You are on page 1of 19

Available online at www.sciencedirect.

com

Composites: Part B 39 (2008) 907925 www.elsevier.com/locate/compositesb

An experimental investigation into the progression of damage in pin-loaded bre metal laminates
R.M. Frizzell, C.T. McCarthy *, M.A. McCarthy
Composites Research Centre, Materials and Surface Science Institute, Department of Mechanical and Aeronautical Engineering, University of Limerick, Limerick, Ireland Received 5 October 2007; received in revised form 22 January 2008; accepted 24 January 2008 Available online 21 February 2008

Abstract An investigation into the progression of damage in pin-loaded bre metal laminate (FML) joints is presented. A parametric study was carried out on the commercially available FML, GLARE, in which edge distance to hole diameter (e/d) and width to hole diameter (w/d) ratios were varied and corresponding failure modes and loads identied. It was found that bearing failure was the dominant failure mode, and obtaining shear-out and net-tension failures required very low e/d and w/d ratios, respectively. Even in specimens that failed by shear-out or net-tension, substantial bearing damage was present. The catastrophic nature of the bearing failure mode under load control was demonstrated. Damage progression was examined using microscopy on specimens loaded to ultimate failure and to percentages of their ultimate failure load. Initial non-linearity in all types of specimens was found to be primarily due to plastic deformation of the aluminium plies and micro-buckling of bres in the 0 layers at the bearing plane. For specimens loaded to percentages of failure, the glass layers were found to recover some of their strain after removal of the pin, while the aluminium layers did not recover. At higher loads, the rate of stiness loss in both shear-out and net-tension specimens slowed due to the activation of tensile loading of bres, until just prior to failure when the bres started to break. In the bearing specimens, nal failure involved complete separation of the outer aluminium plies from the glass layers, which under load control was found to be an unstable process. The study provides supporting information for the development and validation of numerical damage models for FML joints. 2008 Published by Elsevier Ltd.
Keywords: C. Damage mechanics; D. SEM; E. Joints; Fibre metal laminates

1. Introduction Fibre metal laminates (FMLs) are a family of advanced aerospace materials that consist of alternating layers of thin metal sheets and bre-reinforced plastic. FMLs originated at Fokker/TU Delft in The Netherlands during the late 1970s and since then have undergone extensive development [14]. One of the rst large-scale demonstrators of FMLs was on a Fokker 50 lower wing panel, where their weight saving ability was demonstrated [1]. Early FMLs, known as ARALL, were based on Aramid bres while
Corresponding author. Tel.: +353 (0) 61 23 4334; fax: +353 (0) 61 20 2944. E-mail address: conor.mccarthy@ul.ie (C.T. McCarthy). 1359-8368/$ - see front matter 2008 Published by Elsevier Ltd. doi:10.1016/j.compositesb.2008.01.007
*

more recent versions (since the mid-1980s) utilise high strength S2-glass bres. The works of Vlot et al. [1] and Gunnink et al. [2] provide comprehensive reviews of the development of these materials. FMLs boast a large number of favourable characteristics, such as excellent fatigue performance, high residual and blunt notch strengths and low density [1]. In addition, their burn-through resistance is superior to that of aluminium alloys [4] and they have good corrosion resistance (since corrosion stops at the prepreg layers [5]). FMLs can also be machined and riveted like conventional aluminium alloys, and Woerden et al. [6] found that cutting/drilling procedures typically used for aluminium did not cause any undesirable eects when used on FMLs, such as bre pullout or delamination. This is in contrast to conventional

908

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

composites where lateral supports and specialised cutting tools are often necessary [7]. It has also been found that the tensile strength of GLARE increases with increasing strain rate, and McCarthy et al. [8,9] found that glassbased FMLs performed excellently under high-velocity impacts such as those due to bird strikes on aircraft. Such characteristics make these materials attractive for structural applications, particularly in the aerospace industry. For example, the glassbre based FML GLARE has found extensive application on the Airbus A380, with approximately 380 m2 of the material used in the upper fuselage skin, with further use in the D-nose of both the horizontal and vertical tail planes possible [10]. For prediction of in-service behaviour of FMLs it is necessary to investigate how they behave in mechanically-fastened joint congurations. Bolted and riveted joints represent potential weak points in a structure due to the highly localised stresses induced, and their ecient design is therefore critical to the load-carrying capability and overall weight of the structure. The performance of FMLs when riveted or bolted is also important from repair considerations. The eects of varying the width, edge distance and thickness on the bearing strength of aramid and glassbre based FML joints have been investigated by a number of authors [1115]. Wu and Slagter [15] found that dierent strengths and failure modes were produced by varying these parameters. They concluded that in order to develop the full bearing strength of FMLs, designers should use a minimum edge distance to hole diameter ratio (e/d) of 3 and a minimum width to hole diameter ratio (w/d) of 4. They also investigated the eects of lateral constraint on the bearing strength of FML joints and found that lateral constraint could increase the bearing strength by more than 20%, since it restrains delamination growth. It is generally accepted that bre-reinforced plastics (FRPs), such as those used in FMLs, fail in a progressive manner, where damage accumulates in the laminate as the load is increased. A number of previous studies have investigated damage progression in FRPs [1621] and the nature of the damage and the failure mode has been shown to depend on a number of factors such as material type, layup and geometry. While the failure mechanisms in conventional composites are fairly well understood, the introduction of metallic layers to form a FML signicantly increases the complexity of damage progression due to the large mismatch in mechanical properties between the dierent constituent layers. A number of studies have examined the failure sequences in notched and unnotched FMLs [2224] under tensile loading. Quasi-static tensile tests of unnotched specimens have revealed that following initial yielding, FMLs exhibit essentially linear isotropic hardening behaviour, with bre failure and delamination occurring just before ultimate failure [22]. An investigation carried out to determine the failure sequence of GLARE blunt-notched specimens (including open hole and centre-notched specimens)

found that failure began with plasticizing of the aluminium followed by debonding between the aluminium and the glass layers [23]. The blunt notch failure mechanisms of GLARE were also investigated by De Vries [24] who found that the dominant failure mechanisms were bre/ matrix debonding, matrix cracking and delamination between the aluminium and the bre layers. Both SEM and C-scan imaging have been used in these studies to examine the extent of damage, and some studies [12,24] have utilised chemical etching to remove the outer aluminium layer so that defects in the exposed glass bre layers could be examined visually. Concerning bolted congurations, Van Rooijen et al. [12] presented a numerical model for predicting the damage progression in FMLs subjected to bolt bearing loads. The results of this model were compared to experimental bearing stress versus displacement results for validation purposes. In [25] Van Rooijen also investigated the pin bearing strength of FMLs, which the author argues is important for design purposes since it is impossible to obtain full lateral restraint with all in-service joints. The pin-bearing test method leads to low joint strengths in FMLs since it does not provide any lateral support to prevent delamination [26]. Since the pin-bearing test method represents a severe test case for joint design it is of interest to develop a thorough understanding of the progression of damage in pin-loaded FMLs. Hence, the aim of this paper is to examine the progression of damage from initiation up to ultimate failure in FML pin-loaded joints. Three distinct failure modes, bearing, shear-out and net-tension, were induced by varying the joint dimensions. Once the appropriate dimensions to produce each failure mode were determined, samples for each mode of failure were loaded to ultimate failure and to percentages of the ultimate failure load, and examined by sectioning and microscopy. This approach is similar to that used in [27] for bolted CFRP laminates. The results from the present work are currently being used by the authors in the development and validation of an advanced damage model for FML joints. 2. Experimental methods The specimen geometry and loading is shown in Fig. 1. The FML laminate was held at one end while load was introduced through the pin. In bolted structures, full lateral support (e.g. from washers or bolt heads) cannot always be guaranteed due to loosening or tipping of the bolt, or bending of the laminates. The pin-bearing test procedure without lateral supports, as used here, is generally considered to represent a worst case scenario for in-service conditions. The material chosen was GLARE 3 3/2 0.4 L glass bre based FML, which was purchased from a commercial materials supplier. This variant of GLARE is in use on the upper fuselage of the Airbus A380. The lay-up consisted of four layers of 0.125 mm thick glass bre-reinforced prepreg (S2 glass bres and FM 94 resin) and

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

909

Fig. 1. Pin-bearing test conguration and typical joint geometry.

three layers of 0.4 mm thick 2024 aluminium alloy. The stacking sequence was [AL/0/90/AL/90/0/AL] (where AL represents an aluminium layer and 0 and 90 refer to the orientation of the prepreg layers) with a total nominal laminate thickness of 1.7 mm. The specimens were all cut from a single sheet of FML, thus ensuring consistency in material composition. The laminates were initially cut oversize using a dedicated composite cutting machine with a diamond coated blade and were then machined to nal specication using a computer aided precision milling machine with carbide cutters, thus ensuring high quality edges and geometric accuracy. All joints had a nominal length L = 135 mm and the hole was drilled undersize using a standard tungsten carbide drill bit and then reamed to a nal diameter of 6 mm using a straight edged reamer. Since all joints had a xed 6 mm diameter hole, various width to hole diameter (w/d) ratios and edge distance to hole diameter (e/d) ratios were achieved by varying the joint width and edge distances, respectively. The three failure modes to be examined were bearing, shear-out and net-tension as illustrated in Fig. 2ac, respectively. Bearing failure is dened as local crushing of the material adjacent to the hole and normally occurs when

the e/d and w/d ratios are large. Based on previous work on composites [27] and FMLs [15] values of w/d = 6 and e/d = 6 were used to promote bearing failure. Shear-out failure occurs when a plug of material separates from the laminate ahead of the pin, as shown in Fig. 2b, and normally occurs when the e/d ratio is small (or in conventional composites, if an excessive amount of 0 plies are used). As explained by Hart-Smith [28] shear-out failure is a special case of bearing failure. It is expected therefore to nd some amount of bearing damage in each failed shear-out specimen. Shear-out failure can therefore occur after some bearing damage has initiated. In this paper a shear-out failure is deemed to have taken place when the ultimate rupture of the specimen occurs by a plug of material separating from the laminate ahead of the pin. This is the dening characteristic of the shear-out failure mode. In order to nd the e/d ratio needed to promote shear-out failure a parametric study was conducted. Hart-Smith [28] suggests that for conventional composites (i.e. without metal layers) an e/d ratio of 3 or greater is sucient to prevent the shear-out failure mode (for commonly used lay-ups). Hence, the e/d ratio was reduced from an initial value of 3 until a shearout failure was observed (see Table 1 for the various ratios tested). When an appropriate value of e/d to promote shear-out failure was found, this value was used to manufacture a number of samples for use in the main (microscopy) test series. Net-tension failure involves a fracture across the width of the joint, as illustrated in Fig. 2c, and normally occurs when the w/d ratio is small. Wu and Slagter [15] found that w/d values as low as 2 were necessary to promote net-tension failures in GLARE 2 joints. In the present work, a parametric study was carried out where the w/d ratio was varied until net-tension failure was observed (see Table 1 for the ratios used). The w/d value at which net-tension failure occurred was then used for specimens in the main (microscopy) study. A schematic of the experimental set-up is shown in Fig. 3. A neat-t 6 mm hardened steel pin was inserted into the hole and then loaded under displacement control via a specially designed testing frame. The testing frame was connected to the testing machine via universal ball joints which prevented any bending moments from being transferred through the joint. No washers or other lateral sup-

Table 1 Geometric ratios examined for bearing, shear-out and net-tension failure modes Failure mode under examination Bearing Shear-out Test no. 1 1 2 3 1 2 3 w/d 6 6 6 6 3 2.5 2.0 e/d 6 3 1.5 1.3 6 6 6 Failure mode detected Bearing Bearing Bearing Shear-out Bearing Bearing Net-tension

Net-tension Fig. 2. Failure modes under investigation.

910

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

with watered silicon carbide paper and polishing with diamond suspension uid. 3. Results and discussion 3.1. Failure modes From Table 1 it can be seen that e/d and w/d ratios of 6 resulted in bearing failure as expected. However, somewhat contrary to expectations, a very small e/d ratio (1.3) was needed to cause shear-out (an e/d ratio of 1.5 did not result in shear-out). For net-tension failure a quite small w/d ratio (2) was required. For the remainder of this paper, data is presented only for three joint congurations, namely the joints that failed by bearing (w/d = 6, e/d = 6), shear-out (w/d = 6, e/d = 1.3) and net-tension (w/d = 2, e/d = 6).
Fig. 3. Schematic showing salient features of experimental set-up.

ports were used. The deection of the pin was measured using an Epsilon displacement gauge with a telescopic arm. The displacement gauge was connected to a National Instruments DAQ system and relayed to a PC using Labview 7.0 software. For each of the three nal congurations (producing the three dierent failure modes), three repeat experiments were performed until ultimate failure and the results were then used to determine the average ultimate failure loads. To investigate the progression of damage, samples were tested to 60%, 80%, 90%, 95% and 100% of the average ultimate failure loads. The specimens were sectioned at specic planes within the laminate, and then examined using a scanning electron microscope (SEM). The planes were marked prior to testing and then exposed using a diamond coated wafering blade. The sectioned material was mounted and cast in epoxy resin, and nished by grinding

3.2. Failure loads and bearing stress versus bearing strain response for each failure mode Fig. 4 shows an example of each failure mode and Table 2 contains the average failure load (dened as the maximum load reached) and the ultimate bearing strength for the dierent joint types. The bearing stress rb is dened as rb P dh 1

Table 2 Average failure loads and bearing strengths for the FML joints Specimen Bearing Shear-out Net-tension Average ultimate load (N) 6017 5817 5217 Ultimate bearing strength (MPa) 563 549 489 Coecient of variance (%) 1.29 0.98 1.53

(a) Bearing
600

(b) Shear-out

Bearing Stress (MPa)

500 400 300

(c) Net-tension
200 100 0 0 5 10 15 20 25 30

Bearing Strain (%)


Fig. 4. Bearing stress versus bearing strain curves for each of the specimen types tested. An example of each failure mode is also shown.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

911

where: P is the applied load, d is the hole diameter and h is the specimen thickness. It is noted that damage due to the pin motion causes material to broom outwards ahead of the pin. This increases the bearing surface area and could aect the values of bearing stresses calculated. This

eect is not taken into account in this paper due to the diculties involved in measuring the change in bearing surface in real time. The ultimate bearing strength ru b is the value of rb when P equals its maximum value. As can be seen the joints designed to fail in bearing were

140 120 100

Bearing Shear out Net tension

Bearing Stress (MPa)

80 60 40 20 0 -20 0 -40 -60 5 10 15 20

25

30

Bearing Strain (%)

Fig. 5. Comparison of the bearing stiness versus bearing strain curves for the three joint types.

Fig. 6. Typical micrograph of the FML: (a) view showing the dierent layers; (b) close-up showing ply orientations after bre buckling.

912

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

the strongest, followed by the shear-out joints and nally the net-tension joints.

The bearing stress versus bearing strain responses of the various joint congurations are also shown in Fig. 4. In this gure the bearing strain is dened as ebr dp Kd 2

Fig. 7. Schematic showing the mechanisms behind the protrusion of the glass layers beyond the aluminium layers at high load.

Fig. 8. Plane examined for the bearing specimens (w/d = 6, e/d = 6).

where dp is the pin displacement, d is the hole diameter and K = 1.0 for double shear joints [29]. The data in Table 2 shows excellent repeatability for each of the three specimen congurations with coecients of variance for the ultimate load being in the range of 1 1.5%. For each of the three congurations little variation was found in the entire loaddeection responses of the three tests to failure. For this reason only a single representative curve for each failure mode is shown in Fig. 4. From Fig. 4a, the bearing specimens show a linear response up to approximately 385 MPa. At this point, the stiness begins to decrease and keeps decreasing until a maximum stress of approximately 560 MPa is reached. This stiness reduction indicates the onset and growth of damage and plasticity in the joint. The bearing stress then drops o, but in a gradual manner, in line with the general viewpoint of bearing failure as a non-catastrophic failure mode. The tests were stopped after approximately 1 mm pin displacement or 18% bearing strain. The apparent non-catastrophic nature of the failure of the bearing specimens is discussed further in Section 3.3.1. The bearing stress versus bearing strain response for the shear-out specimens in Fig. 4b shows a similar region of linearity, with a more extended non-linear region. For these joints the stiness starts to reduce at approximately 280 MPa and continues thereafter until catastrophic shear-out failure at approximately 550 MPa. At this point, a signicant downturn is observed in the curve and thereafter the joint was unable to sustain any further loading. The prolonged region of non-linearity indicates the develop600

Bearing Stress (MPa)

500 400 300 200 100 0 0 140 120 5 10 15 20 25 30

Bearing Strain (%)

Bearing Stiffness (MPa)

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Fig. 9. SEM micrograph of a bearing specimen (w/d = 6, e/d = 6) at 80% of the average failure load. Magnication: 60. (Note: that the large white area in the gure above is dust.) Shown on the right are the stressstrain and stinessstrain curves with diamond symbol indicating point corresponding to the micrograph.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

913

ment and growth of extensive sub-critical damage before ultimate shear-out failure occurs. This highly non-linear response is in contrast to the response of joints made from carbon to bre-reinforced plastic that fail in shear-out; in [27,30] such joints were found to follow an almost linear loaddeection response until the rst peak in the load deection curve signied the onset of shear-out failure. Clearly the ductile behaviour of the metal layers in the FML joints has a signicant eect on the shear-out specimen response. Finally, the bearing stress versus bearing strain curves for the net-tension specimens are shown in Fig. 4c. As can be seen, the response is linear up to approximately 290 MPa and thereafter the stiness continuously reduces until ultimate net-tension failure occurs at approximately

490 MPa. At this point, the specimens suddenly rupture by a crack growing from the hole edge out towards the edge of the laminate and the joint is unable to sustain any further loading. This catastrophic loss in the joints load-carrying capacity is typical of the net-tension failure mode. Additional insight can be gained by examining the variation in joint stiness during the tests. Loss of stiness is an indication of progressive damage development and has been calculated here in a similar fashion to that used by McCarthy et al. [29]. This involves calculating the stiness as a moving average of nine adjacent data points from the bearing stress versus bearing strain curves. This smoothing operation is necessary to smooth out the very noisy stiness data and reveal the interesting overall trends.

600

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

Bearing Strain (%)


140 120

Bearing Stiffness (MPa)

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Fig. 10. SEM micrograph of a bearing specimen (w/d = 6, e/d = 6) at 90% of the failure load: (a) 60 magnication; (b) 400 magnication.

914

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

The stiness variation was analysed for the three test repeats of each joint conguration. As noted above, little variability was found in the stressstrain curves for the three repeats of each conguration, so not surprisingly little variation was found within each set of three stiness curves either. Fig. 5 therefore shows one representative stiness curve for each joint type. Immediately evident from the gure is the dierences in initial stiness of the three joint congurations, with the bearing specimen having the highest initial stiness followed by the shear-out specimen and then the net-tension specimen. All three joint types exhibit an initial period of relatively constant stiness corresponding to the linear elastic region of the bearing stress versus bearing strain responses in Fig. 4. As the strain is increased there is a point for each conguration where the stiness begins to decrease and the bearing strain

value at which this occurs seems to be similar for all joint types. This suggests that the same mechanisms may be responsible for the initial onset of non-linearity in each of the joints. This issue will be further investigated in the next section. Concerning the rate at which stiness is lost for each joint type, the bearing joints exhibit a steady and steep rate of stiness loss up to bearing failure (which occurs at approximately 8% bearing strain), after which a period of negative stiness occurs as the load drops o with increasing strain (as seen in Fig. 4). The net-tension joints display a much more gradual rate of stiness loss until just before nal failure (at around 13% bearing strain) when the rate of loss suddenly increases. Interestingly, the shear-out joints initially show a similar rate of stiness loss to the bearing specimens (between about 2.5% and 4.5% bearing strain),

600

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

Bearing Strain (%)


140 120

Bearing Stiffness (MPa)

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Fig. 11. SEM micrograph of a bearing specimen (w/d = 6, e/d = 6) at 95% of the failure load: (a) 60 magnication; (b) 250 magnication.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

915

but then the rate of stiness loss reduces signicantly and in fact between about 7% and 12% bearing strain, the shearout and net-tension joints possess almost identical stiness and rate of change of stiness. These results suggest that the mechanisms causing stiness loss in the shear-out joints in the early stages of deformation are similar to those in the bearing joints, but later on, mechanisms similar to those in the net-tension joints start to dominate the stiness behaviour. The reasons behind these observations will be examined more fully in the microscopy study in the following section. 3.3. Scanning electron microscopy (SEM) This section presents SEM micrographs of the specimens after sectioning. The micrographs were taken at load levels corresponding to 60%, 80%, 90%, 95% and 100% of the average failure loads for each joint type, given in Table

2. They thus map the damage mechanisms occurring as load increases. One sample for each load level was examined using SEM for each of the three joint types under consideration. For convenience of interpretation, a bearing stress versus bearing strain curve and a bearing stiness versus bearing strain curve are presented next to the micrographs, with a symbol indicating where on each graph the micrograph corresponds to (for example, see Fig. 9). Before presenting the results, some of the salient features observed in the micrographs are described. Fig. 6a shows a typical micrograph highlighting the dierent layers that form the laminate, while Fig. 6b shows another micrograph at a higher magnication, where it is possible to identify the orientation of the bres. The 90 bres run perpendicular to the plane of view and so only bres ends (which appear circular) can be seen. The 0 layers run perpendicular to the 90 bres, and appear as horizontal segments. As the hole edge is approached the 0 bres change from horizontal
600

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

Bearing Strain (%)


140 120

Bearing Stiffness (MPa)

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Fig. 12. SEM micrograph of a bearing specimen (w/d = 6, e/d = 6) at 100% of the failure load (i.e. the peak load in the test).

600

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

Bearing Strain (%)


140 120

Bearing Stiffness (MPa)

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Fig. 13. SEM micrograph of a bearing specimen (w/d = 6, e/d = 6) tested beyond the point where the peak load occurs. Magnication: 30.

916

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

segments to elliptical shapes. This eect occurs after bre buckling and is due to the bres being rotated out of the plane of view. Regions of delamination, on occasion, show up as bright white areas in the SEM images, due to particles of glass (that get removed during the grinding process) gathering in the cracks and then becoming highly charged by the SEM beam. Extensive plastic (i.e. non-recoverable) strains also occur in all layers as can be clearly seen in Fig. 6. Of note is the

600

Bearing Stress (MPa)

500 400 300 200 100 0 0 2 4 6 8 10

protrusion of the glass layers (particularly the 0 layer) beyond the ends of the aluminium layers at the right of Fig. 6. Fig. 7 postulates the process by which this occurs. During loading all layers are in contact with the pin so deform by the same amount. However, on unloading (removal of the pin), the glass layers evidently recover their elastic strain and spring back. For the aluminium layers a greater proportion of the total strain is plastic rather than elastic so a much lower level of springback occurs. To accommodate the dierent levels of springback in each layer the bonding between the layers must be lost, i.e. delamination must occur so this protrusion of the glass layers is only seen at high load levels where delamination has occurred. 3.3.1. Bearing joints (w/d = 6, e/d = 6) For the bearing joints most of the damage is contained in front of the bolt as seen in Fig. 4a. Hence to examine the damage progression, the bearing plane, highlighted in Fig. 8, was exposed and examined using SEM. The SEM images at 80%, 90%, 95%, and 100% of the failure load are shown in Figs. 912, while Fig. 13 shows the joint when it is deformed beyond the point where the peak load occurs. The dierence between the micrographs at 60% and 80% of the failure load is minimal, so the 60% micrograph is not shown. At 80% of the failure load, the bearing stiness versus bearing strain graph in Fig. 9 shows that the joint is well into its non-linear range of behaviour and has experienced approximately a 50% loss in stiness compared to its initial value. From the micrograph, plastic deformation of the aluminium layers is clearly visible. A small amount of bre-kinking is also evident in the 0 plies (the poisson expansion of the aluminium layers encourages the upper 0 ply to kink downwards and the lower 0 ply to kink upwards). Very minor delamination is evident between the aluminium and 0 plies allowing slight protrusion of the glass layers from the aluminium layers when the pin loading is removed, as seen at the right of the gure. No evidence of bre fracture or matrix cracking is present. The initial stiness loss can thus be attributed to the aluminium alloy entering its non-linear region of behaviour (and thus signicantly reducing in modulus) and (most likely to a lesser extent, given the small thickness of the glass layers) kinking of the 0 glass layers.

Crosshead Displacement (mm)

Fig. 14. (a) Bearing stress versus crosshead displacement curve for test carried out to 9.5 mm crosshead displacement and (b) macro image of the specimen.

Fig. 15. Macro image of specimen tested to failure under load control.

Fig. 16. Planes examined for the shear-out specimens (w/d = 6, e/d = 1.3).

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

917

At 90% of the failure load (Fig. 10) the same damage phenomena exist, but at higher levels. Plastic deformation in the aluminium plies has increased and delamination between the outer aluminium layers and their adjacent 0 prepreg layers is now clearly evident. Delamination has also started between the 0 and 90 glass plies. Fibre kinking has now progressed to a microbuckle in the 0 glass layers and some bres have broken as a result. At 95% of the failure load (Fig. 11) the delaminations have grown signicantly, along with increased plastic deformation of the aluminium plies and micro-buckling of the 0 glass layers. Fig. 12 shows the joint tested to 100% of the failure load (i.e. to the peak load seen in the test, before the load starts to drop o). In terms of damage there is quite a large difference from the 95% loading case. The bres in the 0 layers have completely buckled with extensive bre failure, and the associated loss in stiness has allowed the pin to bear deep into the sample, leading to large shear cracks running at an angle through the full thickness of both the 0 and 90 plies, and causing the aluminium layers to bend outwards. Between the 95% and 100% load level, the stiness has continued to drop and reaches zero at the point this micrograph was taken. Fig. 13 shows what happens when the pin displacement increases beyond the point where the peak load occurs. The material now oers no resistance to increased pin displacement and in fact the load drops o. All the previous damage mechanisms are suddenly hugely amplied. Critically,

delamination grows rapidly and the aluminium layers completely separate from the glass layers and bend far out of plane causing complete loss of joint stiness. The glass layers also separate from each other and massive damage occurs in the bre and matrix phases of both layers. The sudden and sharp growth in damage between Figs. 12 and 13 prompts the question of what happens if the pin displacement continues beyond the level in Fig. 13. To investigate this issue, a bearing GLARE specimen was loaded to 9.5 mm machine stroke. Because of fears of damage to the displacement gauge, it could not be used in this experiment. The experiment, as for all experiments presented thus far, was performed under displacement control. Fig. 14 shows the resulting bearing stress versus crosshead displacement curve and a macro-image of the specimen. As can be seen, after the initial peak load is reached, the load drops to about 45% of the peak value and never recovers above 55% of the peak value. This is in contrast to the behaviour of the carbonbre composites seen in [3033] where after initial load drop-o, continued pin displacement resulted in the load increasing again to close to, or even higher than, the original peak value. However, it is noted that the FML tested here is not unique in that other material systems show the same severe load drop after bearing failure occurs [34]. This load-drop o should mean that if the FML was tested under load control, the failure should be sudden and complete once the rst load peak is reached. This was conrmed by a test done under load control once the

600

140 120

Bearing Stiffness (MPa)

Bearing Stiffness (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 -40 -60 0 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 17. SEM micrographs of the shear-out plane in a shear-out specimen (w/d = 6, e/d = 1.3) at 80% of the average failure load (35 Magnication).

918

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

maximum load was reached, the pin immediately sliced completely through the specimen (see Fig. 15 for the failed specimen). The reason for this behaviour appears to be that once delamination of the metal layers from the glass layers begins, as seen in Fig. 13, the thin outer metal layers are unsupported and hence can be easily cut through, causing them to coil up and oer little resistance to further deformation. Evidence for this theory is seen in Fig. 15 which shows the load-controlled specimen and the two coiled-up remains of the portion of the outer aluminium plies in front of the pin. The coiled up metal remains have

very few bres attached to them, indicating the aluminium layers completely delaminated from the glass layers, as the pin sliced through the laminate. The practical implication of this is that designers should be wary of considering bearing failure in this material as a non-catastrophic failure mode. Of course, the addition of lateral support through washers and bolt torquing would change this behaviour and cause a signicant increase in the ultimate load reached [25], but in the pin-loaded situation, once the peak load is reached under load control, failure appears to be catastrophic.

600

140 120

Bearing Stiffness (MPa)

Bearing Stiffness (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 -40 -60 0 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 18. SEM micrographs of the shear-out plane in a shear-out specimen (w/d = 6, e/d = 1.3) at 90% of the failure load: (a) 35 magnication; (b) 180 magnication.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

919

3.3.2. Shear-out joints (w/d = 6, e/d = 1.3) For the shear-out joints, bearing damage was evident in front of the bolt, in addition to damage along the shear-out planes, as shown in Fig. 4b. Hence, for this joint type, both the bearing and shear-out planes, shown schematically in Fig. 16, were sectioned and examined. The bearing plane micrographs for the shear-out specimen revealed that the damage mechanisms for this joint

at the bearing plane were very similar to that of the bearing specimens discussed previously (Section 3.3.1). In the interests of brevity then the bearing plane micrographs are not shown. However, it should be noted that even though the mechanisms that occur are similar there are dierences in the stress, stiness and strain at which these mechanisms occur. Note also that in the fully failed specimen, extensive damage was seen with delamination between all plies,

600

140 120

Bearing Stiffness (MPa)

Bearing Stiffness (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 -40 -60 0 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 19. SEM micrographs of the shear-out plane in a shear-out specimen (w/d = 6, e/d = 1.3) at 95% of the failure load: (a) 35 magnication; (b) 250 magnication.

920

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

600

140 120

Bearing Stiffness (MPa)

Bearing Stiffness (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 -40 -60 0 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 20. SEM micrographs of the bearing plane in a shear-out specimen (w/d = 6, e/d = 1.3) at 100% of the failure load: (a) 35 magnication; (b) macro image.

severe plastic deformation of the aluminium layers and large areas of matrix cracking, particularly in the 90 layer. Thus even though the nal failure mechanism in this specimen was shear-out and the e/d ratio was only 1.3, a substantial amount of bearing damage occurred. The SEM images at the shear-out plane at 80%, 90%, 95% and 100% of the average failure load are shown in Figs. 1720, respectively. Micrographs taken at 60% of the failure load showed minimal dierence to those taken at 80% of the failure load, so are not shown. At 80% of the failure load there is little evidence of damage, as shown in Fig. 17. However, when the load reaches 90% of the failure load (Fig. 18) there is clearly signicant damage present in the form of matrix cracking, bre failure and small areas of delamination between the 0 layers and the outer aluminium plies. Interestingly the 80% and 90% cases span a range of 812% bearing strain. This is the range where the rate of stiness loss for the shear-out specimen changes from being similar to that in the bearing specimen to being similar to that in the net-tension specimen (see Fig. 5). We postulate that the reason for the signicant reduction in the rate of stiness loss at around 5% bearing strain in the shear-out specimen, is because the bres in the glass plies start to become loaded in tension

Fig. 21. Planes under examination in the net-tension specimens (w/d = 2, e/d = 6).

as the hole elongates. This increases the resistance to further loading. The reason of the overall joint stiness does

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

921

not actually increase, is because the bearing plane (see Section 3.3.1) continues to experience reductions in stiness, osetting the increase in stiness due to the tension loading of the bres in the shear-out plane. The damage at the shear-out plane increases after this and at 95% of the failure load a signicant amount of

matrix cracking in the 90 layers and bre failures in the 0 layers are evident, as shown in Fig. 19. A fully failed specimen is shown in Fig. 20 where it can be seen that a plug of material ahead of the pin is forced out of the specimen. The damage mechanisms present at this point include extensive delamination, bre failure in the 0 layers, matrix

600

140 120

Bearing Stiffness (MPa)

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 22. SEM micrograph of the shear-out plane in a net-tension specimen (w/d = 2, e/d = 6) at 80% of the average failure load.

600

140 120

Bearing Stiffness (MPa)

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 23. SEM micrographs of the shear-out plane in a net-tension specimen (w/d = 2, e/d = 6) at 90% of the failure load (37 magnication).

922

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

cracking in the 90 layers and failure of the aluminium layers, leading to complete rupture through the thickness of the specimen. 3.3.3. Net-tension joints (w/d = 2, e/d = 6) The net-tension joints also displayed bearing damage at the hole, in addition to damage across the net-tension plane, as shown in Fig. 4c. While it would initially appear logical to examine the net-tension plane, it was found that damage at this plane occurred very rapidly at or close to the nal failure load so it was more benecial to examine the formation of this damage at the hole edge. Hence, a plane parallel to the hole edge (i.e. the shear-out plane) was sectioned and examined, as illustrated in Fig. 21. In addition, bearing damage at the bearing plane was also examined. As for the shear-out joints, the micrographs of the bearing plane are not shown since it was found that

the damage progression in the net-tension joints was quite similar to that of the bearing joints analysed in Section 3.3.1. Therefore the damage mechanisms shown in Figs. 912 are common to the net-tension specimens also. Note that at 100% of the failure load, damage in the bearing plane was signicant even though this is a net-tension specimen with a w/d ratio of only 2. The SEM images at the shear-out plane at 80%, 90%, 95% and 100% of the average failure load are shown in Figs. 2225, respectively. Micrographs taken at 60% of the failure load showed minimal dierence to those taken at 80% of the failure load, so are not shown. From Figs. 22 and 23, it can be seen that very little damage is present at the shear-out plane up to 90% of the average failure load. However, by 95% (Fig. 24) the amount of damage has increased noticeably, and includes bre failures. The micrographs at 80%, 90% and 95% of the failure

600

140 120

Bearing Stiffness (MPa)

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 24. SEM micrographs of the shear-out plane in a net-tension specimen (w/d = 2, e/d = 6) at 95% of the failure load: (a) 35 magnication; (b) 180 magnication.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

923

600

140 120

Bearing Stiffness (MPa)

Bearing Stress (MPa)

500 400 300 200 100 0 0 5 10 15 20 25 30

100 80 60 40 20 0 -20 0 -40 -60 5 10 15 20 25 30

Bearing Strain (%)

Bearing Strain (%)

Fig. 25. SEM micrographs of the shear-out plane in a net-tension specimen (w/d = 2, e/d = 6) at 100% of the failure load: (a) 35 magnication; (b) 180 magnication.

load span a range of 711% bearing strain. Over this range, the stiness of the shear-out and net-tension specimens is very similar (see Fig. 5). Thus it seems that the overall stiness of the shear-out and net-tension specimens over this range of bearing strain is primarily inuenced by a combination of tension loading of the bres in the shear-out plane and compressive plastic deformation of the aluminium layers at the bearing plane, resulting in a similar overall stiness for the joints. Fig. 25 shows the shear-out plane at 100% of the failure load for net-tension specimens. Fig. 4c shows that one side of the net-tension joint failed completely while the other side remained intact. Fig. 25 shows the side of the tension specimen that did not fully rupture. Assuming that the damage level was similar on both sides of the hole just before catastrophic rupture, Fig. 25 thus indicates the level and type of damage present in the specimen just before

nal failure. Clearly, extensive bre damage is present in the 0 ply, extending through almost half of the ply thickness in the region highlighted. A matrix crack in one of the 90 layers extends fully through the thickness of the ply. The lack of damage below 95% of the failure load followed by the sudden occurrence of signicant damage at 95% highlights the catastrophic nature of the net-tension failure mode. 4. Concluding remarks In this paper, the damage and failure processes in pinloaded specimens made from the glass-based FML GLAREhave been examined in detail. Bearing stress, strain and stiness variations have been presented and scanning electron microscopy was used to investigate the damage progression with increasing load. The nal failure of the

924

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925

specimens was a result of the accumulation of several mechanisms, including plastic deformation of the aluminium plies, bre buckling and fracture, delamination between some or all the plies and matrix cracking. Bearing, shear-out and net-tension failure modes were all successfully promoted by varying the e/d and w/d ratios of the joints. However, to achieve shear-out a very low e/d ratio (1.3) was required and to obtain net-tension failures a relatively low w/d ratio (2) was needed. Even in those specimens substantial damage occurred at the bearing plane, indicating this materials high susceptibility to bearing failure. Bearing specimens were found to have the highest initial stiness and ultimate strength, followed by shear-out specimens, and then net-tension specimens. All three types of specimen exhibited initiation of non-linear behaviour at a similar bearing strain (or pin displacement) value. Micrographs indicated that, in all three specimen types, this initial non-linear behaviour was due primarily to plastic deformation of the aluminium plies, in addition to some bre-kinking in the 0 glass plies, under compressive loading at the bearing plane. Bearing specimens showed a steep drop-o in stiness up to the maximum load, while net-tension specimens showed a more gradual stiness reduction until just before nal failure. Interestingly, shear-out specimens initially showed a similar rate of stiness loss to the bearing specimens, but later in the loading sequence mirrored the slower rate of stiness loss of the net-tension specimens. This reduction in the rate of stiness loss in the latter stages of the shear-out specimen tests is believed to be due to the activation of tensile loading of the glass bres as the hole elongates. Evidence for this was seen in the development of broken bres in the range of 712% bearing strain for both the shear-out and net-tension specimens. In all three joint types, at high loads the glass layers (particularly the 0 plies) appeared in the bearing plane micrographs as protruding out from between the aluminium layers at the hole surface. This is believed to be due to the bres recovering the elastic portion of their strain in a springback eect when the pin loading is removed, while the aluminium plies do not spring back due to the fact that most of their strain is plastic in nature. To allow protrusion of the glass layers, delamination would have to have occurred between the layers. For the bearing specimens, once the rst peak load was reached, further displacement of the pin resulted in large reductions in force under displacement-controlled loading, and even with very large pin displacements, the load did not recover to anywhere near the initial peak value. Under force control, once the initial peak load was reached (at just 0.45 mm pin displacement), catastrophic failure occurred with the pin almost instantaneously slicing through to the end of the specimen. Microscopy indicated that at the initial peak load, the outer aluminium plies completely delaminated from the glass layers, and the force-controlled test indicated that this delamination grew in an unstable

fashion causing sudden failure of the joint. This behaviour is in contrast to behaviour seen in some carbonbre composites, where a more gradual failure is generally seen in bearing mode failure specimens. The work performed in this paper is currently being used in the development of a computational damage model for FML joints and this work will be reported on in a future publication. From the results in this paper, it is clear that in order for any numerical model to realistically predict FML joint behaviour it must be capable of predicting both in-plane damage, such as matrix and bre damage, as well as inter-ply delaminations. Due to the large proportion of metal layers in the FMLs tested here, the elasticplastic behaviour of these layers must also be correctly characterised. Acknowledgement The authors would like to acknowledge the Irish Research Council for Science, Engineering and Technology (IRCSET) for funding this work. References
[1] Vlot A, Vogelesang LB, De Vries TJ. Towards application of bre metal laminates in large aircraft. Aircraft Eng Aerosp Technol 1999;71(6):55870. [2] Gunnink JW, Vlot A, De Vries TJ, Van Der Hoeven W. GLARE technology development 19972000. Appl Compos Mater 2002;9:20119. [3] Vlot A. Impact loading on bre metal laminates. Int J Impact Eng 1995;18(3):291307. [4] Vogelesang LB, Vlot A. Development of bre metal laminates for advanced aerospace structures. J Mater Process Technol 2000;103:15. [5] Vlot A, Gunnink JW. Fibre metal laminates: an introduction. Dordrecht: Kluwer Academic Publishers; 2001. [6] Woerden HJM, Sinke J, Hooijmeijer PA. Maintenance of GLARE structures and GLARE as riveted or bonded repair material. Appl Compos Mater 2003;10:30729. [7] Abrao AM, Faria PE, Campos Rubio JC, Reis P, Paulo Davim J. Drilling of ber reinforced plastics: a review. J Mater Process Technol 2007;186(13):17. [8] McCarthy MA, Xiao JR, Petrinic N, Kamoulakos A, Melito V. Modelling of bird strike on an aircraft wing leading edge made from bre metal laminates part 1: material modelling. Appl Compos Mater 2004;11:295315. [9] McCarthy MA, Xiao JR, McCarthy CT, Kamoulakos A, Ramos J, Gallard JP, et al. Modelling of bird strike on an aircraft wing leading edge made from bre metal laminates part 2: modelling of impact with SPH bird model. Appl Compos Mater 2004;11:31740. [10] Wu G, Yang JM. The mechanical behaviour of GLARE laminates for aircraft structures. JOM 2005;57(1):729. [11] Wu HF, Wu LL, Slagter WJ. An investigation on the bearing test procedure for bre-reinforced aluminium laminates. J Mater Sci 1994;29:4592603. [12] Van Rooijen RGJ, Sinke J, De Vries TJ, Van Der Zwaag S. The bearing strength of ber metal laminates. J Compos Mater 2006;40(1):519. [13] Meola M, Squillace A, Giorleo G, Nele L. Experimental characterization of an innovative glare bre-reinforced metal laminate in pin bearing. J Compos Mater 2003;37(17):154352.

R.M. Frizzell et al. / Composites: Part B 39 (2008) 907925 [14] Icten BM, Sayman O. Failure analysis of pin-loaded aluminiumglass-epoxy sandwich composite plates. Compos Sci Technol 2003;63:72737. [15] Wu HF, Slagter WJ. Parametric studies of bearing strength for bre metal laminates. J Aircraft 1994;31(4):93645. [16] Yan Y, Wen WD, Chang FK, Shyprykevich P. Experimental study on clamping eects on the tensile strength of composite plates with a bolt lled hole. Compos: Part A 1999;30:121529. [17] Hollmann K. Failure analysis of bolted composite joints exhibiting in-plane failure modes. J Compos Mater 1996;30(33):35883. [18] Xiao Y. Bearing deformation behaviour of carbon bismaleimide composites containing one and two bolted joints. J Reinf Plast Compos 2003;22(2):16982. [19] Wang HS, Hung CL, Chang FK. Bearing failure of bolted composite joints. Part 1: experimental characterisation. J Compos Mater 1996;30(12):1284313. [20] Vangrimde B, Boukhili R. Descriptive relationships between bearing response and macroscopic damage in GRP bolted joints. Compos Part B 2003;34:593605. [21] Lawlor VP, Stanley WF, McCarthy MA. Characterisation of damage development in single shear bolted composite joints. Plast Rubber Compos 2002;31(3):12633. [22] Petrinic N. High strain rate mechanical properties of GLARE. CRAHVI deliverable report D1.1.8. UK: University of Oxford; 2002. [23] Beumler T. Flying GLARE a contribution to aircraft certication issues on strength properties in non-damaged and fatigue damaged GLARE structures. PhD thesis, Delft University of Technology; 2004. [24] De Vries TJ. Blunt and sharp notch behaviour of glare laminates. PhD thesis, Delft University; 2001.

925

[25] Van Rooijen RGJ. Bearing strength characteristics of standard and steel reinforced GLARE. PhD thesis, Delft University of Technology; 2006. [26] Slagter WJ. On bearing strength of bre metal laminates. J Compos Mater 1992;26(17):254266. [27] Camanho PP, Bowron S, Matthews FL. Failure mechanisms in bolted CFRP. J Reinf Plast Compos 1998;17(3):20533. [28] Hart-Smith LG. Mechanically fastened joints for advanced composites-phenomenological considerations and simple analysis. In: Proceedings of 4th conference on brous composites in structural design. San Diego; November, 1978. p. 54374. [29] McCarthy MA, Lawlor VP, Stanley WF, McCarthy CT. Bolt-hole clearance eects and strength criteria in single-bolt, single-lap, composite bolted joints. Compos Sci Technol 2002;62:141531. [30] Maikuma H, Kubomuro K. Bearing strength and damage progress for PAN-based and pitch-based carbon ber composites. J Compos Mater 1993;27(18):173961. [31] Barboni R, Carlini S, Gaudenzi P, Scarponi C. Importance of 3D stress eld in the failure of pin-loaded joints. In: Proceedings of 7th ICCM international conference on composite materials. Beijing; August, 1989. p. 18792. [32] DiNicola AJ, Fantle SC. Bearing strength behaviour of clearance-t fastener holes in toughened graphite/epoxy laminates. Compos Mater Test Des ASTM STP 1206 1993;11:22037. [33] Aktas A, Dirikolu MH. The eect of stacking sequence of carbon epoxy composite laminates on pinned-joint strength. Compos Struct 2003;62:10711. [34] Caprino G, Giorleo G, Nele L, Squillace A. Pin-bearing strength of glass mat reinforced plastics. Compos: Part A 2002;33:77985.

You might also like