- Email: [email protected]
metal joints
Fatigue research on bonded carbon fibre composite/metal joints M. A. SMITH and R. HA R D Y
This paper describes the objectives and philosophy of research being conducted on the fatigue performance of carbon fibre composite (cfc) in structural form. The research programme on simple composite/metal joints utilising bonding is discussed and results are presented for fatigue tests on a 2%” bonded scarf joint under both constant amplitude loading and a simple form of variable amplitude loading. These are discussed in terms of failure modes, cumulative damage behaviour and the effects of testing frequency. This work forms part of the initial stages of a programme on the fatigue of composite joints and future areas of work are briefly outlined.
Interest in carbon fibre composite (cfc) as an aircraft structural material has been generated by the possibility of both weight and cost savings. Consequently a large amount of research is currently aimed at developing the composite into a useable aircraft material and applying it in specific aircraft components. Within this field, the role of the Fatigue Laboratory in the Structures Department is to identify forms of construction which may be sensitive to fatigue and to evaluate the behaviour of typical structural features. Joints are a particular area of interest and a review of work was conducted’ in 1974 which identified bonding and bolting as being the main methods of jointing under consideration. It was decided that initial studies would be on bonded joints and that work on bolted joints would follow.
DESCRIPTION
A scarf joint was chosen as an example of a good bonded joint which provides a gradual transfer of load between the adherends, thereby giving lower stress concentrations at the ends of the adhesive layer than an equivalent size lap joint. It also provides a smooth (aerodynamic) surface to the component which is important for aerospace useage. The angle of the scarf is 2’%” and this was chosen as a compromise between the need for as low an angle as possible and machining practicability. A diagram of the specimen is shown in Fig. 1 and a photograph in Fig. 2. 7
The intention was to test the joints under constant amplitude loading and other forms of loading more representative of service conditions. The resulting information on performance and failure mechanisms could then be useful in the development of life estimation methods and in aiding decisions on the forms of loading best suited to full-scale testing and the acquisition of design data. The amount of work completed to date on the bonded joint goes some way towards these objectives and in this paper information is presented on the cumulative damage behaviour of a particular form of bonded joint. Structures
Department,
RAE,
Farnborough,
OCTOBER
1977
Carbon
flbre
composite
Hants, England.
This paper was presented at the SEE Fatigue Group conference on ‘Fatigue of frp’ held on 29 June 1977 at City University, London.
COMPOSITES.
OF SPECIMENS
Fig. 1
Diagram of 23’2,”scarf joint between
aluminium
and cfc
255
the tensile mode. Thus the stresses applied to the unidirectional and the angle-ply joints differed slightly in proportion to their static strengths, shown in Table 2. It is known3 that significant temperature rises are produced by testing at above 10 Hz. This factor, as well as the frequency effect found in adhesive joints (to be described later), resulted in a compromise between speed of testing and the desirability of testing at representative structural frequencies. Thus all work, unless otherwise specified, was carried out at a frequency of 5 Hz. This work will be extended to include testing under realistic Bight-by-flight loading generated from an on-line computer. It is proposed to use ‘FALSTAFF’, the standard fighter loading spectrum4 for these tests.
Failure
modes
X Composlle/odhesrve 0 Alummrum cluddmg
.i_i_ IO5
I
IO6
IO'
Endurance Icycles)
TEST EQUIPMENT
Testing was performed both in-house and under contract at GEC, Whetstone. In both cases a 25 kN Instron electrohydraulic fatigue machine with hydraulic wedge grips was used. The narrow band random loading was generated by means of a sine-random generator and Burr-Brown filter.
Fig. 5 Fatigue results for unidirectional cfclaluminium joints under constant amplitude loading (mean stress MN/m’ (40% of static strength): frequency - 5 Hz)
2%” scarf 155.7
PRESENTA TION OF RESULTS
The constant amplitude S-N data for the joints between aluminium and unidirectional cfc is shown in Fig. 5, and between aluminium and angle-ply cfc in Fig. 6. Corresponding data from tests under narrow band random loading are shown in Figs 7 and 8 respectively. These last two graphs include curves showing life predicted from the constant amplitude data by Miner’s Rule. Detailed results are shown in Tables 3 and 4. FAILURE MODES AND CUMULATIVE BEHA VIOUR
DAMAGE
There were two main failure modes and these are shown in Figs 9 and 10. The failure shown in Fig. 9 is that of debonding followed by delamination. In this case the adhesive had failed by fatigue at one end of the joint length (the area of highest stress concentration) thus leading to progressive delamination and specimen failure. The mode of failure shown in Fig. 10 is different in that the fatigue failure had initiated in the aluminium side of the joint at the point marked leading almost instantaneously to failure by delamination of the composite. The failure in the aluminium originated from the outer, clad surface and then propagated across the section of the aluminium. Although this mode cannot be attributed to the cfc or the adhesive, it nevertheless is associated with the joint in that it results from the
Fig. 4 Narrow band random loading provides some of the features experienced by bonded joints in-service
256
Fig. 6 Fatigue results for angle-ply cfclaluminium 2%’ scarf joints under constant amplitude loading (cfc lay-up - 0” f 60”: mean stress 167.6 MN/m’ (40% of static strength): frequency - 5 Hz)
stress distribution within the joint. This type of failure occurred almost exclusively in the specimens with angle-ply composite and it seems significant that the failure point occurred at a point likely to be influenced by the centre unidirectional pre-preg layers - an obvious region of high stress. Figs 5 and 6 show a definite fatigue effect for the scarf joint with both the unidirectional and the angle-ply lay-ups. Generally the scatter appears to be comparable with metals. As described in the previous paragraph, two types of failure mode were observed, ie debonding-delamination or cladding initiated. For the unidirectional lay-up (Fig. 5) the failure mode was the debonding type with one exception, but for the angle-ply lay-up (Fig. 6) the failure depended on the stress level, the debonding mode predominating at high stresses and the cladding mode at low stresses. Thus the log mean curve drawn through the points in Fig. 6 is effectively a transition between the curves associated with the two failure modes - a situation analogous to the performance of plain specimens of clad aluminium alloy,’ where the cladding influences failure only at low stresses. When the curves for the unidirectional and angle-ply specimens are compared on the basis of alternating stresses on the same graph (Fig. 1 l), it is seen that the angle-ply specimens have the lower strength, contrary to the situation found in the static testing of these joints.
COMPOSITES.
OCTOBER
1977
Table 1. Specimen surface preparation prior to bonding Aluminium
alloy
IL 73)
Wash and wipe the surface with trichloroethane immerse
in a chrome
sulphuric
acid solution
for 30 minutes
at 60°-65’C The solution Fig. 2
Tests were carried out on bonded scarf joints
Throughout the programme the carbon tibre used was Type II (manufactured by Coutaulds) with the surface treatment appropriate at the time of the start of the work (1973). This fibre is in general use throughout the aircraft industry. However, some changes in the matrix resin system for the cfc have occurred. The programme started with Union Carbide ERLA 46 17 which was subsequently withdrawn from the market. While awaiting the results of various resin evaluation programmes the work continued in the established Shell DX210 system. As the DX210 resin possesses disadvantages when operating even at moderately elevated temperatures, it was decided to transfer to the Fothergill and Harvey Code 69 resin once this became widely accepted. In this programme it is considered that the fatigue strength is a function more of the adhesive than of the properties of the cfc matrix resin system and that results are likely to be comparable even with different matrix resins. Two lay-ups of cfc were used - unidirectional (0”) and angleply (0’ + 60”: containing 50% 0” fibre). The angle-ply lay-up was selected to be biased for greater strength in the axial direction (Fig. 3 shows the details). This material is a form considered to be generally representative at the time of its selection, although current thinking prefers 0” ? 45”.
is made up from:
Concentrated
sulphuric
acid (SG 1.84)
Chromium
trioxide
Cold water
- to make up to
crystals
Rinse with clean, cold running Dry at room temperature 50°C
600 ml 200 gm 4000
water for about
for about
15 minutes,
ml
15 minutes or at about
for 5 minutes.
Bond immediately.
Carbon
fibre composite
Abrade
with wet, fine silicon carbide
Wipe with acetone
paper
to remove all dust
Bond
Table 2. Static tensile strength results for aluminium/cfc 2%’ scarf joints Unidirectional specimen: Average static tensile strength
=
389.4 MN/m2
=
419.0
Angle-ply specimen: Average static tensile strength
MN/m*
The metal adherend was clad aluminium alloy to BS L73 a representative aircraft structural metal. The adhesive used throughout was AF 126/2 manufactured by The Minnesota Mining and Manufacturing Company. This is an epoxy-based adhesive with a nitrile modified epoxy filler curing at 120°C and was chosen after static evaluation tests at the beginning of the work. The surface pre-treatments for
the aluminium and the cfc are listed in Table 1. The specimen was assembled in a jig to ensure axial alignment of the two parts and cured in a heated platen press under a pressure of 0.345 MN/m’. Curing the joint at 120°C produced a residual thermal stress in the specimen after it had cooled - resulting in slight bowing. This stress was relieved after a few fatigue cycles and is assumed not to prejudice the results. In any case, it would be extremely difficult in this type of specimen to relieve the thermal stresses during the bonding phase.
0”
Plain rectangular end plates of aluminium alloy were bonded to each end of the specimen to enable it to be held in the wedge grip end fittings of the fatigue machine. FORMS
(3,6)
(2.7)
(1.4,5,8)
O”~60’ 8ply laminate Half
Fig. 3
0”
8
holf
f60”
Stacking sequence for angle-ply laminate
COMPOSITES.
OCTOBER
1977
OF LOADING
Two forms of loading were used for the testing - constant amplitude and narrow band random (Fig. 4). The constant amplitude loading provided basic stress-life (S-N) data on the fatigue performance of the specimen, for use in life prediction calculations, whilst the narrow band random loading provided a convenient form of variable amplitude loading containing some of the features experienced in service.’ The mean stress selected for the tests was 40% of the static strength of the joints, chosen to allow a useful range of alternating stress amplitudes whilst maintaining loading in
257
Referring to the narrow band random loading results in Figs 7 and 8 and the comparison of the log mean curves in Fig. 12, it is apparent that the unidirectional specimens were generally stronger than the angle-ply ones as was found for the constant amplitude loading. Also, there was some tendency in the angle-ply specimens for cladding modes of failure at the lower stresses but, as far as can be judged from the limited number of results, cladding failures do not predominate as they did for constant amplitude loading.
60
0
Alummwm
cloddlng
0 104
The second curves on Figs 7 and 8 are the predicted lives of the specimens, computed from the constant amplitude data using Miner’s Rule. In both cases it can be seen that the predicted life curves overestimate life at the lower stress levels and underestimate life at the higher stress levels. The cross-over of the predicted and achieved life curves occurs approximately at the point below which cladding failure modes appear in the angle-ply specimens. At the lower stress levels the value of 2 n/N is approximately 0.3 for both types of specimens and at the higher stress values the corresponding figure is about 3.0.
105
1
I
106
IO’
Endurance
(cycles)
Fig. 7 Fatigue results for unidirectional cfc/aluminium 2%’ scarf joints under narrow band random loading (mean stress - 155.7 MN/m’ : frequency - 5 Hz)
60,
I
Table 3. Fatigue test results for aluminiumhfc lay-up) 2%” scarf joint* Alternating stress
f
IO-
X
Composlfe
0
Alumlnlum
c
/adhesw
0 I04
-20
cloddlng
105 Endurance
/
I
IO6
IO’
A
108
=
(2 MN/m* 1
Constant
(cycles)
Fig. 8 Fatigue results for angle-ply cfc/aluminium under narrow band random loading (mean stress frequency 5 Hz)
Port
E aI f
Part
2%” scarf joints 167.6 MN/m’ :
B
+% of static strength
amplitude
(unidirectional
Endurance cycles x IO’
Failure mode
3.3 1.7 4.4 2.3 13.0 13.0 10.0 8.5 140.0 38.0 840.0 270.0
A
loading
137.7 137.7 137.7 137.7
35.4 35.4 35.4 35.4
110.1
28.3
110.1
28.3
110.1
28.3
110.1
28.3
82.6
21.2
82.6
21.2
55.1
14.1
55.1
14.1
A A A A A A A A
8 A A
Aluminium
Narrow Part
Port
band
random
46.7 46.7 46.7 38.9 38.9 31.1 31.1 23.4 23.4 15.6
A
B
Adhewe
Peaklrms
258
Debondingdelamination
failure mode
A -
15.0 19.0 69.0 50.0 53.0 230.0 55.0 56.0 960.0 2400.0
12 12 12 10
10 8 8 6 6 4
l Mean stress 155.7 frequency - 5 Hz
Fig. 9
loading
g
MN/m’
A A A A A A A A A A
(40% of static strength of joint):
3.5
Debondingdelamination
failure:
B - Aluminium
COMPOSITES.
cladding failure
OCTOBER
1977
There are several interesting l
failure in the metal can be precipitated by a stress concentration induced by the lay-up of the composite
l
for the angle-ply material the predominance of initial failure in the metal at low stresses of constant amplitude
Table 4. Fatigue test results for aluminium/cfc lay-up) 2X” joint* Alternating f (MN/m’
)
stress
Endurance
Failure mode
loading
118.5
28.3
0.94
A
118.5
28.3
1.7
A
104.7
25
8.5
A
104.7
25
14.0
A
104.7
25
5.3
A
104.7
25
3.8
A
104.7 88.9
25 21.2
9.8
B
88.9
21.2
16.0
A
83.8
20
42.0
83.8
20
29.0
A B
83.8
20
21.0
A
83.8
20
29.0
B
83.8
20
29.0
A
62.8 62.8
15 15
110.0 480.0
B B
62.8
15
65.0
B
98.0
B B
62.8
10.0
15
0
there is a pronounced difference in the cumulative damage at high and low stress levels
l
the cumulative damage characteristics of the unidirectional and angle-ply specimens are very similar despite differences in the modes of failure under constant amplitude loading.
(angle-ply
cycles x IO4
*‘A of static strength
Constant amplitude
is not found under random loading, presumably because high stresses occur occasionally at low rms stress levels
features to these results:
A
62.8
15
210.0
62.8
15
62.0
B
59.3
14.1
210.0
B
59.3
14.1
260.0
B
59.3
14.1
800.0
B
59.3
14.1
180.0
A
41.9
10
850.0
B
EFFECT OF TESTING FREQUENCY
Another interesting aspect of this work was the discovery, at an early stage of the programme, of a strong frequency effect for the aluminium to angle-ply cfc joints (the effect was not investigated for the unidirectional joints). Fig. 13 presents the data obtained from the investigation of the effect of testing frequency on fatigue endurance, and the detailed results are shown in Table 5. As can be seen, by increasing the test frequency from 0.5 Hz to 100 Hz, an increase in endurance of about 1000: 1 was obtained for the same conditions of fatigue loading. By contrast, it is usually found that metal specimens show an increase of about 2: 1 in endurance for a similar change in frequency. All the joint surfaces of the failed specimens were found to be of similar appearance indicating that the mode of failure was the same. This was the debonding-delamination type of failure which was mentioned earlier and is consistent with testing at the higher stress levels. These results warrant special attention, particularly in relation to accelerated fatigue testing. Ironically, the testing time for these tests remained approximately the same when the frequency was increased, suggesting that there may be a creep effect contributing to the failure process, although a sustained loading of 251.4 MN/m*
Part
A
Port B
Narrow band random loading 41.9
IO
23.0
A
41.9
10
31.0
A
41.9
10
34.0
A
41.9
10
31.0
A
33.5
8
52.0
A
33.5
8
76.0
A
33.5
8
82.0
B
33.5
8
90.0
B
25.1 25.1
6
110.0
A
6 4
390.0
B
16.8
1500.0
A
16.8 16.8
4 4
* Mean stress - 167.6 frequency - 5 Hz Peak/rms A -
ratio
e
MN/m’
runout
1200.0
runout
layer
Fatigue surface of olumlnlum
‘art B
‘art
A
(40% of static strength of joint):
3.5
Debonding-delamination
COMPOSITES.
2000.0
CladdIng
OCTOBER
failure:
1977
I3 -
Aluminium
cladding failure
Fig. 10
Cladding-initiated
failure mode
259
(60% of static strength)
at room temperature did not produce a failure. It is planned to extend the investigation to metalto-metal and cfc-to-cfc bonded scarf joints to determine the extent to which the observed effect was associated with the composite or the adhesive.
60
-___-
1
r
All the results discussed above are the first stage in the programme investigating the fatigue behaviour of bonded scarf joints and further results will be presented when available. GENERAL
REMARKS
REGARDING
BONDING
Mean
stress
40%
:
Frequency
There are some general points to be made regarding bonded joints. Firstly, the surface preparation of the adherends before bonding is a process requiring meticulous care and consistent quality is needed if the resulting bond is to be
_-.-/A
108 Endurance
(Hz) 100
240.0
100
31.0
100
34.0
100 50
37.0
50
45.0
50
22.0
50
26.0
5
29.0
5 5
29.0 42.0
5
29.0
5 0.5
21.0 0.037
0.5
0.064
0.5
0.11
0.5
0.042
167.6 MN/m’ (40% * 83.8 MN/m2
E ;; ao: =. :60: 40%
of stottc
IO6
strong enough to achieve the stresses quoted in this report, because there is at present no satisfactory ndt method of measuring the strength of a bond after it has been manufactured. Secondly, when the adherends are of dissimilar material, current high strength adhesives, curing at elevated temperatures, produce a residual thermal stress in the component upon cooling. Thirdly, there are practical problems of applying a suitable bonding pressure to components of complex shape. Despite these problems, bonded joints can be envisaged as having applications in areas where there is a single, well-defined stress direction and hence a unidirectional lay-up can be used, which produces a strong bond. Also the weight of a bonded joint is likely to be less than that of a bolted one of comparable strength. In addition, there are no stress concentrations arising from drilled holes as in the case of mechanical joints: tibre layers are not penetrated and there is a better load transfer path.
strength
20’ IO5
IO6 Endurance
11 Comparison scarf joints
260
105 (cycles)
5 Hz
Frequency
Fig.
104
Fig. 13 Fatigue results showing a frequency effect for angle-ply cfc/aluminium 2X0 scarf joints under constant amplitude loading; (mean stress - 167.6 MN/m’ (40% of static strength): alternating stress - 83.8 MN/m’ (mean to peak 20% of static strength))
IOO-
stress
IO3
Endurance
6
IO4
/
102
t P
Mean
loading results for 2%’
100
oli
2120.
z
of narrow band random
28.0
l Loading - constant amplitude: Mean stress of static strength of joint); Alternating stress (5 20% of static strength of joint)
z 5 40.
(cycles)
Endurance cycles x IO4
Frequency
2 :
strength
_
Fig. 12 Comparison scarf joints
Table 5. Effect of testing frequency on fatigue endurance of an aluminium/cfc 2%’ scarf joint with angle-ply lay-up of 0” + 60”*
of stotlc
5Hz
of constant
I
I
IO’
IO8
REMARKS
The work described in this paper, on the fatigue behaviour
(cycles)
amplitude
CONCLUDING
loading results for 2%”
of bonded scarf joints between clad aluminium and unidirectional or angle-ply cfc, forms the initial phase of a
COMPOSITES.
OCTOBER
1977
programme on the fatigue of bonded joints. Information has been obtained on modes of fatigue failure, cumulative damage behaviour and the effect of testing frequency and can be summarised as follows: bonded joints between aluminium and unidirectional cfc are slightly stronger in fatigue than those between aluminium and angle-ply cfc the general mode of failure is a progressive debonding and delamination. However, in angle-ply specimens this process is aggravated by damage in the metal, initiating from the clad surface Miner’s Rule underestimates life at higher stresses but overestimates at lower stresses for both types of specimen under narrow band random loading there is a marked effect of testing frequency life for this type of bonded joint.
on fatigue
The work on bonded scarf joints will be extended to include testing under a representative flight-by-flight loading. Also the effect of bonding cfc to unclad aluminium, titanium and steel is being investigated under various forms of loading. In parallel with the work on bonded joints, work has started on mechanical joints which will clearly form an important part
COMPOSITES.
OCTOBER
1977
of future cfc aircraft structure. Problems such as the effect of environment on fatigue life and the effect of compressive loadings will need to be considered in the future. ACKNOWLEDGEMENT
This article is published by permission of the Controller, Majesty’s Stationery Office, Crown Copyright reserved.
Her
REFERENCES Sage, G.N. and Smith, M.A. ‘A survey of metal-cfrp joints with reference to fatigue problems’, RAE Technical Memorandum, Struciures 852 (1974) Kirkby, W.T. and Edwards, P.R ‘A method of fatigue life prediction using data obtained under random loading conditions’, RAE Technical Report 66023 (1966)
Sturgeon, J.B. ‘Tensile testing of f 45” orientations to determine the shear response of carbon fibre reinforced plastics’, RAE Technical Report 76056 (1976)
Lowak, H., Schutz, D. Huch, M., and Schtttz, W. ‘Standardisiertes eingelflug programm fur Kampfflugzenge - FALSTAFF - LBF Nr 3045’ IABG No 568 (1976) Edwards, P.R. and Earl, M.G. ‘A comparative study of the fatigue performance of notched specimens of clad and unclad aluminium ahoy, with and without a pre-stress’, ARC CP No 1361 (1977)
261
This paper describes the objectives and philosophy of research being conducted on the fatigue performance of carbon fibre composite (cfc) in structural form. The research programme on simple composite/metal joints utilising bonding is discussed and results are presented for fatigue tests on a 2%” bonded scarf joint under both constant amplitude loading and a simple form of variable amplitude loading. These are discussed in terms of failure modes, cumulative damage behaviour and the effects of testing frequency. This work forms part of the initial stages of a programme on the fatigue of composite joints and future areas of work are briefly outlined.
Interest in carbon fibre composite (cfc) as an aircraft structural material has been generated by the possibility of both weight and cost savings. Consequently a large amount of research is currently aimed at developing the composite into a useable aircraft material and applying it in specific aircraft components. Within this field, the role of the Fatigue Laboratory in the Structures Department is to identify forms of construction which may be sensitive to fatigue and to evaluate the behaviour of typical structural features. Joints are a particular area of interest and a review of work was conducted’ in 1974 which identified bonding and bolting as being the main methods of jointing under consideration. It was decided that initial studies would be on bonded joints and that work on bolted joints would follow.
DESCRIPTION
A scarf joint was chosen as an example of a good bonded joint which provides a gradual transfer of load between the adherends, thereby giving lower stress concentrations at the ends of the adhesive layer than an equivalent size lap joint. It also provides a smooth (aerodynamic) surface to the component which is important for aerospace useage. The angle of the scarf is 2’%” and this was chosen as a compromise between the need for as low an angle as possible and machining practicability. A diagram of the specimen is shown in Fig. 1 and a photograph in Fig. 2. 7
The intention was to test the joints under constant amplitude loading and other forms of loading more representative of service conditions. The resulting information on performance and failure mechanisms could then be useful in the development of life estimation methods and in aiding decisions on the forms of loading best suited to full-scale testing and the acquisition of design data. The amount of work completed to date on the bonded joint goes some way towards these objectives and in this paper information is presented on the cumulative damage behaviour of a particular form of bonded joint. Structures
Department,
RAE,
Farnborough,
OCTOBER
1977
Carbon
flbre
composite
Hants, England.
This paper was presented at the SEE Fatigue Group conference on ‘Fatigue of frp’ held on 29 June 1977 at City University, London.
COMPOSITES.
OF SPECIMENS
Fig. 1
Diagram of 23’2,”scarf joint between
aluminium
and cfc
255
the tensile mode. Thus the stresses applied to the unidirectional and the angle-ply joints differed slightly in proportion to their static strengths, shown in Table 2. It is known3 that significant temperature rises are produced by testing at above 10 Hz. This factor, as well as the frequency effect found in adhesive joints (to be described later), resulted in a compromise between speed of testing and the desirability of testing at representative structural frequencies. Thus all work, unless otherwise specified, was carried out at a frequency of 5 Hz. This work will be extended to include testing under realistic Bight-by-flight loading generated from an on-line computer. It is proposed to use ‘FALSTAFF’, the standard fighter loading spectrum4 for these tests.
Failure
modes
X Composlle/odhesrve 0 Alummrum cluddmg
.i_i_ IO5
I
IO6
IO'
Endurance Icycles)
TEST EQUIPMENT
Testing was performed both in-house and under contract at GEC, Whetstone. In both cases a 25 kN Instron electrohydraulic fatigue machine with hydraulic wedge grips was used. The narrow band random loading was generated by means of a sine-random generator and Burr-Brown filter.
Fig. 5 Fatigue results for unidirectional cfclaluminium joints under constant amplitude loading (mean stress MN/m’ (40% of static strength): frequency - 5 Hz)
2%” scarf 155.7
PRESENTA TION OF RESULTS
The constant amplitude S-N data for the joints between aluminium and unidirectional cfc is shown in Fig. 5, and between aluminium and angle-ply cfc in Fig. 6. Corresponding data from tests under narrow band random loading are shown in Figs 7 and 8 respectively. These last two graphs include curves showing life predicted from the constant amplitude data by Miner’s Rule. Detailed results are shown in Tables 3 and 4. FAILURE MODES AND CUMULATIVE BEHA VIOUR
DAMAGE
There were two main failure modes and these are shown in Figs 9 and 10. The failure shown in Fig. 9 is that of debonding followed by delamination. In this case the adhesive had failed by fatigue at one end of the joint length (the area of highest stress concentration) thus leading to progressive delamination and specimen failure. The mode of failure shown in Fig. 10 is different in that the fatigue failure had initiated in the aluminium side of the joint at the point marked leading almost instantaneously to failure by delamination of the composite. The failure in the aluminium originated from the outer, clad surface and then propagated across the section of the aluminium. Although this mode cannot be attributed to the cfc or the adhesive, it nevertheless is associated with the joint in that it results from the
Fig. 4 Narrow band random loading provides some of the features experienced by bonded joints in-service
256
Fig. 6 Fatigue results for angle-ply cfclaluminium 2%’ scarf joints under constant amplitude loading (cfc lay-up - 0” f 60”: mean stress 167.6 MN/m’ (40% of static strength): frequency - 5 Hz)
stress distribution within the joint. This type of failure occurred almost exclusively in the specimens with angle-ply composite and it seems significant that the failure point occurred at a point likely to be influenced by the centre unidirectional pre-preg layers - an obvious region of high stress. Figs 5 and 6 show a definite fatigue effect for the scarf joint with both the unidirectional and the angle-ply lay-ups. Generally the scatter appears to be comparable with metals. As described in the previous paragraph, two types of failure mode were observed, ie debonding-delamination or cladding initiated. For the unidirectional lay-up (Fig. 5) the failure mode was the debonding type with one exception, but for the angle-ply lay-up (Fig. 6) the failure depended on the stress level, the debonding mode predominating at high stresses and the cladding mode at low stresses. Thus the log mean curve drawn through the points in Fig. 6 is effectively a transition between the curves associated with the two failure modes - a situation analogous to the performance of plain specimens of clad aluminium alloy,’ where the cladding influences failure only at low stresses. When the curves for the unidirectional and angle-ply specimens are compared on the basis of alternating stresses on the same graph (Fig. 1 l), it is seen that the angle-ply specimens have the lower strength, contrary to the situation found in the static testing of these joints.
COMPOSITES.
OCTOBER
1977
Table 1. Specimen surface preparation prior to bonding Aluminium
alloy
IL 73)
Wash and wipe the surface with trichloroethane immerse
in a chrome
sulphuric
acid solution
for 30 minutes
at 60°-65’C The solution Fig. 2
Tests were carried out on bonded scarf joints
Throughout the programme the carbon tibre used was Type II (manufactured by Coutaulds) with the surface treatment appropriate at the time of the start of the work (1973). This fibre is in general use throughout the aircraft industry. However, some changes in the matrix resin system for the cfc have occurred. The programme started with Union Carbide ERLA 46 17 which was subsequently withdrawn from the market. While awaiting the results of various resin evaluation programmes the work continued in the established Shell DX210 system. As the DX210 resin possesses disadvantages when operating even at moderately elevated temperatures, it was decided to transfer to the Fothergill and Harvey Code 69 resin once this became widely accepted. In this programme it is considered that the fatigue strength is a function more of the adhesive than of the properties of the cfc matrix resin system and that results are likely to be comparable even with different matrix resins. Two lay-ups of cfc were used - unidirectional (0”) and angleply (0’ + 60”: containing 50% 0” fibre). The angle-ply lay-up was selected to be biased for greater strength in the axial direction (Fig. 3 shows the details). This material is a form considered to be generally representative at the time of its selection, although current thinking prefers 0” ? 45”.
is made up from:
Concentrated
sulphuric
acid (SG 1.84)
Chromium
trioxide
Cold water
- to make up to
crystals
Rinse with clean, cold running Dry at room temperature 50°C
600 ml 200 gm 4000
water for about
for about
15 minutes,
ml
15 minutes or at about
for 5 minutes.
Bond immediately.
Carbon
fibre composite
Abrade
with wet, fine silicon carbide
Wipe with acetone
paper
to remove all dust
Bond
Table 2. Static tensile strength results for aluminium/cfc 2%’ scarf joints Unidirectional specimen: Average static tensile strength
=
389.4 MN/m2
=
419.0
Angle-ply specimen: Average static tensile strength
MN/m*
The metal adherend was clad aluminium alloy to BS L73 a representative aircraft structural metal. The adhesive used throughout was AF 126/2 manufactured by The Minnesota Mining and Manufacturing Company. This is an epoxy-based adhesive with a nitrile modified epoxy filler curing at 120°C and was chosen after static evaluation tests at the beginning of the work. The surface pre-treatments for
the aluminium and the cfc are listed in Table 1. The specimen was assembled in a jig to ensure axial alignment of the two parts and cured in a heated platen press under a pressure of 0.345 MN/m’. Curing the joint at 120°C produced a residual thermal stress in the specimen after it had cooled - resulting in slight bowing. This stress was relieved after a few fatigue cycles and is assumed not to prejudice the results. In any case, it would be extremely difficult in this type of specimen to relieve the thermal stresses during the bonding phase.
0”
Plain rectangular end plates of aluminium alloy were bonded to each end of the specimen to enable it to be held in the wedge grip end fittings of the fatigue machine. FORMS
(3,6)
(2.7)
(1.4,5,8)
O”~60’ 8ply laminate Half
Fig. 3
0”
8
holf
f60”
Stacking sequence for angle-ply laminate
COMPOSITES.
OCTOBER
1977
OF LOADING
Two forms of loading were used for the testing - constant amplitude and narrow band random (Fig. 4). The constant amplitude loading provided basic stress-life (S-N) data on the fatigue performance of the specimen, for use in life prediction calculations, whilst the narrow band random loading provided a convenient form of variable amplitude loading containing some of the features experienced in service.’ The mean stress selected for the tests was 40% of the static strength of the joints, chosen to allow a useful range of alternating stress amplitudes whilst maintaining loading in
257
Referring to the narrow band random loading results in Figs 7 and 8 and the comparison of the log mean curves in Fig. 12, it is apparent that the unidirectional specimens were generally stronger than the angle-ply ones as was found for the constant amplitude loading. Also, there was some tendency in the angle-ply specimens for cladding modes of failure at the lower stresses but, as far as can be judged from the limited number of results, cladding failures do not predominate as they did for constant amplitude loading.
60
0
Alummwm
cloddlng
0 104
The second curves on Figs 7 and 8 are the predicted lives of the specimens, computed from the constant amplitude data using Miner’s Rule. In both cases it can be seen that the predicted life curves overestimate life at the lower stress levels and underestimate life at the higher stress levels. The cross-over of the predicted and achieved life curves occurs approximately at the point below which cladding failure modes appear in the angle-ply specimens. At the lower stress levels the value of 2 n/N is approximately 0.3 for both types of specimens and at the higher stress values the corresponding figure is about 3.0.
105
1
I
106
IO’
Endurance
(cycles)
Fig. 7 Fatigue results for unidirectional cfc/aluminium 2%’ scarf joints under narrow band random loading (mean stress - 155.7 MN/m’ : frequency - 5 Hz)
60,
I
Table 3. Fatigue test results for aluminiumhfc lay-up) 2%” scarf joint* Alternating stress
f
IO-
X
Composlfe
0
Alumlnlum
c
/adhesw
0 I04
-20
cloddlng
105 Endurance
/
I
IO6
IO’
A
108
=
(2 MN/m* 1
Constant
(cycles)
Fig. 8 Fatigue results for angle-ply cfc/aluminium under narrow band random loading (mean stress frequency 5 Hz)
Port
E aI f
Part
2%” scarf joints 167.6 MN/m’ :
B
+% of static strength
amplitude
(unidirectional
Endurance cycles x IO’
Failure mode
3.3 1.7 4.4 2.3 13.0 13.0 10.0 8.5 140.0 38.0 840.0 270.0
A
loading
137.7 137.7 137.7 137.7
35.4 35.4 35.4 35.4
110.1
28.3
110.1
28.3
110.1
28.3
110.1
28.3
82.6
21.2
82.6
21.2
55.1
14.1
55.1
14.1
A A A A A A A A
8 A A
Aluminium
Narrow Part
Port
band
random
46.7 46.7 46.7 38.9 38.9 31.1 31.1 23.4 23.4 15.6
A
B
Adhewe
Peaklrms
258
Debondingdelamination
failure mode
A -
15.0 19.0 69.0 50.0 53.0 230.0 55.0 56.0 960.0 2400.0
12 12 12 10
10 8 8 6 6 4
l Mean stress 155.7 frequency - 5 Hz
Fig. 9
loading
g
MN/m’
A A A A A A A A A A
(40% of static strength of joint):
3.5
Debondingdelamination
failure:
B - Aluminium
COMPOSITES.
cladding failure
OCTOBER
1977
There are several interesting l
failure in the metal can be precipitated by a stress concentration induced by the lay-up of the composite
l
for the angle-ply material the predominance of initial failure in the metal at low stresses of constant amplitude
Table 4. Fatigue test results for aluminium/cfc lay-up) 2X” joint* Alternating f (MN/m’
)
stress
Endurance
Failure mode
loading
118.5
28.3
0.94
A
118.5
28.3
1.7
A
104.7
25
8.5
A
104.7
25
14.0
A
104.7
25
5.3
A
104.7
25
3.8
A
104.7 88.9
25 21.2
9.8
B
88.9
21.2
16.0
A
83.8
20
42.0
83.8
20
29.0
A B
83.8
20
21.0
A
83.8
20
29.0
B
83.8
20
29.0
A
62.8 62.8
15 15
110.0 480.0
B B
62.8
15
65.0
B
98.0
B B
62.8
10.0
15
0
there is a pronounced difference in the cumulative damage at high and low stress levels
l
the cumulative damage characteristics of the unidirectional and angle-ply specimens are very similar despite differences in the modes of failure under constant amplitude loading.
(angle-ply
cycles x IO4
*‘A of static strength
Constant amplitude
is not found under random loading, presumably because high stresses occur occasionally at low rms stress levels
features to these results:
A
62.8
15
210.0
62.8
15
62.0
B
59.3
14.1
210.0
B
59.3
14.1
260.0
B
59.3
14.1
800.0
B
59.3
14.1
180.0
A
41.9
10
850.0
B
EFFECT OF TESTING FREQUENCY
Another interesting aspect of this work was the discovery, at an early stage of the programme, of a strong frequency effect for the aluminium to angle-ply cfc joints (the effect was not investigated for the unidirectional joints). Fig. 13 presents the data obtained from the investigation of the effect of testing frequency on fatigue endurance, and the detailed results are shown in Table 5. As can be seen, by increasing the test frequency from 0.5 Hz to 100 Hz, an increase in endurance of about 1000: 1 was obtained for the same conditions of fatigue loading. By contrast, it is usually found that metal specimens show an increase of about 2: 1 in endurance for a similar change in frequency. All the joint surfaces of the failed specimens were found to be of similar appearance indicating that the mode of failure was the same. This was the debonding-delamination type of failure which was mentioned earlier and is consistent with testing at the higher stress levels. These results warrant special attention, particularly in relation to accelerated fatigue testing. Ironically, the testing time for these tests remained approximately the same when the frequency was increased, suggesting that there may be a creep effect contributing to the failure process, although a sustained loading of 251.4 MN/m*
Part
A
Port B
Narrow band random loading 41.9
IO
23.0
A
41.9
10
31.0
A
41.9
10
34.0
A
41.9
10
31.0
A
33.5
8
52.0
A
33.5
8
76.0
A
33.5
8
82.0
B
33.5
8
90.0
B
25.1 25.1
6
110.0
A
6 4
390.0
B
16.8
1500.0
A
16.8 16.8
4 4
* Mean stress - 167.6 frequency - 5 Hz Peak/rms A -
ratio
e
MN/m’
runout
1200.0
runout
layer
Fatigue surface of olumlnlum
‘art B
‘art
A
(40% of static strength of joint):
3.5
Debonding-delamination
COMPOSITES.
2000.0
CladdIng
OCTOBER
failure:
1977
I3 -
Aluminium
cladding failure
Fig. 10
Cladding-initiated
failure mode
259
(60% of static strength)
at room temperature did not produce a failure. It is planned to extend the investigation to metalto-metal and cfc-to-cfc bonded scarf joints to determine the extent to which the observed effect was associated with the composite or the adhesive.
60
-___-
1
r
All the results discussed above are the first stage in the programme investigating the fatigue behaviour of bonded scarf joints and further results will be presented when available. GENERAL
REMARKS
REGARDING
BONDING
Mean
stress
40%
:
Frequency
There are some general points to be made regarding bonded joints. Firstly, the surface preparation of the adherends before bonding is a process requiring meticulous care and consistent quality is needed if the resulting bond is to be
_-.-/A
108 Endurance
(Hz) 100
240.0
100
31.0
100
34.0
100 50
37.0
50
45.0
50
22.0
50
26.0
5
29.0
5 5
29.0 42.0
5
29.0
5 0.5
21.0 0.037
0.5
0.064
0.5
0.11
0.5
0.042
167.6 MN/m’ (40% * 83.8 MN/m2
E ;; ao: =. :60: 40%
of stottc
IO6
strong enough to achieve the stresses quoted in this report, because there is at present no satisfactory ndt method of measuring the strength of a bond after it has been manufactured. Secondly, when the adherends are of dissimilar material, current high strength adhesives, curing at elevated temperatures, produce a residual thermal stress in the component upon cooling. Thirdly, there are practical problems of applying a suitable bonding pressure to components of complex shape. Despite these problems, bonded joints can be envisaged as having applications in areas where there is a single, well-defined stress direction and hence a unidirectional lay-up can be used, which produces a strong bond. Also the weight of a bonded joint is likely to be less than that of a bolted one of comparable strength. In addition, there are no stress concentrations arising from drilled holes as in the case of mechanical joints: tibre layers are not penetrated and there is a better load transfer path.
strength
20’ IO5
IO6 Endurance
11 Comparison scarf joints
260
105 (cycles)
5 Hz
Frequency
Fig.
104
Fig. 13 Fatigue results showing a frequency effect for angle-ply cfc/aluminium 2X0 scarf joints under constant amplitude loading; (mean stress - 167.6 MN/m’ (40% of static strength): alternating stress - 83.8 MN/m’ (mean to peak 20% of static strength))
IOO-
stress
IO3
Endurance
6
IO4
/
102
t P
Mean
loading results for 2%’
100
oli
2120.
z
of narrow band random
28.0
l Loading - constant amplitude: Mean stress of static strength of joint); Alternating stress (5 20% of static strength of joint)
z 5 40.
(cycles)
Endurance cycles x IO4
Frequency
2 :
strength
_
Fig. 12 Comparison scarf joints
Table 5. Effect of testing frequency on fatigue endurance of an aluminium/cfc 2%’ scarf joint with angle-ply lay-up of 0” + 60”*
of stotlc
5Hz
of constant
I
I
IO’
IO8
REMARKS
The work described in this paper, on the fatigue behaviour
(cycles)
amplitude
CONCLUDING
loading results for 2%”
of bonded scarf joints between clad aluminium and unidirectional or angle-ply cfc, forms the initial phase of a
COMPOSITES.
OCTOBER
1977
programme on the fatigue of bonded joints. Information has been obtained on modes of fatigue failure, cumulative damage behaviour and the effect of testing frequency and can be summarised as follows: bonded joints between aluminium and unidirectional cfc are slightly stronger in fatigue than those between aluminium and angle-ply cfc the general mode of failure is a progressive debonding and delamination. However, in angle-ply specimens this process is aggravated by damage in the metal, initiating from the clad surface Miner’s Rule underestimates life at higher stresses but overestimates at lower stresses for both types of specimen under narrow band random loading there is a marked effect of testing frequency life for this type of bonded joint.
on fatigue
The work on bonded scarf joints will be extended to include testing under a representative flight-by-flight loading. Also the effect of bonding cfc to unclad aluminium, titanium and steel is being investigated under various forms of loading. In parallel with the work on bonded joints, work has started on mechanical joints which will clearly form an important part
COMPOSITES.
OCTOBER
1977
of future cfc aircraft structure. Problems such as the effect of environment on fatigue life and the effect of compressive loadings will need to be considered in the future. ACKNOWLEDGEMENT
This article is published by permission of the Controller, Majesty’s Stationery Office, Crown Copyright reserved.
Her
REFERENCES Sage, G.N. and Smith, M.A. ‘A survey of metal-cfrp joints with reference to fatigue problems’, RAE Technical Memorandum, Struciures 852 (1974) Kirkby, W.T. and Edwards, P.R ‘A method of fatigue life prediction using data obtained under random loading conditions’, RAE Technical Report 66023 (1966)
Sturgeon, J.B. ‘Tensile testing of f 45” orientations to determine the shear response of carbon fibre reinforced plastics’, RAE Technical Report 76056 (1976)
Lowak, H., Schutz, D. Huch, M., and Schtttz, W. ‘Standardisiertes eingelflug programm fur Kampfflugzenge - FALSTAFF - LBF Nr 3045’ IABG No 568 (1976) Edwards, P.R. and Earl, M.G. ‘A comparative study of the fatigue performance of notched specimens of clad and unclad aluminium ahoy, with and without a pre-stress’, ARC CP No 1361 (1977)
261