BMI composite at high operating temperatures

Composites: Part A 41 (2010) 902–912

Contents lists available at ScienceDirect

Composites: Part A journal homepage: www.elsevier.com/locate/compositesa

Bonded repairs for carbon/BMI composite at high operating temperatures A.N. Rider *, C.H. Wang 1, P. Chang Air Vehicles Division, Defence Science and Technology Organisation, 506 Lorimer St., Fisherman’s Bend, Victoria 3207, Australia

a r t i c l e

i n f o

Article history: Received 25 October 2009 Received in revised form 9 March 2010 Accepted 10 March 2010

Keywords: A. Prepreg B. Adhesion C. Stress concentrations D. Mechanical testing

a b s t r a c t The increased use of carbon reinforced polymer composite structures in civilian and military aircraft has produced new challenges for on-aircraft bonded repairs. Carbon/bismaleimide (BMI) composite structure provides an added complexity associated with the application of adhesively bonded repairs. Commercially available carbon/BMI composite usually requires a post-cure temperature in excess of 220 °C to achieve the high strength properties at elevated operating temperatures. Application of bonded repairs in situ often places an upper limit on the temperatures that can be employed for curing the adhesive at 177 °C. Consequently, the adhesive bond needs to achieve similar mechanical properties to the parent matrix material without the benefit of the high post-cure temperature. The current work examines a range of repair options that can be used to recover strength and the selection of adhesives and processes to successfully apply the repairs using vacuum assisted pressure. Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved.

1. Introduction Carbon/bismaleimide (BMI) composite may be employed in critical load-bearing aircraft structures that experience high operating temperatures. The upper temperature testing by the manufacturer of CycomÒ 5250-4 prepreg system with IM7 carbon fibres was around 177 °C in the wet condition [1]. By contrast CycomÒ IM7/977-3 has an upper use temperature around 132 °C [2]. The relative compression and short beam shear strengths for these two systems indicates the carbon/BMI system can offer 15–30% wet strength improvement over IM7/977-3 at 132 °C or equivalent wet strength at 20 °C to 60 °C higher temperatures [1,2]. Despite the advantages offered by carbon/BMI for higher temperature structural applications, difficulties arise in undertaking in situ repairs where the high temperature properties need to be restored. The upper cure temperature for on-aircraft repairs may be limited to 177 °C to remain below the auto-ignition temperature for the aviation fuel and prevent high temperature transmission into aluminium substructure. High internal vapour pressures generated during high temperature cure may also lead to separation of the composite skin from honeycomb skins. Whilst the carbon/BMI achieves the high temperature properties with a post-cure temperature in excess of 220 °C, the repair adhesive needs to re-establish the BMI properties with the 177 °C cure temperature limit. A further complication for in situ repairs is pressure application during adhesive cure. Pressure may be applied using * Corresponding author. Tel.: +61 3 9626 7393; fax: +61 3 9626 7174. E-mail address: [email protected] (A.N. Rider). 1 Present address: School of Aerospace, Mechanical and Manufacturing Engineering, RMIT University, GPO Box 2476, Melbourne, Victoria 3001, Australia.

vacuum bag arrangements, to reduce the foot-print of the repair equipment. As most adhesives are designed for use in high pressure autoclaves, vacuum assisted pressure may cause high bondline porosity [3]. Previous work on bismaleimide adhesive using vacuum pressurisation found that voiding could be reduced if the vacuum was released at the adhesive flow temperature and a small positive pressure was maintained during cure [4]. If small positive pressure can be applied during aircraft repair this technique could be considered, however, additional equipment requirements would need to be assessed. The bonded repair should restore the original stiffness, static strength, durability and damage tolerance [5]. Creep of moisture laden adhesive under high temperature loading also needs consideration [6,7]. Common bonded repair designs [8,9] performed on the aircraft structure may be limited to options such as an external doubler, step-lap or scarf repair (Fig. 1). The bonded doubler strength may be compromised by geometrically non-linear bending, which increases the stress concentration adjacent to the damage cut-out region [10]. By contrast, the scarf or step-lap configuration should minimise secondary bending effects, but will be more difficult to apply. Design of the bonded repair must consider the full temperature and humidity environment that the component will experience. At high operating temperatures the effect of moisture on the adhesive and composite may be significant. The strength of the adhesive and composite matrix is closely related to their glass transition temperatures, Tg, which are highly dependent on the cure temperature and moisture state. Unlike a metallic patch over metallic structures, bonded composite repairs are permeable to moisture throughout the length of the joint and consequently the bondline will absorb

1359-835X/$ - see front matter Crown Copyright Ó 2010 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesa.2010.03.006

903

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

Fig. 1. Repair configurations that may be employed in cases where only single-sided access is available (a) doubler, (b) step-lap, (c) scarf and (d) scarf-doubler bonded repairs.

moisture at a significantly faster rate than equivalent metallic structures. Consequently, the limiting strength for composite repairs bonded to high temperature composite materials, such as carbon/ BMI, may often be in the hot/wet condition [10]. Previously, work has examined the use of bismaleimide adhesive in conjunction with a second low temperature adhesive to provide optimal joint properties over a large temperature range [4,11,12]. The use of a low temperature adhesive with the high temperature BMI adhesive worked for dissimilar adherends where the overall joint strength was limited by the low temperature strength of the brittle BMI adhesive. Whilst this work provided a novel solution to the low temperature strength of some BMI adhesives it is inapplicable to the current study where similar composite adherends are used. Additionally, BMI adhesives cannot easily be used for repair, due to their high post-cure temperature, which precludes on-aircraft use. The following work examines the design and application of bonded repairs to carbon/BMI composite structure. Repair options considered single-sided access, cure temperature limited to 177 °C and vacuum pressurisation for usage temperatures between 55 °C (cold/dry) and 177 °C in the wet condition (hot/wet). Potential adhesives were screened using dynamic mechanical thermal analysis (DMTA) and differential scanning calorimetry (DSC) [13] and surface treatment options were examined. Validation of repair design was performed using representative joints. The overall aims of the work were to develop practical repair techniques for high temperature, highly loaded carbon/BMI composite structure where the repairs were required to be applied in situ and the repair adhesive cure was limited to 177 °C.

2. Experimental 2.1. Characterisation and testing Cytec’s oxyamide film adhesive FM32 (488 g/m2) and epoxy film adhesive FM355 (366 g/m2) were cured 177 °C /4 h and 177 °C/1 h, respectively. AF131-2 (366 g/m2) epoxy film adhesive from 3 M was cured 90 min at 177 °C. Cytec IM7-G/5250-4 carbon/BMI unidirectional prepreg tape (145 g/m2) was cured at 586 kPa and 177 °C for 6 h with a 226 °C/ 6 h post-cure. The adhesive glass transition temperature (Tg) used single cantilever bending of 0.1% strain at 1.0 Hz and a ramp rate of 5 °C/min. The Tg was measured from the onset of the storage modulus change for the adhesive, in the dry and moisturised state. Differential scanning calorimetry (DSC) measured the extent of cure from the residual exotherm of the cured adhesive samples. Mini-scarf joints used 12 ply unidirectional composite adherends with a 3° scarf angle (Fig. 2) and the surface treatment procedures detailed in Table 1 were characterised using contact angle and surface analysis measurements [13]. The 3° scarf angle was machined using a computer controlled router with a diamond encrusted tool-bit. Bonding at 275 kPa or 100 kPa used methods provided in Table 1 with an envelope bag to restrain adhesive flow. Conditioned joints had 1.2% moisture uptake. Thick adherend lap shear testing used Al2024–T351 aluminium in accordance with the methods detailed in ASTM D5656 [14] and a grit-blast and epoxy silane treatment [15]. Moisture conditioning took 75 to 100 days at 70 °C/95% R.H. Mode II testing used the three-point bend end-notch flexure test [16] with 15-ply unidirectional laminates bonded using FM32.

25mm

[0]12

3°

Fig. 2. Mini-scarf configuration used to examine adhesive shear strength of composite bonded joints.

1.62 mm

904

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

Table 1 Surface treatment and bonding methods used to construct the adhesive bonded joints using either autoclave or vacuum assisted pressure. Description Surface treatment Abraded

Routed surfaces were sanded with 400 grit Al2O3 paper, cleaned with methyl ethyl ketone (MEK) and deionised water, oven dried at 110 °C Oxygen plasma treat ‘‘abraded” composite at 140–200 W at a rate between 2.5 mm and 5.0 mm/s using 30 L/min of Helium and 0.5– 1.5 L/min of Oxygen from a standoff distance of 2.5 mm.

Plasma Lay-up method B-staged Taped + Hot Debulk with CerexÒ

Adhesive heated at 80 °C/80 kPa vacuum 20 min prior to adhesive joint construction Flashbreaker tape fully encloses bondline Joint is debulked at 80 °C and 100 kPa vacuum for 20 min, cooled, and taped fully. CerexÒ breather veil is placed on the adhesive surface prior to placement of the two adherends.

2.2. Composite joint design

where

The standard joint configurations examined are shown in Fig. 1a and c and a novel design concept previously described [10,17] was also investigated, in which a doubler was bonded over a scarf joint to produce a scarf-doubler configuration, as shown in Fig. 1d. Typically, 30-ply orthotropic laminates of IM7/5250-4 with a stacking sequence [45/02/45/90]3s, referred to as stiff laminates, were used. Additionally, [0]12, [45/90/45/03]3s and [45/0/45/90]4s, referred to as unidirectional, hard and quasi-isotropic, respectively, were also examined. An attempt was made to understand the distribution of the adhesive shear stresses along the scarf joint using linear finite element (FE) analysis. The shear stress along the adhesive bondline was extracted at the mid-point of the adhesive layer and the stress concentration, Kt, was measured as the maximum adhesive bondline stress divided by the average shear stress. Fig. 3 shows that the FE analysis predicts that the maximum Kt, would decrease as the total percentage of the zero-degree plies in the laminate increases. The analysis assumed that the stacking sequences of the scarf insert and doubler matched the parent laminate. The Kt values in the adhesive peak at the zero-degree plies along the scarf bondline. The predicted doubler joint strengths used equations detailed below [10,18], which account for geometric non-linear bending of the single-sided doubler. The maximum uniaxial tensile stress, rmax, that a one-sided doubler joint (Fig. 1a) can carry corresponds to when the peak adhesive shear strain reaches the failure value, which is given by Eq. (1); a detailed derivation is included in Appendix A.

S¼

Kt (Shear stress /average shear stress)

rmax

ð1 þ SÞsy ¼ ks ts

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2Ga cf  1;

ð1Þ

sy

Ed td ; Es ts

ks ¼

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   4Ga 1 1 þ : g Es t s Ed t d

1.5

1.0 y = -0.73 ln(x) + 4.4

0.5

0 0

20

40 60 80 Percentage of 0° plies (%)

100

Fig. 3. Linear finite element analysis of Kt for carbon/BMI laminates with different 0 degree ply percentages.

ð3Þ

The adhesive yield shear stress, sy, is determined from the shear stress–strain curve of the adhesive. The parameters Ga and cf denote, respectively the adhesive shear modulus and failure strain. The adhesive thickness corresponds to g and the subscripts s and d represent parameters pertinent to the parent structure and doubler, with E and t denoting the modulus and thickness, respectively. The double-sided doubler joint strength (a second matching doubler is bonded to the configuration shown in Fig. 1a, where the combined doubler thickness equals the parent thickness) was obtained by replacing ks in Eq. (1) with Eq. (4) for the shear transfer parameter, k, representing the case of no outof-plane bending [20],

k¼

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   Ga 1 1 þ : g Es ts Ed td

ð4Þ

In the case of the scarf joints (Fig. 1c), Eq. (5) was used to predict the maximum joint strength [19]:

rmax ¼

sf cos a sin a

;

ð5Þ

where sf is the ultimate shear strength determined from the thick adherend lap shear test, and a is the scarf angle. The predicted maximum scarf-doubler joint strength used Eq. (6) [10]:

rmax ¼

2.0

ð2Þ

2ð1 þ SÞsf ; sinð2aÞ

ð6Þ

where a is the scarf angle and S in the stiffness ratio of the doubler and parent laminate as given by Eq. (2). The laminate stiffness was determined using the laminate theory with manufacturer’s datasheet [1] for the carbon/BMI composite at the different test conditions. Eq. (6) assumed that the secondary bending effect caused by the addition of the doubler is relieved by the geometrically non-linear deflection. Some linear FE analysis of the bondline stress distribution was also undertaken for the stiff laminate in the scarf and scarf-doubler joint configuration to assess Eqs. (5) and (6) and estimate peak stresses that may develop along the scarf bondline. Fig. 4 shows that due to the stiffness variation in the thickness direction, the adhesive stress concentration occurs near where zero-degree plies terminate on the scarf, which has been identified previously [19]. However, the analysis indicates that the stresses in the scarf and

905

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

Kt (Shear stress /average shear stress)

1.80 1.60 1.40 1.20 1.00 0.80 0.60 Scarf-doubler joint

0.40

Scarf joint

0.20 0.00 0.0

0.2

0.4 0.6 Relative distance along scarf

0.8

1.0

Fig. 4. The distribution of adhesive shear stresses for adhesively bonded scarf and scarf-doubler joints for a [45/02/45/90]2s laminate with bending constraint.

scarf-doubler joint are similar, which supports the assumptions in Eq. (6). It should be noted that the Kt determined from the FEA analysis was relatively insensitive to laminate thickness when bending was constrained. Eqs. (5) and (6) assume that the scarf bond line stress is uniform, whereas the FE analysis shows that peak stresses along the adhesive may develop, which could potentially reduce the predicted strength of the joint. It might be expected that the extent of adhesive plasticization would increase the bondline stress uniformity and that the Kt would, therefore, be more significant at lower operating temperatures. A more detailed non-linear FEA model would be required to fully assess the effect of adhesive properties on the bondline Kt at the maximum joint strength. FEA used StressCheckÒ with 2-D elements in a 3-D coordinate reference system. The model was constrained at one end and loaded with unit stress at the opposing end. Model convergence suggested that the results had an error less than 1%. Convergence was assessed by completing analyses for polynomial levels two to six, meaning that the element order increased from two to six for the analysis.

3. Results 3.1. Adhesives, surface treatment and bonding Table 2 shows the extent of reaction and Tg for the candidate repair adhesives. Based on the wet Tg for the adhesives, the mechanical performance under hot/wet conditions should decrease in the order FM32, AF131-2 and FM355. The extent of cure for FM32, appears to be correlated with the wet Tg (Table 2), however, there is a deviation for the 2 h/177 °C cure, where the degree of cure measured with DSC would suggest a higher Tg value than measured with the DMTA. The FM355 result does not fall on the general FM32 curve, suggesting that the correlation between degree of cure and Tg would need to be established for each adhesive type.

Table 2 also indicates that generally the wet Tg reduces to a temperature around 80% of the dry Tg. Once again the 2 h/177 °C cure result for FM32 shows a bigger drop than expected. Interestingly, the 12 h/160 °C cure indicates less of a change, suggesting that subtle changes in the cure mechanisms occur between 160 °C and 177 °C. Table 3 indicates the plasma treatment alters the composite surface chemistry and enables the water to spread, corresponding to an increase in the surface energy of the composite. X-ray photoelectron spectroscopy [13] shows the plasma treatment increases the oxygen content of the surface, which corresponds with the increased wettability. Fig. 5 shows abrasion of the machined composite surface with alumina sandpaper helps remove loosely embedded carbon fibres that have been fractured during the machining step. The abrasion also removes loose resin flakes from the matrix. Loosely bound fibres and matrix left from the machining would lead to a reduction in bond strength. Techniques to enable a reduction in bondline voiding were also developed. Sources of voiding include the volatile materials contained in the adhesive and entrapped air at the adhesive-to-adherend interface. Analysis of FM32 volatile gases using Fourier transform infrared (FT-IR) spectroscopy revealed the evolution of acetone, water, and toluene during adhesive cure. Separate weight-loss measurement revealed that the total weight of the evolved gases was less than 0.5%. Efforts were then made to remove or react the volatiles from the FM32 by using vacuum heating at 80 °C and 80 kPa for 20 min. The extent of volatile evolution was observed in a vacuum oven for the staged and as-received adhesive to determine the vacuum pressure at which cure could proceed without inducing significant voiding. Fig. 6 shows that as the heating temperature increases, the vacuum pressure at which volatiles evolve decreases. However, the staging of the adhesive can increase the vacuum pressure, particularly at the lower temperatures, at which volatile evolution occurs. A 80 kPa vacuum would minimise void formation during cure. The adhesive staging at 80 °C and 80 kPa also is successful at maximising the vacuum pressure before volatiles evolve rapidly, enabling higher vacuum pressure to be applied during repair. The process of removing solvent and reacting excess volatile constituents in the adhesive will also stage the adhesive. DSC measurements of the FM32 after different staging conditions (Table 4) show that staging for 20 min between 80 °C and 120 °C produced reactions ranging between 6% and 21%, accompanied by moderate rises of the Tg. The increase in the extent of reaction reduced the adhesive tack significantly, although the adhesive flows during cure were found, by separate measurement, to not be significantly

Table 3 Surface elemental compositions and contact angle of IM7/5250-4 laminate after abrading with alumina sandpaper, solvent cleaning and plasma treatment. IM7/5250-4 Surface

%C

%O

%N

Contact angle (degrees)

Abraded Abraded + plasma

81.4 73.9

17.4 22.7

3.3 3.3

65 ± 10 <3

Table 2 Extent of reaction and, dry and wet glass transition temperatures for the three candidate repair adhesives. Adhesive

Cure

Extent of reaction (%)

Dry Tg (°C)

Wet Tg (°C) (3.7% H20)

Wet Tg/dry Tg (%)

FM355 AF131-2 FM32

1 h/177 °C 1.5 h/177C 4 h/177 °C and 2 h/204 °C 4 h/177 °C 2 h/177 °C 12 h/160 °C 4 h/160 °C

92.0 – 97.6 93.5 91.5 89.0 86.8

207 225 234 233 233 186 180

169 174 182 178 166 167 159

82 77 78 76 71 90 80

906

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

Fig. 5. SEM images taken of the IM7/5250-4 composite surface after (A) CNC routing and (B) alumina paper abrasion.

100 Staged

Volatile evolution pressure (-kPa)

98

As received 96 94 92 90 88 86 84 82 80 70

80

90

100 110 120 Temperature (°C)

130

140

150

icantly reduced voiding and the air permeable fabric layers were proven very effective in removing entrapped air at the interface between glass slide and adhesive. Previous work on BMI adhesives suggested that void reduction was also possible if vacuum was removed at the adhesive flow temperature and small positive pressure was maintained during cure [4]. The application of the second adherend at the flow point provided and even better method of reducing voids [4]. The present technique provided a better and more practical method to achieve a similar outcome by enabling entrapped air to be removed from both interfaces in situ, whilst staging the adhesive prior to joint fabrication removes the volatile components that can produce voiding during the high temperature cure. Further, the use of higher vacuum pressures can facilitate fit-up of the scarf and doublers on aircraft structure with complex curvature.

Fig. 6. Volatile evolution pressure for FM32 adhesive heated between 80 °C and 140 °C.

3.2. Composite bond strength and adhesive properties

altered. It is clear from the results shown in Table 4 that the total cure reaction of staged adhesive was clearly reduced if the initial staging was too aggressive. Mechanical strength tests of the scarf joints under hot/dry and hot/wet conditions showed that the average adhesive shear strength under these conditions (Table 4) decreased significantly when the degree of cure induced by the staging exceeded 10%. Methods to assist in removing the entrapped air at the adhesive-to-adherend interface used adhesive sandwiched between glass slides that was cured with either OptimatÒ carbon fibre veil (Optimat 20304A from Technical Fibre Products (12 g/m2) and polyester binder) or CerexÒ nylon non-woven fabric (CerexÒ 230336: Type 23-Cerex/03–10 g/m2, 76 lm thickness) at the two adhesive-glass interfaces. The glass joint was vacuum debulked at room temperature and, then at 80 °C and 100 kPa vacuum for 20 min (Table 1). Fig. 7 shows that the adhesive staging signif-

Scarf joints bonded with FM32, AF131-2 and FM355 in the moisture conditioned state (Fig. 8) shows that the shear strength increases with decreasing testing temperature. The relative ranking of the adhesives is FM32, AF131-2 followed by FM355. The relative performance is consistent over the temperature range. At 177 °C, FM32 showed clear benefit relative to the other two adhesives. FM355 and AF131-2 would need to operate at approximately 20 °C lower temperature than FM32 to achieve equivalent shear strength. FM32 would be preferred in higher temperature repair applications. All fracture surfaces appeared to propagate at the adhesive-composite interface. Work with BMI adhesive suggested that in the dry state, the strength of single lap shear (SLS) joints actually decreased at lower temperatures [4]. The brittle nature of BMI adhesives leads to low joint strength in the SLS configuration due to high peel stresses that develop at the joint edges. In the present work the use of a unidirectional ‘‘mini”-scarf joint significantly reduces the peel stresses and can enable the strength of

Table 4 The influence of staging conditions on adhesive properties before and after cure. Staging

Extent of reaction (%)

Tg (°C)

Total cure (%)

Flow (%)

Shear strength (MPa) 177/Dry

177/Wet

80 °C/20 min 100 °C/20 min 120 °C/20 min

6 10 21

9.8 14.6 19.2

94 90 79

73 72 67

28 24 14

15 15 10

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

No staging, debulked

Staged at 80°C/-80 kPa vacuum, 20 min., debulked

Staged at 80°C/-80 kPa vacuum, 20 min., debulked. Optimat® carbon fibre veil at adhesive/glass interface

Staged at 80°C/-80 kPa vacuum, 20 min., debulked. Cerex® at adhesive/glass interface

907

Fig. 7. Voiding observed between glass plates bonded with FM32 in which the glass plates were taped to constrain adhesive flow and debulked 20 min at 80 °C and 100 kPa vacuum.

Average shear stress (MPa)

30

FM32 AF 131-2

26

FM355

22 18 14

the curve and set the shear yield strength equal to the ultimate strength, resulting in a decrease in the adhesive modulus, Ga, and an increase in cy, strain yield, for the ideal elastic–plastic cases. FM32 indicates a substantial increase in shear modulus and a reduction in the failure strain for the hot/dry and hot/wet conditions compared to FM355. Mode II fracture toughness for the FM32 bonded laminates shows a strong trend between the GII value and the test condition. The failure modes also showed a change from the composite-adhesive interface for the hot/dry and hot/wet conditions to interlaminar failure of the composite for cold/dry and room temperature testing.

10

3.3. Composite joint strength 6 115

130

145 160 Test temperature (°C)

175

Fig. 8. Bond strength for composite scarf joints tested under hot/wet conditions using FM32, AF131-2 and FM355 film adhesives.

the adhesive at lower temperatures to be achieved. Attention to scarf-tip geometry, however, is critical, particularly, if trying to maximise strength for joints bonded to the brittle composite matrix, employed in the current work, at lower test temperatures. Results from thick adherend lap shear testing for FM355 and FM32 (Table 5) were used for joint strength predictions, however, these values are only fully relevant when the composite joint fails within the adhesive layer, which may often only occur when the joint is tested in the hot/wet condition. In Table 5 the subscripts f represent failure for the stress and strain, whereas, y represents the yield point from the ideal elastic–plastic fitted curve. As temperature and moisture content increases the adhesive shear strength and modulus decreases rapidly. Previous work on BMI adhesive in the dry state found that the adhesive shear strength and modulus also decreased with increasing temperature [12]. The fitting of the ideal curve maintained the same area underneath

Fig. 9 shows the relative joint strength for 40 plies [45/02/45/ 90]4s laminates with either a single-sided full-thickness doubler or double-sided 20 ply doublers bonded with FM355. The predicted strength based on Eq. (1) is reasonably close for the hot/wet case, but underestimates the hot/dry strength. In the case of the double-sided repair, both hot/dry and hot/wet results are close to the predicted values determined using the shear transfer parameter, k, defined in Eq. (4). The unnotched laminate strength, rUNS, was calculated using laminate theory based on quoted unidirectional laminate properties [1]. Clearly, more experimental points and different parent laminate thicknesses would be required to improve confidence in the model, however, for the laminate thickness examined, the model provides an approximate estimate when joint failure proceeds in the adhesive. Failure surfaces for both the hot/dry and hot/wet single-sided doubler joints (Fig. 10) indicated interlaminar failure of the parent laminate, suggesting out-of-plane bending led to peel stresses and the consequent composite fracture mode, explaining the low strength. The double-sided 20 plies doubler joints showed a significant increase in strength for the hot/dry and hot/wet cases, corresponding to more failure in the adhesive layer (Fig. 10), suggesting a reduction of the peel stresses present for the single-sided case.

908

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

Table 5 Shear stress and shear strain properties for FM355 and FM32 measured using the thick adherend metal lap shear test and the mode II fracture toughness (FM32), GII, determined using the end-notch flexure test. Adhesive

Property

56 °C/Dry

RTA

177 °C/Dry

177 °C/Wet (75 days)

FM355

Ga-ideal (MPa) sf (MPa) cf (mm/mm) cy-ideal (mm/mm)

555 (64) 48 (2) 0.09 (0.01) 0.09 (0.01)

213 (15) 64 (2) 0.55 (0.05) 0.21 (0.01)

62 (5) 26 (1) 0.75 (0.02) 0.42 (0.02)

32 (12) 13 (1) 0.51 (0.03) 0.39 (0.03)

FM32

Ga-ideal (MPa) sf (MPa) cf (mm/mm) cy-ideal (mm/mm) GII (J/m2)

509 (30) 52 (5) 0.12 (0.03) 0.10 (0.02) 670

525 (120) 49 (5) 0.15 (0.04) 0.09 (0.01) 810

249 (46) 25 (1) 0.28 (0.06) 0.10 (0.02) 2150

102 (6) 15 (1) 0.30 (0.02) 0.14 (0.01) 2185

0.9 0.8

σmax /σ UNS

0.7 0.6 0.5 0.4 0.3

1400 1200 Joint strength (MPa)

Model-double sided: Hot/Dry Model-double sided: Hot/Wet Double sided: Hot/Dry Double sided: Hot/Wet Model-single sided: Hot/Dry Model-single sided: Hot/Wet Hot/Dry-single sided Hot/Wet-single sided

1.0

1000

0.2

800 600 400 200

0.1

0 -100

0.0 0

2

4

6

8

10

-50

0 50 100 Test temperature (°C)

150

200

Parent laminate thickness, ts (mm) Fig. 9. The relative joint strength as a function of laminate thickness for single- and double-sided doubler repairs under hot/dry and hot/wet conditions.

The results demonstrate the difficulties in recovering the pristine laminate strength using single-sided doubler repairs. Recovering 50% of the original strength would require laminate thickness less than 1 mm, whereas for a double-sided repair, laminates up to 3 mm thick might be repaired. Scarf repairs that would reduce the geometrically non-linear bending were investigated. Fig. 11 shows the measured joint strength for FM355 adhesive bonded using unidirectional, stiff,

Double-sided: Hot/Wet

Fig. 11. Scarf joint strength for FM355 for dry (filled symbols) and wet (open symbols) test conditions. Points are designated for unidirectional (N), hard (), stiff (j) and quasi-isotropic laminates (d). The predicted strength is designated by the solid (dry) and broken (wet) lines.

hard and quasi-isotropic laminates over the range of test conditions. The strength of repairs operating under the hot/wet condition would at best only recover around 20% of the unnotched laminate tensile strength, which would limited the repairs to lightly loaded areas. The hot/wet and hot/dry strength all fall on or below the predicted strength based on Eq. (5). For all test conditions, the joint strength of the quasi-isotropic laminate falls below

Single-sided: Hot/Wet

Single-sided: Hot/Dry

Fig. 10. Fracture surfaces for bonded doubler butt joints.

909

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

1000

800 Joint strength (MPa)

the predicted strength. Based on Fig. 3, the quasi-isotropic laminate would be predicted to have the highest stress concentration and the extent of adhesive plasticization at failure would reduce as a function of test temperature. The general trend in Fig 11 is consistent with a bondline stress concentration affecting joint strength. However, there is insufficient data to confirm this trend in Fig 11 and, additionally, the room temperature and cold/dry joint did not fail in the adhesive layer, limiting the validity of Eq. (5). These initial results, however, suggest that bondline Kt needs to be considered when estimating the strength of scarf joints between orthotropic composite laminates [19]. Fig. 12 shows the measured strength for FM32 bonded scarf joints, prepared with the stiff laminates over the different test conditions. The joint strength at the hot/wet condition is more than 150% greater than the equivalent FM355 joint and is consistent with the increase observed for the mini-scarf joints using the two adhesives (Fig. 8), confirming the correlation in the wet Tg of the two adhesives and joint strength. The predicted strength from Eq. (5) is close to that measured for both the hot/wet and hot/dry cases, suggesting that bondline stress may have been relatively uniform at failure under these test conditions. This contrasted with FM355 scarf joints at hot/wet and hot/dry, which both showed the predicted strength of the stiff laminate joints was higher than the measured strength (Fig. 11). Given the FM355 plasticization at failure would be expected to much greater than FM32 at hot/wet and hot/dry, it might have been expected that FM355 bondline uniformity would be greater and have led to a better prediction. The much lower than predicted strength for FM32 scarf joints at room temperature (RTA) and cold/dry (CD) (Fig. 12) was associated with failure within the composite laminate. Joint strength, therefore, may not necessarily have been limited by the adhesive strength and Eq. (5) would not be valid. It is interesting to note, however, that, if the Kt determined in Fig. 4 was applied to the measured strengths of the FM32 joints at CD and RTA, a much closer prediction would be achieved, which may suggest that bondline stress distribution is influencing joint strength. In order to improve the joint strength for the hot/wet condition, a novel scarf-doubler repair scheme was investigated. In typical repairs a doubler or a scarf repair will be used, but not together. A scarf repair may have a thin doubler bonded externally to provide protection at the tips from impact damage, but a structural doubler would not be considered. The current design is considered where maximum strength is required and the adhesive properties are limited at the operating temperature. The design relies critically on the doubler failing after the scarf and significant attention to doubler taper is required. Previous results for FM355 indicated

600

400

200

0 0.0

0.2

0.4

0.6

0.8

1.0

Edtd/Ests Fig. 13. FM355 bonded scarf-doubler joint strength for stiff laminates under hot/ dry (j) and hot/wet (h) conditions. The predicted strength is designated by the solid (dry) and broken (wet) lines.

the scarf-doubler repair could increase joint strength in proportion to the ratio of the doubler and parent laminate stiffness [10]. Fig. 13 shows the strength for scarf-doubler joints bonded with FM355 for the hot/wet and hot/dry condition, where failure occurred in the adhesive layer or at the interface between the adhesive and composite matrix. For the hot/wet case it can be seen that as the doubler stiffness increases, the joint strength increases proportionally, indicating the increased stiffness of the doubler can increase the scarf joint strength. For a full-thickness doubler, around 40% of the stiff laminate unnotched tensile strength (UNS) could be recovered for the hot/wet condition, which would make the repair more applicable for structure experiencing higher strain in service. The increase in hot/wet strength is, generally, less than the strength predicted using Eq (6). The increase in the hot/dry strength of the joint is considerably less than that predicted by Eq (6), although the experimental points indicate a clear increase in strength provided by the doubler. The influence of doubler design and bondline stresses may need to be considered further for the development of a more accurate model to predict the scarfdoubler joint strength. Fig. 14 shows the strength as a function of doubler stiffness for scarf-doubler joints bonded using FM32. Both the hot/wet and hot/ dry strength approximately double when the stiffness of the doubler matches the parent laminate, which is the increase predicted by Eq. (6). By using FM32, almost 60% of the UNS can be recovered, 1200

1200 1000 Joint strength (MPa)

Joint strength (MPa)

1000 800 600 400

800 600 400 200

200 0 0 -100

0.0 -50

0 50 100 Test temperature (°C)

150

200

Fig. 12. FM32 bonded scarf joint strength for stiff laminates under dry (filled symbols) and wet (open symbols) test conditions. The predicted strength is designated by the solid (dry) and broken (wet) lines.

0.2

0.4 0.6 Edtd/Ests

0.8

1.0

Fig. 14. FM32 bonded scarf-doubler joint strength for stiff laminates under hot/dry (j), hot/wet (h), room temperature (+) and cold/dry (s) conditions. The predicted strength is designated by the grey solid (hot/dry), grey broken (wet) and black solid (room temp.) lines.

910

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

which would make the repair very applicable to highly loaded aircraft structure. In contrast the strength of the RTA and CD tested scarf-doubler joints bonded with FM32 (Fig. 14) is only 150% greater than the scarf joint for the full-thickness doubler of matching stiffness. The RTA and CD joints both failed in the first ply layer of the composite parent, with some indication that the doubler had failed before the scarf in the case of the scarf-doubler joints. Both these factors would limit the applicability of Eq. (6). The predicted strength for the RTA case is shown in Fig. 14 and is considerably higher than the measured strength for both scarf and scarf-doubler joints. 4. Discussion 4.1. Adhesive characterisation and bonding The selection of a suitable adhesive and the development of a robust method for curing bonded joints using vacuum assisted pressure required a number of important steps. Initially, the selection of a suitable adhesive indicated that DMTA could provide a good estimate of the adhesive wet Tg (Table 2). Mechanical testing confirmed a good correlation existed between adhesives’ hot/wet shear strength and the wet Tg (Fig. 8). The reduction in joint strength worsened when the Tg value fell below the testing temperature. FM32 hot/wet strength of 15 MPa compared to 12 MPa for FM355 or AF131-2 indicated that FM32 was the best candidate for high temperature repairs. The vacuum processing of FM32 was facilitated by removal of volatile components using a vacuum staging process. However, staging the adhesive beyond 10% of the full cure could have adverse effects on the adhesive mechanical properties (Table 4). DSC measurements indicated that extended staging could reduce the maximum level of cure possible and reduce flow during cure, suggesting that the minimum viscosity point was increased (Table 4). The use of a permeable fabric layer at the adhesive-substrate interface also was effective at reducing entrapped air. Staging the adhesive also helped the effectiveness of the permeable fabric layer by reducing the adhesive tack and enabling the fabric to provide a continuous path for the air to escape from the interface prior to adhesive beginning to flow (Fig. 7). The use of plasma treatment to increase the composite surface energy was beneficial in improving bonding by facilitating improved wetting of the adhesive during cure and evacuation of air at the composite to adhesive interface. As indicated by Young’s equation (cs/ v = cl/v cos h + cs/l), the competing surface tension forces, c, where s, l and v represent the solid, liquid and vapour phases, respectively, control the contact angle. As the surface energy of the composite (cs/v) increases, following plasma treatment, the contact angle decreases, therefore, the adhesive may readily displace air as it wets the composite surface. For a contaminated surface, adhesive wetting of the composite is limited and air remains at the adhesive-composite interface. The plasma treatment increases surface energy by removing surface contaminants left by abrasion and solvent cleaning and introducing hydrophilic groups to the matrix (Table 3). Abrasion is an important step in removing loosely bound fibres and matrix from the machining step, which, if not removed, would otherwise reduce the bond strength. 4.2. Bonded repair design Analysis of two traditional repair configurations and the associated experimental validation confirmed that neither bonded single-sided doublers or scarf repairs were capable of achieving significant strength recovery under hot/wet conditions. In the case of external doubler repairs, the contribution of secondary bending

caused a significant reduction in joint strength (Fig. 9). Typical scarf joints examined using both FM355 and FM32 adhesive also offered limited strength recovery for the hot/wet condition. Most notably, addition of an external doubler to the scarf joint enabled considerable improvement in joint strength and supported the simple model used to predict the joint strength for the hot/wet and hot/dry cases. Essentially, for a correctly designed doubler, the scarf bondline will fail first in the hot/wet condition, leading to an increase in the scarf joint strength that is proportional to the ratio of the doubler and parent stiffness. The scarf-doubler configuration would clearly be a labour intensive repair but would not impact on the aerodynamic performance or component balance and weight any more significantly than a straight forward external doubler repair. The novel repair solution would provide excellent strength recovery for highly-loaded, high temperature composite structure and obviate problems associated with the limitation of 177 °C curing adhesives. The 177 °C curing adhesive systems examined in this work exhibited relatively poor strength under the hot/wet operating environment that some BMI composite structure may experience, necessitating the novel scarfdoubler repair design solution. The scarf joint strength for FM355 adhesive (Fig. 11) was relatively close to that predicted by the simple model (Eq. (5)) for the hot/dry and hot/wet cases where failure occurred in the adhesive, with results for the different laminates being relatively similar. This may indicate that bond stresses predicted by linear FE analysis become more uniform with adhesive plasticization prior to failure. More significant deviations from predicted strength were observed at RTA and cold/dry, particularly for the quasi-isotropic laminate. The quasi-isotropic laminate was predicted to have the highest stress concentration along the bondline (Fig. 3) and, despite the limited amount of data may indicate bondline stress dependency on laminate stacking sequence would need to incorporated into predictive models with higher accuracy. In a similar way, under hot/dry and hot/wet conditions, the scarf joints bonded with FM32 (Fig. 12) had strengths relatively similar to those predicted by the model and significant deviation from the predicted strength at RTA and cold/dry. If the predicted strength was modified to account for the bondline stress concentration estimated from elastic FE analysis (Fig. 4) then a closer prediction could be achieved for the RTA and cold/dry samples. However, as the joint failed in the composite, the model would no longer be valid. Nevertheless, the fact that the use of a stress concentration factor could improve the RTA and cold/dry strength predictions, may indicate that the bondline stresses need to be accounted for in developing improved models for joint strength prediction over a range of operating conditions and temperatures. The bonded scarf-doubler joint strengths could be relatively accurately predicted with the simple model (Eq. (6)) for the hot/ wet test condition, confirming the validity of the model. The hot/ dry strength for the scarf-doubler joints was accurately predicted for the FM32 joints but the overprediction for the FM355 joints may indicate that the model accuracy depended on the adhesive properties. Further work would be needed to see how flexible the model was for a range of adhesives with different ductility at given testing temperatures. The FM32 bonded scarf-doubler joints showed significantly lower than predicted strength at RTA and cold/dry test temperatures. The lower strength was associated with first ply failure in the composite laminate and it was suspected that the doubler had failed prior to the scarf, preventing full scarf strength from being achieved. The RTA and cold/dry strength was still significantly higher than the hot/wet strength and, therefore, would not be the limiting design strength condition. However, it is important to improve the doubler design to increase the strength of the scarf-doubler joint at RTA and cold/dry and develop more accurate models to predict joint strength

911

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

1.0

σ Measured / σ Predicted

0.8

Scarf

0.6 Scarf-Doubler

0.4

0.2

0.0 500

700

900

1100 1300 1500 1700 1900 2100 2300 GII (J/m2)

Fig. 15. FM32 bonded joint strength versus GII for hot/wet (d) and hot/dry (N), RTA () and cold/dry (j) conditions. Open symbols represent scarf data and filled symbols represent scarf-doubler data. The broken and solid lines represent linear fits to the two sets of data.

for these conditions. The influence of doubler taper and scarf-tip geometry needs to be considered carefully in trying to maximise joint strength for carbon/BMI composites at RTA or CD temperature and will be the subject of a future publication. A method for predicting joint strength could potentially establish an empirical relationship between mode II fracture toughness and the measured strength. Fig. 15 shows that the measured strength normalised to the predicted strength, determined using Eqs. (5) and (6), show a direct correlation with GII. The two fitted lines to the experimental data show a difference in gradients for the scarf and scarf-doubler joints. The scarf joints at the cold/dry and RTA condition achieved around 60% of the predicted strength, whereas, the scarf-doubler joints at cold/dry and RTA only achieved around 45% of the predicted strength, which may have been due to premature doubler failure. The mode II fracture toughness failure samples also closely replicated those observed in the joint tests. At the hot/dry and hot/wet conditions, where the shear stresses are more uniform due to plastic yielding of the adhesive, the GII value is potentially representative for a range of laminates and would provide a good estimate of adhesive mode II properties. In the RTA and cold/dry cases, the joints tended to indicate failure at the first zero ply in the parent laminate and the GII values generated from the unidirectional laminate may also be representative of the fracture toughness of the joint by replicating the failure mode. Further work would require the comparison of the mode II fracture toughness of the adhesive and laminate separately to determine any difference with the joint configuration. Clearly a range of GII values would also need to be considered in conjunction with different stacking sequences to establish how applicable the GII value may be for estimating joint strength in the scarf and scarf-doubler cases. The scarf-doubler joints provide a good option for improving the joint strength of bonded repairs under hot/wet conditions and despite the lower-than-expected strength at RTA and cold/ dry, the hot/wet condition would still the provide the limiting design strength, at least at temperatures approaching 177 °C.

duce adhesive tack provided a continuous path for entrapped air to be removed under vacuum and offered a novel solution to reduce bondline voiding in vacuum cured adhesive joints. Plasma treatment also aided in removing entrapped air by improving adhesive wettability of the composite surface. For the first time a novel scarf-doubler joint design has been examined and demonstrated to provide the highest strength for the hot/wet test condition compared to traditional joint configurations currently in regular use. Single-sided doubler repairs and scarf joints provide limited repair capability at the hot/wet condition due to secondary bending effects and limited strength of the adhesives at temperatures approaching their wet Tg value. A simple model was used to predict the joint strength for scarf and scarf-doubler joints and was found to be relatively accurate for the hot/wet and hot/dry cases for two different adhesives. The ability to predict scarf and scarf-doubler joint strengths for the RTA and cold/dry condition would require the development of a more complex model that could account for composite failure modes. The scarf-doubler repair solution provides the ability to recover significant levels of strength in high temperature composite structure for in situ repairs, where adhesive options are limited to 177 °C curing systems. Acknowledgments The authors thank Mr. David Dellios, Mr. John Nugent, Mr. Aaron Charon and Mr. Jack Singhavong for contribution in fabrication and testing of the composite and metal joints. Appendix A. Derivation of the adhesive bond strength for a onesided doubler joint Consider the case of an adhesively bonded doubler, which is sufficiently long so that stress transfer occurs in a region close to the ends of the doubler. In this case the actual length of the doubler has no influence on the adhesive stresses and the hence the load carrying capacity of the bonded structure. Referring to Fig. 1a for a structure that is subjected to an applied stress rapplied , the peak adhesive shear stress in the adhesive near the cut-out is given by the following expression, when the adhesive remains elastic,

smax ¼

k rapplied ts 1þS

ðA1Þ

Assuming the yield stress of the adhesive is sy , plastic deformation will occur in the adhesive when the applied stress exceeds the following limit,

rp ¼ ð1 þ SÞ

sy

ðA2Þ

kts

Beyond the adhesive yielding point, the peak adhesive shear strain is given by [20]

"



rapplied cmax ¼ 1þ 2G rp sy

2 #

For an adhesive having a shear failure strain of cf , the bond strength of the joint is 5. Conclusions Screening of suitable adhesives for high temperature repairs indicated that the DMTA analysis of the wet Tg value was an accurate predictor of the adhesive strength. Methods to apply the adhesives using vacuum assisted cure techniques were developed in which volatile materials were removed or pre-reacted using vacuum staging. The placement of a permeable fabric layer at the adhesive-composite interface in combination with staging to re-

rmax

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2Gcf sy 2Gcf ¼ rp  1  ð1 þ SÞ 1 sy kt s sy

References [1] CycomÒ 5250-4 Prepreg System. Cytec engineered materials, technical data sheet, 17p. .

912

A.N. Rider et al. / Composites: Part A 41 (2010) 902–912

[2] CycomÒ 977-3 Toughened Epoxy System. Cytec engineered materials, technical data sheet, 5p (Rev. E). [last accessed 15.11.95]. [3] Pearce PJ, Arnott DR, Camilleri A, Kindermann MR, Mathys GI, Wilson AR. Cause and effect of void formation during vacuum bag curing of epoxy film adhesives. J Adhes Sci Technol 1998;12(6):567–84. [4] da Silva LFM, Adams RD, Gibbs M. Manufacture of adhesive joints and bulk specimens with high-temperature adhesives. Int J Adhes Adhes 2004;24:69–83. [5] MIL-HDBK-17-3F. Composite materials handbook: Supportability, vol. 3. Department of Defence [chapter 8]. [6] Baker AA, Chester RJ, Hugo GR, Radtke TC. Scarf repairs to highly strained graphite/epoxy structure. Int J Adhes Adhes 1999;19:161–71. [7] Oplinger DW. Mechanical fastening and adhesive bonding. In: Peters ST, editor. Handbook of composites. London: Chapman & Hall; 1998. p. 610–66. [8] Hart-Smith LJ. Advances in the analysis and design of adhesive bonded joints in composite aerospace structures. In: SAMPE process engineering series, vol. 19. Asuza: SAMPE; 1974, p. 722–37. [9] Hart-Smith LJ. Adhesively bonded joints in fibrous composite structures. Douglas aircraft paper 7740. In: The international symposium on joining and repair of fibre-reinforced plastics, Imperial College, London; 1986. [10] Wang CH, Rider AN, Chang P, Charon A, Baker AA. Structural repair techniques for highly-loaded carbon/BMI composites. In: Proceedings of SAMPE fall technical conference, SAMPE Publishing; 2007. ISBN 978-1-934551-01-1.

[11] da Silva LFM, Adams RD. Measurement of the mechanical properties of structural adhesives in tension and shear over a wide range of temperatures. J Adhes Sci Technol 2005;19:109–41. [12] da Silva LFM, Adams RD. Adhesive joints at high and low temperatures using similar and dissimilar adherends and dual adhesives. Int J Adhes Adhes 2007;27:216–26. [13] Rider AN, Wang CH, Chang P, Brack N. Surface treatment and application processes for epoxy bonding to bismaleimide composites. In: SAMPE fall technical conference, SAMPE Publishing; 2007. ISBN 978-1-934551-01-1. [14] ASTM D5656-04e1. Annual book of ASTM standards, vol. 15.06. Adhesives; 2004. [15] Rider AN, Chalkley P. The durability of an off-optimum cured aluminium joint. Int J Adhes Adhes 2004;24:95–106. [16] Davidson BD, Sun X. Geometry and data reduction recommendations for standardized end notched flexure test for unidirectional composites. J ASTM Int 2006;3(9) [paper id: JAI100285]. [17] Baker AA. Development of a hard-patch approach for scarf repair of composite structure. DSTO-TR-1892, DSTO, Melbourne; 2006. [18] Wang CH, Rose LRF. A crack bridging model for bonded plates subjected to tension and bending. Int J Solids Struct 1999;36(13):1985–2014. [19] Wang CH, Gunnion AJ. On the design methodology of scarf repairs to composite laminates. Compos Sci Technol 2008;68:35–46. [20] Duong C, Wang CH. Composite repair-theory and design. Elsevier Science; 2007.