Вы находитесь на странице: 1из 16



Paul R Spendley 1, Stephen L Ogin 1, Paul A Smith 1 and Andrew B Clarke 2

Faculty of Engineering and Physical Sciences, University of Surrey,

Guildford, GU2 7XH, United Kingdom
QinetiQ, Cody Technology Park, Ively Road, Farnborough,
Hampshire, GU14 0LX, United Kingdom

This paper presents an experimental study which examines the tensile and compressive response of a
quasi-isotropic carbon fibre reinforced polymer (CFRP) laminate subjected to ambient conditions and
environmental extremes associated with aircraft design. Specimens with and without holes are
considered within the context of structural features and associated design allowables. The data obtained,
relating to the effect of sample replicates and specimen geometry on the variability in failure strength,
support the concept of a reduced qualification test programme for CFRP.

Early aircraft adopted natural composites in the form of wood and bamboo to produce
efficient lightweight structures. These materials were often limited by their anisotropic
nature. With the development and availability of lightweight alloys during the 1930s,
the aircraft design engineer was able to consider multi-directional loading. Metallic
materials led this weight-dominated industry until the development of fibre reinforced
plastic or, more specifically, the commercialisation of CFRP in the 1980s. This material
offered enormous advantages over aluminium alloys, although these benefits were often
associated with the unidirectional strength of the CFRP rather then an appropriate multidirectional lay-up associated with an application there is a compromise to be met
between optimum unidirectional properties and an appropriate balanced lay-up. Over
the years, a number of standard coupon test methods and a process of determining a
material design allowable have been developed, based on methods for metallic
materials. The mechanical properties measured using these test methods often show
significant variability and as such lead to considerable conservatism or safety margin
during the design of a safety-critical component such as an aircraft. Ultimately, this
conservatism compromises the efficient use of these materials. This paper is concerned
with the generation of typical CFRP materials data for aerospace design and considers
the variability in CFRP coupon test results as a consequence of test configuration [1].
Generating design allowables based on test data alone can result in artificially low
design strengths, as mentioned above, although the conservative nature of airworthiness
certification means that this is an acceptable approach [2, 3].
In the period since the certification approach for composite structures was developed [46], the knowledge and understanding of composite laminate behaviour has increased
significantly (e.g. [7-9] with respect to notched compression) and hence advances in the
development of improved design allowables are timely. Currently, the variability and

lack of significant reference data on specific CFRP laminates often suggest it is

necessary to test a large number of specimens in order to determine the failure strength
with an acceptable degree of confidence. The effect of environmental extremes (based
on application) and structural features (e.g. fastener holes) on the laminate is addressed
by further testing or combining knock-down factors. Practical and economic constraints
can mean that only a comparatively small number of non-ambient tests are performed.
Consequently, the use of small sample methods and knock-down factors [10] for
calculating design strengths has been proposed as a 'quick' method, although this may
often result in over-conservative design allowables.
The present study examines the tensile and compressive response of notched and
unnotched quasi-isotropic CFRP laminate specimens under ambient and non-ambient
conditions. The test methods are based essentially on existing ASTM methods and the
effect of specimen number together with the influence of a circular notch on the
variability in failure strength are of particular interest here.
2.1 Material preparation
The pre-preg used in this study, high tensile strength PAN-based carbon fibres within a
toughened epoxy matrix, had an approximate mass of 140 g/m 2 and a cured ply
thickness of 0.125 mm. Sample quasi-isotropic composite plates with ply orientations
of (+45/0/-45/90)4s and measuring approximately 300 mm x 300 mm x 4 mm were
fabricated in a high quality environment. Tensile and compressive specimens with
nominal dimensions of 280 mm x 32 mm and 132 mm x 32 mm (recommended by
ASTM), respectively, were cut from the various plates. Specimens for drilling and for
subjecting to the various environmental conditioning regimes were selected from the
tension and compression sample sets in a randomised manner. Where notches were
required, holes of 6 mm diameter were drilled carefully in the centre of the specimen
using sickle-point drill bits. The hot wet (HW) samples were exposed to an
environment of 80C, 90% RH until saturation was achieved, typically after about 1000
hours. The complete test matrix is shown in Table 1.

2.2 Mechanical testing

All tests were performed using an Instron universal testing machine with suitably rated
hydraulic grips and a 250 kN load cell. Specimens were tabbed prior to testing, using
end tabs cut from a 45 glass fibre reinforced polymer laminate. The tests were based
on ASTM guidelines (see Table 1). A temperature cabinet was used to test the HW aged
samples at a nominal temperature of 80oC. Cold-dry specimens were pre-cooled to
-55oC prior to testing and a test temperature of -55oC was maintained by surrounding the
samples with solid carbon dioxide pellets.

Table 1: Test matrix showing the abbreviations used for the unnotched and notched
specimens tested at room temperature ambient, hot-wet and cold-dry conditions,
together with the number of specimens tested at each condition.
Test condition and number of specimens, n
Specimen type

(room temperature, ambient:
22C 2, 60% RH)

(Hot-wet: aged at 80C
and 90% RH, tested at

tested at

PT (Unnotched tension)




PC (Unnotched compression)




OHT (Notched tension






OHC (Notched compression)


(i) aASTM D3039/D3039M, bASTM D3410/D3410M 1995, cASTM D5766, dASTM
D6484/D6484M, e ASTM D5229/D5229M-92
(ii) Test specimen identification is constructed from specimen type and environmental
condition, e.g. Unnotched tensile (PT) tested under Cold Dry (CD) conditions becomes
2.3 Strain Measurement
For all specimens, the strain was measured using a longitudinal strain gauge and the
modulus was calculated using the data over the load range 0.1Pmax to 0.5Pmax, where Pmax
denotes the specimen failure load. For some specimens, both longitudinal and
transverse strain gauges were used and for some of the compression tests, strain gauges
were bonded to both faces of the specimens in order to verify that there was no
significant specimen bending. For the unnotched specimens, the single longitudinal
gauge was located at the centre of the coupon gauge length. For the notched specimens,
the strain gauge was located midway between the top of the hole and the grips (i.e.
position B in Figure 1). This proved satisfactory for the tension specimens because the
coupons were sufficiently long that the strain at this location essentially corresponded to
the far-field strain. For the shorter compression specimens, however, this appears not to
be so, because the results from the OHCRT coupons (the first group of notched samples
to be tested in compression) led to a much higher calculated value of the Youngs
modulus, suggesting that the strain was depressed as a result of the proximity to the
hole/grips. Subsequent finite element analysis confirmed that this was the case.
Consequently, for the other notched compression testing (OHCHW and OHCCD) the
strain gauge location was moved so that the gauge was located towards the mid-point of
the remaining ligament rather than above the mid-point of the hole (i.e. position A in
Figure 1).

Figure 1 Schematic diagram showing location of strain gauges for the open hole
compression tests (OHCRT, OHCHW, OHCCD). Dimension shown in parentheses
indicates location of gauge in the longer open hole tension test coupons (OHT,
OHTHW, OHTCD). Strain gauges for the unnotched specimens were positioned at the
centre of the specimen.
3.1 Mechanical response
The stress-strain data (up to 0.5 % strain) from all the experiments for the three test
conditions (Cold Dry, CD; Room Temperature, RT; Hot Wet, HW) are shown in Figures
2 to 13 and the resulting mechanical properties (modulus, strength) are summarised in
Table 2. From all the tests, there are perhaps two samples (one from the PCHW batch
and one from the OHTCD) that appear anomalous, showing particularly non-linear
stress-strain responses that result in low modulus values. It is possible that these
samples had poorly bonded gauges or slipped in the grips, although the failure strength
for the PCHW specimen was also low.
Overall the modulus data in Table 2 seem reasonable. Looking at the PT and OHT data
sets at the three conditions, we note that the average modulus increases slightly with
decreasing test temperature as would be expected. The compression data show a
broadly similar trend, when the OHCRT data, for which it is suggested that there was an
influence of the hole or the grips, are excluded (Included in Figure 11 are the data from
the single specimen with gauges present at locations A and B the increased strain at
location B is clear). In general the compression moduli are some 10 % lower than the
corresponding tension values. This is perhaps in part a reflection of a greater degree of
non-linearity in the compression stress-strain response over the relevant load range.

Figure 2 Graph of stress against strain for PTRT coupons.

Figure 3 Graph of stress against strain for PTHW coupons

Figure 4 Graph of stress against strain for PTCD coupons.

Figure 5 Graph of stress against strain for PCRT coupons.

Figure 6 Graph of stress against strain for PCHW coupons.

Figure 7 Graph of stress against strain for PCCD coupons.

Figure 8 Graph of stress against strain for OHTRT coupons.

Figure 9 Graph of stress against strain for OHTHW coupons.

Figure 10 Graph of stress against strain for OHTCD coupons.

Figure 11 Graph of stress against strain for OHCRT coupons (dashed curves indicate
examples strain measurements at position A and B).

Figure 12 Graph of stress against strain for OHCHW coupons.

Figure 13 Graph of stress against strain for OHCCD.


Table 2 Summary of moduli and strength values for the various specimen types tested.
Average results are shown one standard deviation. The effect of excluding the
questionable data points in the PCHW and OHTCD sample sets are indicated by the
figures in parenthesis.
Specimen type and

Mean modulus, GPa

one standard deviation
50.8 2.4
49.9 2.6
51.2 2.6
47.6 2.5
44.9 (46.3) 5.7 (3.3)
48.9 3.5
45.5 2.1
50.8 2.3
51.9 (52.7) 4.2 (2.9)
62.1 4.7*
44.1 2.9
45.2 2.4

Mean strength, MPa

one standard deviation
765 64
785 50
760 55
547 61
508 (518) 59 (49)
595 48
362 18
418 21
375 (373) 25 (25)
309 17
300 19
392 25

Proximity to the hole and/or the top grip appears to diminish strain readings, leading to
an unrealistically high value of Youngs modulus
The strength results are shown graphically in figure 14, in which the mean strengths and
standard deviations are shown for unnotched and notched tension and compression
specimens at each of the three test conditions. The number of specimens tested, n, is
also indicated for each type of test.
The basic trends of the strength data are much as would be expected. The (mean)
tensile strength of the unnotched coupons does not appear to change significantly with
test condition (i.e. PTCD, PTRT, PTHW), which is as expected for a fibre-dominated
property. The compression strength of the unnotched coupons is significantly lower
than the tensile strength, by about 30 % at the RT condition. Moreover the compression
strength decreases progressively from CD, through RT to HW. This is likely to be a
reflection of the progressive lowering of matrix properties with increasing temperature
and moisture content of the matrix (e.g. 11, 12) and an associated reduction in the
resistance of the composite to fibre micro-buckling. Residual stresses will also change
significantly over the test temperature range [13] and this might influence damage
development and strength.
Turning to the notched properties, these are substantially lower than the unnotched
properties, as would be expected for these notch-sensitive laminates. The strength
reductions under RT conditions of 40 50% are more than twice those that would be
expected simply on the basis of the reduction in the net-section cross-sectional area
(19%). The notched strength trends with test condition differ for the tension and
compression samples. The notched tensile strength shows an increase for the HW
condition compared to the RT condition, while the notched compression strength shows
a decrease. Degradation of the matrix and the fibre-matrix interface is likely to promote

matrix cracking/splitting damage mechanisms which act to reduce the stress

concentration at the notch edge and so delay fibre fracture in tension until higher levels
of applied laminate stress. Under compressive loading, however, the matrix and
interface degradation are likely to lead to fibre micro-buckling at lower levels of applied
stress than in the RT condition.

Figure 14 Histograms showing mean failure strengths (MPa) for each specimen type
against test condition (error bars represent one standard deviation).

3.2 Observations on strength variation

An alternative method of presenting the variability of the test data is to plot the
probability density function against failure strength for each sample type and test
condition (figure 15). This method of presenting data allows an appraisal of the scatter
associated with each data set. Figure 15 shows clearly that there is significantly reduced
scatter in the results for the notched data sets for all testing conditions. All of the
unnotched data sets have relatively large standard deviations and there is considerable
overlap between the data sets. The results show that for the notched specimens the
stress concentration reduces the scatter of the results, in addition to reducing the
strength of the specimens.


Figure 15 Test data plotted as probability density functions (x = strength value, =

mean and = standard deviation).

Figure 16 Test data sampled by plotting running average against the number of
specimens (n). (Mean failure strength values presented in parentheses)


Another method of representing the test results is to examine the running average of the
strengths. Figure 16 shows the running average for each type of test and provides an
indication of the number of specimens required for a reliable strength measurement.
Examining the unnotched tension results first, which of course give the highest strength
values, each condition requires about 12 or 13 specimens before the mean value
stabilises. The notched data sets, on the other hand, appear to reach a stable value after
approximately 5 to 7 strength values have been obtained. This suggests that the use of
small sample methods employing notched specimens may provide an economical
method of measuring strengths, providing the unnotched strengths can be derived from
the notched-strength values.
For design purposed, the B-basis value is often important. This is given by the simple
B k b

The parameter kb is the one sided tolerance factor for a 95 % confidence level. Figure
17 shows the ratio of the mean failure strength for each data set to the corresponding Bbasis value. This ratio gives another way of looking at the spread of the data, with
larger ratios indicating greater spread. The largest ratio is seen for the HW (unnotched)
compression test case and then, to a lesser extent, the two other unnotched compression
conditions. The notched compression tests show (on average) only slightly larger
strength ratios than the notched tension tests. Figure 18 shows the running B-basis
value as a function of sample number. It is clear that compared to Figure 16 (the
running average) data from a larger number of samples are needed before the B-basis
value can be identified, but that, as with Figure 16, convergence is quicker for the
notched data than the unnotched data.

Figure 17 Mean failure strength normalised by B-basis value for each data set.

Figure 18 Test data sampled by plotting B-basis value against the number of specimens
(n). (B-Basis values are shown in parentheses)

Strength measurements have been conducted on notched and unnotched specimens from
a CFRP quasi-isotropic laminate under ambient and non-ambient conditions. The
results show that notched specimens containing a centrally placed 6 mm diameter hole
exhibit significantly less variation in tensile and compressive failure strength than
unnotched specimens. Sampling the data with respect to specimen quantity suggests a
mean strength may be obtained with fewer tests if notched specimens are used, provided
an appropriate method is available to extract unnotched data from the results.

The authors would like to thank Mr Peter Haynes and Dr David Jesson of the University
of Surrey for their invaluable experimental support and advice. Further thanks go to Ms
Linda Clowes and Mr Kevin Denham of QinetiQ for assistance in producing high
quality samples. Finally, gratitude is extended for financial support provided by EPSRC
and QinetiQ (Farnborough).


1- Hart-Smith LJ, Generation of higher composite material allowables using
improved test coupons, 36th International SAMPE Symposium, 1991.
2- Federal Aviation Administration, Composite Aircraft Structure, Advisory
Circular AC 20-107A, 25 April 1984 and companion document by the JAA, ACJ
25.603, Composite Aircraft Structure Acceptable Means of Compliance, 1986.
3- European Aviation Safety Agency., On certification specifications, including
airworthiness codes and acceptable means of compliance, for large aeroplanes,
(CS-25) Decision No. 2003/2/RM, 2003.
4- Rouchon JR, Certification of large airplane composite structures, recent
progress and new trends in compliance philosophy, ICAS congress, 1990.
5- Whitehead RS., Kan HP., Cordero R. and Saether ES. Certification testing
methodology for composite structures, Vols 1 and 2, NADC Final Report,
Contract No. N62269-84-C-024,. 1986.
6- Bristow J., Structural composite airworthiness in civil aircraft. Proceedings of
6th SAMPE Conference, 1985; 39: 1-11.
7- Soutis C., Fleck NA. and Smith PA.,. Failure prediction technique for
compression loaded carbon fibre-epoxy laminate with open holes, Journal of
Composite Materials, 1991; 25:1476-1498.
8- Sutcliffe MPF., Xin XJ., Fleck NA.and Curtis PT., .Composite Compressive
Strength Modeller, Version 1.4a, 1999, Engineering Department, Cambridge
University, Trumpington St, Cambridge, CB2 1PZ.
9- Suemasu H., Takahashi H. and Ishikawa T., On failure mechanisms of
composite laminates with an open hole subjected to compressive load.
Composites Science and Technology, 2006; 66, 634-641.
10- Hallet SJ, Derivation of Design allowables at Airbus Filton site, 2nd
International Conference on Composites Testing and Model Identification, 2004.
11- Potter RT and Purslow D., The effect of environment on the compression
strength of notched CFRP - a fractographic investigation, Composites, 1983;14:
12- Kellas S, Morton J and Curtis PT, The effect of hygrothermal environments
upon the tensile and compressive strengths of notched CFRP laminates Part 1 :
Static loading. Composites, 1990; 21.
13- Sanchez-Saez S, Gomez-del Ro T., Barbero E., Zaera R. and Navarro C., Static
behavior of CFRPs at low temperatures, Composites: Part B, 2002; 33: 383