DESIGN ALLOWABLES FOR NOTCHED AND UNNOTCHED CFRP IN TENSION AND COMPRESSION UNDER DIFFERING AMBIENT

CONDITIONS.

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

1 Faculty of Engineering and Physical Sciences, University of Surrey,Guildford, GU2 7XH, United Kingdom

2 QinetiQ, Cody Technology Park, Ively Road, Farnborough,Hampshire, GU14 0LX, United Kingdom

ABSTRACTThis 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.

1. INTRODUCTION

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 multi-directional 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 [4-6], 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

1

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. EXPERIMENTAL METHODS

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/m2 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 80ºC, 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.

2

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.

Specimen type

Test condition and number of specimens, n

RT(room temperature, ambient:22ºC ±2º, 60% RH)

HW(Hot-wet: aged at 80ºC and 90% RH, tested at 80ºC)e

CD(Cold-dry: tested at-55ºC)

PT (Unnotched tension) 21a 14 14

PC (Unnotched compression) 33b 12 12

OHT (Notched tension 17c 15 15

OHC (Notched compression) 33d 14 15

Notes: (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 PTCD.

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 Young’s 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).

3

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. RESULTS & DISCUSSION3.1 Mechanical responseThe 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.

4

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 1000 2000 3000 4000 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 2 Graph of stress against strain for PTRT coupons.

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 1000 2000 3000 4000 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 3 Graph of stress against strain for PTHW coupons

5

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 1000 2000 3000 4000 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 4 Graph of stress against strain for PTCD coupons.Applied stress versus strain for all PCRT specimens

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

-5000-4000-3000-2000-10000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 5 Graph of stress against strain for PCRT coupons.

6

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

-5000-4000-3000-2000-10000

Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 6 Graph of stress against strain for PCHW coupons.Applied stress versus strain for all PCCD specimens

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

-5000-4000-3000-2000-10000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 7 Graph of stress against strain for PCCD coupons.

7

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 1000 2000 3000 4000 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 8 Graph of stress against strain for OHTRT coupons.

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 1000 2000 3000 4000 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 9 Graph of stress against strain for OHTHW coupons.

8

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 10 Graph of stress against strain for OHTCD coupons.

Applied stress versus strain for all OHCRT specimens

020406080

100120140160180200220240260

280300320340

-5000-4000-3000-2000-10000Strain (x10-6)

App

lied

stre

ss (M

Pa)

B

A

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

9

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

-5000-4000-3000-2000-10000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 12 Graph of stress against strain for OHCHW coupons.

0

20

40

60

80

100

120

140

160

180

200

220

240

260

280

300

-5000-4000-3000-2000-10000Strain (x10-6)

App

lied

stre

ss (M

Pa)

Figure 13 Graph of stress against strain for OHCCD.

10

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 condition

Mean modulus, GPa± one standard deviation

Mean strength, MPa± one standard deviation

PTRT 50.8 ± 2.4 765 ± 64PTHW 49.9 ± 2.6 785 ± 50PTCD 51.2 ± 2.6 760 ± 55PCRT 47.6 ± 2.5 547 ± 61PCHW 44.9 (46.3) ± 5.7 (3.3) 508 (518) ± 59 (49)PCCD 48.9 ± 3.5 595 ± 48

OHTRT 45.5 ± 2.1 362 ± 18OHTHW 50.8 ± 2.3 418 ± 21OHTCD 51.9 (52.7) ± 4.2 (2.9) 375 (373) ± 25 (25)OHCRT 62.1 ± 4.7* 309 ± 17OHCHW 44.1 ± 2.9 300 ± 19OHCCD 45.2 ± 2.4 392 ± 25

*Proximity to the hole and/or the top grip appears to diminish strain readings, leading to an unrealistically high value of Young’s 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

11

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.

PTCD

OHTCD

PCCD

OHCCD

PTRT

OHTRT

PCRT

OHCRT

PTHW

OHTHW

PCHW

OHCHW

0

100

200

300

400

500

600

700

800

900

CD RT HW

Mea

n fa

ilure

stre

ngth

(MP

a)

n=14

n=15

n=12

n=15

n=21

n=17

n=33

n=33

n=14

n=15

n=12

n=14

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.

12

0.000

0.005

0.010

0.015

0.020

0.025

0.030

200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950Failure strength (MPa)

Pro

babi

lity

dens

ity fu

nctio

n (P

)OHCHW

OHTRTW

OHCRTOHCCD

OHTCDOHTHW

PCHW

PCRT

PCCD

PTCD

PTHW

PTRT

2

21exp

21),,(

xxP

Figure 15 Test data plotted as probability density functions (x = strength value, µ = mean and σ = standard deviation).

200

250

300

350

400

450

500

550

600

650

700

750

800

850

900

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25Number of specimens (n)

Run

ning

ave

rage

of f

ailu

re s

treng

th (M

Pa)

PTHW

PTRT

PTCD

PCCD

PCRT

PCHW

OHTHW

OHCCD

OHTCD

OHTRT

OHCRT

OHCHW

(760)

(791)

(765)

(595) (552)

(508)(418)

(392)(375)(363) (307)

(297)

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

13

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 expression:

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 B-basis 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.

OHTCD

PCCD

OHCCD

OHTRT OHTHW

PCHW

OHCHW

PTCDPTRT

PCRT

OHCRTPTHW

1

1.05

1.1

1.15

1.2

1.25

1.3

1.35

CD RT HW

Mea

n fa

ilue

stre

ngth

/B-B

asis

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

14

100

150

200

250

300

350

400

450

500

550

600

650

700

750

2 3 4 5 6 7 8 9 10 1112 13 1415 16 17 1819 20 21 22 2324 25 2627 28 29 30 3132 33 3435 36

Number of specimens (n)

B-B

asis

(MP

a)

PTRT OHTRT

PCRT OHCRT

PTHW OHTHW

PCHW OHCHW

PTCD OHTCD

PCCD OHCCD

(643)

(693)

(645)

(487)

(440)

(384) (375)(341)

(322)(326)

(260)

(272)

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

4. CONCLUDING REMARKS

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.

ACKNOWLEDGEMENTS

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).

15

REFERENCES

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: 206-224.

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 Rıo T., Barbero E., Zaera R. and Navarro C., “Static behavior of CFRPs at low temperatures”, Composites: Part B, 2002; 33: 383–390.

16