins were pushed by the metal plate and penetrated through flange and skin laminates with assistance of vacuum pressure. The pressure was slowly imposed by an adjustable pump. Afterwards, the pin portion remaining in the foam block was sheared off by a sharp blade.
The distribution patterns adopted in this work are shown in Figure 4 on the premise that a fixed number (40) of pins were used. Mostly, pins are uniformly distributed in the reinforcement region in a uniform grid pattern. Type-A is the traditional uniform distribution with an interval of 5 mm between adjacent pins (Figure 4(a)). However, a similar study on stitched joints showed that the distribution pattern of the reinforcement has an influence on the improvement efficacy.
19
Thus, the other two different distribution patterns were developed. Type-B employs a locally densified distribution (pins concentrated in the vicinity of flange edges) by halving the interval in the direction lateral to flange edges (Figure 4(b)). Given the number of Z-pins, only two lines of pins with a distance of 5 mm were inserted at each side of the flange laminate. A modified locally densified distribution (Figure 4(c)) was used as Type-C with a distance of 3 mm between the two pin lines. Control specimens without pins were also fabricated for static and fatigue tests.
Distribution patterns of pins (a) Type-A, (b) Type-B, and (c) Type-C.
The assembly was cured in an autoclave at the temperature of 135℃ with 60 psi pressure for 2 h. It should be noted that no adhesive was used while the flange laminate, the skin laminate, and pins were co-cured together. Tabs were attached with 3M structural adhesive film AF 163-2 L after filler laminates were removed. And finally, specimens with 200 × 25 mm2 in-plane dimension were cut from the panel using a diamond-coated rotating blade. In order to monitor the crack growth, the lateral surfaces of each specimen were coated with white paint.
Microstructure observations
Features in Z-pinned zone (called eye-shaped resin rich area) reported in previous papers 20 were observed in Z-pin reinforced specimens, including the fiber breakage, fiber waviness, micro cracks, voids, and orientation offset (the offset angle of pin’s orientation to the orthogonal direction). The distribution pattern of Z-pins has no obvious influence on the presence and characteristics of those defects. With respect to the orientation offset, the statistical investigation of Chang et al. 21 showed that the pin angle followed Gaussian distribution function and the median value for Ø0.5 mm pin was 23° within the range of 15°–33°. In this study, the offset angle was observed to be 11.8° ± 7.5°. The strategy of using a flat rear end for pins was effective to minimize the offset.
The geometry at flange foot was also observed. As shown in Figure 5, an S-shaped flange end was commonly found in cured specimens with a cavity at the corner. Layers of 90° and ±45° tend to spread with the resin flow under cure pressure in both flange laminate and filler laminates. Even though the interface between the flange laminate and filler laminates is separated by a releasing film, the interaction between resin flows from the two sides leads to the formulation of the end shape. According to the manufacturing experience, the placement method of filler laminates and the flange laminate is a key factor determining the corner geometry. The particular corner concave geometry could be responsible for the crack onset during the test. Thus, the manufacture process was carefully controlled to achieve similar corner shape for different specimen batches.
Microphoto of cured flange end (×5).
Monotonic and cyclic tensile tests
Tension was loaded along 0° fiber direction in a servo-hydraulic 250 KN MTS machine. Static tensile tests were conducted prior to fatigue tests, both of which were undertaken in displacement control. Two ends of the specimen were clamped by wedge grips. The crosshead displacement was zeroed with a <300 N preload and set at 1.5 mm/min rate. Strains in flange and skin laminates were recorded by strain gauge at the center of flange top surface and skin bottom surface. The strain in flange was chosen to monitor the skin-to-flange load transfer. When the flange strain is reduced toward zero, it is used to mean that the failure occurred at skin/flange connection. Debonding was deemed to occur and testing was terminated. Five specimens were tested under static tension for each group. The final failure displacement δf were determined from the static test results for the displacement level definition in fatigue tests.
The tension-tension fatigue tests were conducted with reference to the methodology reported by Cvitkovich et al. 18 for the characterization of skin/stiffener debonding. The cyclic tension was loaded with a frequency of 5 Hz and a displacement R-ratio (δmax/δmin) of 0.1. Two displacement levels, that is, 50% δf (δmax = 0.55 mm, δmin = 0.055 mm) and 60% δf (δmax = 0.66 mm, δmin = 0.066 mm), were used. After the installment of specimen, a <300 N preload was imposed similar to the static tests. Before cycling, the lower grip was moved to the mean displacement location in 30 seconds and held for 5 seconds. Two specimens were tested at each displacement level for each Z-pin reinforced group as well as control group. Cracks which start from the four flange corners were marked as Crack 1, Crack 2, Crack 3, and Crack 4 as shown in Figure 2. Initiations and growths were monitored and documented at fixed intervals during the tests depending on the growing rate. Visual check with help of a light was used to obtain the crack length without interrupting the test. The contrast of light reflection at the crack line against the plain zone helped determine the location of crack tip accurately. Cyclic tests were conducted continuously to a maximum crack length of 20 mm or 106 cycles depending on which one happened first.
Results and discussion
Static tensile tests
Typical responses
Figure 6 shows typical evolutions of flange strain and skin stress related to skin strain for specimens without and with pins. The skin stress was defined as the axial nominal stress with the tensile force divided by cross-sectional area of skin laminate. The flange strain was recorded as ɛf while the skin strain as ɛs. The flange strain started by an initial nonlinear stage (<4000 μɛ) and continued with a linear increase for specimens with and without pins. The nonlinear increase comes from the combined loading condition in the flange laminate including out-of-plane bending moment and in-plane tension, triggered by the tensioned skin laminate. The skin stress increased linearly to the occurrence of damage. The crack initiations were hardly noticed for control specimens followed by a simultaneous debonding failure. Under static tests, specimens reinforced by pins of different distributions experienced similar debonding pattern while the typical curves from the Type-C Z-pinned specimen are shown in Figure 6(b). The load and flange strain dropped along with a “crackle” sound and small cracks appeared at the flange edge, which caused the sudden failure of control specimens. Then the load and ɛf bounced back abruptly. Another reduction of load and ɛf indicated further propagation of the interfacial cracks. Both stress increased and flange strain decreased in a jagged pattern. Finally, flange strain decreased to zero with the presence of fatal crack extending over half of the bonding area. Results showed that specimens endured 2–3 up-and-down cycles before failure corresponding to the successive failure of pin lines.
Typical stress–strain curves of static tensile testing. (a) Specimen without Z-pins (control group) and (b) specimen with pins (from Type-C).
Debonding characterizations
Debonding properties.
On the other hand, the improvement on ultimate debonding strength is increased up to 54.6% compared to the control group. Also the average elongation limit is increased up to12,333 µɛ. The distribution pattern has a significant effect on the ultimate debonding strength. Compare to the uniform distribution (Type-A), the ultimate strength increases of Type-B and Type-C are 11.1% and 28% by concentrating pins close to flange edges. From the results, the average failure displacement (δf) is 1.1 mm for control group and it is used for the fatigue tests on both control specimens and Z-pinned specimens for comparison.
Fatigue behavior
Crack initiation
Summary of the number of cycles to crack initiation.
Crack propagation
The growths of Crack 1, 2, 3, and 4 were plotted along with the base 10 logarithm of cycles (log10N) for the control specimens and specimens reinforced by Type-A, Type-B, and Type-C distributed pins. The typical curves are shown in Figure 7 (60% δf level) and Figure 8 (50% δf level). Similar crack evolution pattern was observed between control group and Type-A group, as well as between Type-B and Type-C group.
Typical crack growths of specimens at 60% δf level. (a) Control specimen, (b) specimen with Type-A Z-pinning, (c) specimen with Type-B Z-pinning, and (d) specimen with Type-C Z-pinning. Typical crack growths of specimens at 50% δf level. (a) Control specimen, (b) specimen with Type-A Z-pinning, (c) specimen with Type-B Z-pinning, and (d) specimen with Type-C Z-pinning.
At 60% δf level, tests were run up to about 2 × 104 cycles on control group, 105 cycles on Type-A group, 4 × 105 cycles on Type-B group, and 7 × 105 cycles on Type-C. A semilinear relationship between crack lengths and log10N exists in control specimens and Type-A reinforced specimens. The cracks evolution in Type-B Z-pinned specimens gradually transits into a three-stage pattern which is more clearly observed in Type-C Z-pinned specimens. At the first few thousands of cycles, cracks constantly propagated until they exceeded the second pin lines. Subsequently, crack propagations were decelerated. And cracks were even suspended in Type-C Z-pinned specimens, which were maintained for up to 3 × 105 cycles. Finally, cracks restarted by a linear continuous propagation. Cracks advances toward the center line without any obstruction at the final stage.
At 50% δf level, the similar linear propagation is found in control and Type-A Z-pinned specimens. But fatigue life is prolonged from 105 cycles to >106 cycles by the reinforcement of uniformly distributed Z-pins. For the two groups with locally densified pins, cracks are again arrested at the edge of Z-pin reinforced region from 2 × 104 to 3 × 104 cycles. Cracks overcome pins arresting at 6 × 105 cycles in Type-B Z-pinned specimens. On the contrary, the stagnation exceeds 106 cycles and cracks are arrested infinitely at the neighborhood of the second pin line in Type-C Z-pinned specimens. Before the crack arresting period, pins maintain the bonding in the center region between pin lines at each side.
The crack-arresting effect comes from the teamwork of the two pin lines at each end. As confirmed by the fact shown in Figure 7(c) and (d) and Figure 8(c) and (d), the suspension of cracks at the first pin line is unnoticeable or less prominent than the one at the second pin line. Even though the interaction between the two pin lines is still not clear, the stable crack propagation usually will not stop until it passed the second pin line, which is the border of Z-pinned region. By contrary, no crack retardation is observed in uniformly distributed pins reinforced specimens since no such a border exists.
Crack growth rate
In these tests, the four cracks did not symmetrically propagate even though the specimen and fixture were designed to be symmetric, especially for Z-pinned samples. Some reasons were deemed to include the asymmetric features from manufacturing and the asymmetric distribution of pins caused by the pin offset angle. For this asymmetric crack growth, two available standards were used before
16
: the crack averaged on the four individual cracks or the main crack first reaching failure, both of which were demonstrated to be alternatives to show the crack evolution trend. We employed the total length as a sum of the average crack length on each side, which was actually identical to the former standard. The total crack length was calculated as follows
The secant method was used to calculate the crack growth rate (CGR) defined as the crack propagation in mm during each cycle, suggested by ASTM E647-15 standard.
23
In Z-pin reinforced specimens, crack growth at pin/laminate interface was difficult to detect, thus the “apparent” CGR was determined on the calculated total crack length at skin/stiffener interface. Figure 9 plots the CGR versus crack length normalized on the whole bonded length, that is, 42 mm. A large amount of oscillation is caused by the inconstant crack propagations in Z-pin reinforced specimens.
Typical crack growth rate versus normalized total crack length. (a) At 60% δf level and (b) at 50% δf level.
From the curves, the highest CGRs are observed in the first 10% crack propagation with similar value regardless of the presence or distribution patterns of pins. For Z-pinned specimens, the similarity exists within the range from flange edge to the first pin line (off-edge distance). This is a further evidence of pins’ ineffectiveness to stop crack propagations outside the pin region. An overall decrease of CGR after the initial period is obtained by the introduction of Z-pinning. A comparison with the control group indicates that the CGR reduction of Type-A Z-pinned specimens becomes more significant at lower displacement level and at longer crack length. The CGR decreased up to three orders of magnitude at specific crack length through uniformly distributed reinforcements. On the other hand, appreciable valleys can be found on the CGR curves respectively around the two pin lines for Type-B and Type-C Z-pinned specimens, which highlights the prominent resistance to fatigue crack enabled by locally densified pins. At those locations, the CGR decreases up to three orders of magnitude compared to the control group. At 50% δf level, the overall crack length is decreased by 46% by adopting a distribution pattern with pins concentrating close to flange edge rather than the uniform distribution. However, at relatively high level (60% δf), advantages of Type-C over Type-B are less apparent than that at low level (50% δf) when a fatigue life longer than 106 cycles is achieved. The increase of displacement amplitude undermines the improvement effect of Z-pinning on fatigue cracking resistance, which was also proved by the experimental works of Warzok et al. 24 on the fatigue degradation of the single-pin reinforced.
Failure observations
Failed specimens were cut along the lateral surface. The surfaces were polished and examined under a microscope. Crack evolution behavior under both static and cyclic loads seemed to be similar with cracks initiated at resin fillet at flange edge (Figure 10(a)). The cracks advanced forward as delamination through the bondline formed between two 0° layers (Figure 10(b)), in spite of the presence of Z-pinning and distribution patterns. The presence of pins introduced a large amount of interface as paths for crack propagations. Typical failure modes include pin/laminate debonding, shear rupture, and inner splitting.
25
Combination of parameters including pin inclination, load conditions, layup sequence, pin strength, and pin/laminate bonding strength determine the mode by which pins fail.7,26 For Z-pinned skin/flange specimens under static tension, pin/laminate debonding was the prevalent failure (Figure 11(a)) while fiber breakage was also observed inside pins (Figure 11(b)). However, no sign of pin breakage was found in specimens under fatigue loading, since a limited displacement was imposed. Transition of pin failure mode from pullout (static load) to transverse fracture (fatigue load) under mode I load was also reported by Cartié et al.
27
Micro photos of cracks at skin/flange bondline. (a) Crack initiation at resin fillet (×10) and (b) crack propagation at bondline (×50). Failure modes of pins. (a) Pin debonding (×5) and (b) fiber breakage in pins (×50).
Conclusions
In summary, the ultimate debonding strength, fatigue crack growth resistance and fatigue life of the skin/flange configuration are significantly improved by Z-pinning under in-plane tension. For the configurations and materials used in this study, the ultimate debonding strength was increased by up to 54.6%. CGR was decreased up to three orders of magnitude once the crack extended into the reinforced region. And without alerting the crack initiation cycles, the fatigue-crack propagation life was increased more than 10 times by introducing Z-pin reinforcement. At 60% displacement level, a >106 cycles fatigue life was achieved.
The comparative tests confirm the advantage of a locally densified distribution over the traditional uniform distribution of pins. Not only a substantial increase of debonding strength, but also long-time crack suspension and lower CGR were achieved by concentrating all Z-pins within a critical area. The continuous crack propagation shifts into a propagation-suspension-propagation pattern through being arrested by the locally densified pins. Considering the manufacturing requirement and in-plane damage of Z-pinning, the maximum density for pins is recommended as 4%, which means the minimum interval for adjacent pins is 2.2 mm for Ø0.5 mm pins (a minimum interval of 2.5 mm was used in this paper).
Pins inhibit the propagation by diverting cracks from the skin/flange bondline to pins/laminate interfaces. Pin debonding failure mode combined with a small amount of pin breakage is found in static specimens. However, when cyclically loaded, pins fail only by debonding.
Footnotes
Acknowledgment
The authors thank the colleagues from Concordia Center for Composites, Concordia University, for technical support.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work was supported by the Aeronautical Science Foundation of China (No. 2015ZE52049); China Scholarship Council and Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).