Professional Documents
Culture Documents
Abstract
In this study, the influence of compaction load, layup speed and temperature on the adhesive properties of automated
fibre placement grade towpreg has been investigated on the ply-tool interface where higher peel forces are required to
permit the deposition of subsequent plies. The automated layup process was simulated on a CNC milling machine, using
a roller assembly and the adhesion properties of the towpreg were determined using a floating roller peel test. The
processing window for the towpreg was determined using a dynamic mechanical analyser and a two-level, full factorial
design of experiments was developed for the three factors, to understand their effects on the peeling force, both
individually and synergistically. The design of experiments analysis indicates a strong temperature effect, with the
towpregs requiring a higher layup temperature to accommodate higher layup speeds. A strong load-temperature inter-
action was detected, with a negative temperature effect at lower loads and a strong positive temperature effect at higher
loads. The predicted factor settings to achieve a peeling force of 246 N/m are, 1 kN compaction load, 65 C layup
temperature, and a layup speed of 120 mm/min. Experimental tests, carried out at the predicted factor settings, agree
well with the analysis, yielding a peel force of 256 N/m with a standard deviation of 25 N/m.
Keywords
Automated fibre placement, peel test, adhesion, tack, characterisation
adhesive properties, commonly referred to as tack. In However, in the case of prepregs, probe tests have
general, the tack levels should be sufficient to hold the been found to be sensitive to changes in surface
prepreg on to the mould surface and to subsequent plies roughness that are caused by fibre architecture and
without slipping, but low enough to be repositioned in resin distribution within the specimen.20 Therefore,
case of any misalignment. In operation, prepregs or this test method may be less suitable for prepreg tack
towpregs are usually heated to make it tackier and pli- characterisation. In a relatively recent study, Crossley
able, to conform to the tool geometry. The tack levels et al.8 have proposed a new method to determine tack
at room temperature are usually provided by the manu- levels, specifically for automated lay-up processes. The
facturer,8 however, information on tackiness at other specifications of the peeling stage are similar to the
temperatures frequently has to be determined by the floating roller peel test as per standard ASTM
part manufacturer. Commonly, in a routine production D3167,21 but additionally the initial stiffness offered
cycle, the lay-up temperature is held constant9,10 or, less during peeling off the backing tape is taken into
commonly, varied linearly according to the feed rate.11 account.
Problems arising with tack during operation are fre- Taken together from literature, the tack levels for
quently tackled by experimenting with the process par- fibre lay-up operations depend primarily on lay-up par-
ameters, lay-up temperature, lay-up speed and ameters, such as the lay-up speed, temperature and
compaction load. Though it may solve the issue tem- compaction load, and any variation in these factors
porarily, such trial and error methods are time consum- will affect the tack properties of the prepreg.
ing and lead to material wastage. Therefore, it is Developing universal models to optimise tack proper-
important to have knowledge of the process and mater- ties, considering all these factors, may be difficult due to
ial parameters prior to planning an AFP schedule. inherent variability in the characteristics of the pre-
Information regarding the processing and material pregs.3 Therefore, each of these factors has been mod-
parameters are available in literature, for example, elled separately. For example, Tierney and Gillespie22
Lukaszewicz et al.3,12 have reported the void content developed a model to predict through-thickness heat
after lay-up is dependent on lay-up speed, temperature transfer and bond strength, based on intimate contact
and pressure. Crossley et al.8,13 reported that tack levels and healing at the ply interface, Lichtinger et al.23
are sensitive to both the lay-up speed and the tempera- developed a parameterised 3D Finite Element model
ture of the lay-up, Calawa and Nancarrow14 reported for the precise prediction of the thermal history
similar results mentioning that a lay-up temperature of during the complete layup process, Hörmann et al.24
35 C should be ideal for most epoxy prepreg materials. investigated how the position and orientation of infra-
Complementary observations have been made by Ahn red emitters, relative to the substrate, influence the
et al.,15 who reported that the tack of an uncured pre- material temperature and subsequently developed a
preg reaches a maximum at 20–25 C above the instant- numerical heat transfer model as a function of the pos-
aneous glass transition temperature of the resin. ition, orientation and power density of the emitter.
Gutowski16 reported that the tack is better at lower Though literature regarding the influence of varying
feed rates, recommending the machine to be operated individual process parameters on the fibre placement
at slow speed when starting a new ply. In a recent process is available, few address their synergistic inter-
study, Rao et al.17 suggested a temperature of 65 C to action on the overall lay-up optimisation. Pitchumani
be optimum when using out-of-autoclave (OOA) tow- et al.25 examined the product quality by varying ther-
preg tapes employed in the fiber placement processes. mal degradation, void content, and dimensional
Several test methods to determine tack levels of changes during the process and identified a window
adhesives are available, of which, the floating roller for tape placement to obtain optimum quality. Aized
peel test involving peeling off a flexible adhesive strip and Shirinzadeh26 used response surface methods to
from a rigid adherend is commonly used. As the vari- identify key influencing parameters in determining the
ation in thicknesses of the adherends influences the test quality of thermoplastic composite prepregs and
values, it is usually limited to 1.63 mm with the speci- reported layup temperature to be the key factor.
men thickness not exceeding 0.63 mm.18 Another Grouve et al.27 have investigated the interrelation
common test used in the pressure sensitive adhesive between process parameters, material properties and
(PSA) industry to characterise tack is the probe inter-laminar bond strength is using laser assisted tape
test,19–21 which consists of a probe coming in contact placement process for thermoplastic tapes and have
with the material. The force and/or energy of separ- demonstrated that an excellent bond quality can be
ation is measured and the stress is defined based on achieved at higher velocities and low input power
the area of the probe surface.8 Several modes of failure using mandrel peel tests.
are frequently observed, occurring in the bulk or at the Here, in order to evaluate the influence of factor
surface, by crack formation and propagation.15 effects on the tack properties of OOA prepregs during
Rao et al. 277
the fibre placement process, a factorial Design of For the peel tests, the tapes were laid on flat alumi-
Experiments (DoE) approach using orthogonal arrays num strips as per ASTM D3167 standard.21 The alu-
is used. The statistical DoE provides the possibility to minum strips were cleaned using acetone and Frekote
vary all the factors systematically in an orthogonal 700-NC mould release agent from Henkel was applied
array, allowing for the estimation of average effects, before layup. The aluminum strips were fastened to a
without the results being distorted by other factor flat mould using double-sided tape sticking tape and
effects. This has with the effect that all the trials are four towpregs were laid during one course. The milling
unique and do not repeat. The physical mechanisms machine software, MastercamÕ X7 from CNC
(e.g. heat transfer, squeeze flow, adhesion) have not Software Inc., was used to define the tool path, simu-
investigated in this work and will be reported in subse- lating the fibre deposition rate in the AFP process.
quent work. The displacement of the roller assembly, to achieve
the required compaction load, was obtained in a separ-
ate experiment with similar setup except that the dis-
Experimental procedure
placement was recorded for 0.5 and 1 kN. This value of
An in-house fibre placement system was designed and displacement was inputted into the software as a con-
built that consisted of a roller and a 20 kN load cell, stant depth of cut over the length of the layup. The
(Omegadyne LCM202) housed between the roller and distance of the air gun from the tape, and its settings,
the tool arbour of a BridgeportÕ milling machine, were also obtained from experiments with similar setup.
Figure 1(a). The layup temperature, just before the A metal ruler was placed and the distance from the tip
compaction on the flat tool was monitored using an of the heat gun and the roller was recorded until the
infrared camera, Fluke-A SC and a hot air gun from temperature of the tape reached 65 C for the whole
BoschÕ was used as the heat source, Figure 1(b). As length of the aluminum strip. After series of trials, a
emissivity affects the temperature measurements, the temperature setting of 110 C and a distance of 75 mm
emissivity factor, the IR camera was calibrated within provided a uniform 65 C at 120 mm/min; however, for
the test environment at 65 C by constantly adjusting 20 mm/min the temperature setting on the heat gun had
the emissivity factor in the camera software to match to be set to 75 C but at the same distance. The carbon
that of the k-type thermocouple and a pre-calibrated fibre OOA towpreg CYCOMÕ 5320-1 was supplied by
Fluke 566 laser temperature gun. An emissivity factor CYTEC Engineered Materials Inc. USA. The nominal
of 0.50 provided the closest match to the thermocouple width of the tape was 6.35 mm5 and the thickness was
and laser temperature gun readings with a variation of 0.275 mm. The nominal fibre areal weight is 145 gsm
1 C. Therefore, a value of 0.50 was chosen as an and the nominal resin content is 33% by weight. The
emissivity factor in this work. Furthermore, timely prepreg spool was thawed overnight and conditioned
monitoring of temperature was carried out using for 2 h at room temperature before the lay-up oper-
Fluke 566 laser temperature gun to check for any ation. Details of the processing conditions for the
deviations. tape can be found in Ref. 5.
200
Peel force (N/m)
150
100
50
0
Trials from y1 -y 8
Figure 2. Peel test results (a) average peel force of all trials with their mean deviations (b) sample peel force traces of trails y5 - y8.
Figure 3. Storage modulus E0 , loss modulus E00 , tan as function of temperature. Measuring frequency of 1 Hz and temperature ramp
of 5 C/min.
chosen following DMA tests and a previous study.17 this work, it is envisaged that the essence of factor influ-
Due to temperature setting limitations on the heat ences on the peel load will remain the same in an indus-
gun, 120 mm/min and 20 mm/min were chosen as high try setting.
and low levels of the layup speed. Though the param- This set-up leads to a typical two-level, three-factor
eters chosen are specific to the AFP simulator used in L8 layout, with eight trials to complete the entire
280 Journal of Reinforced Plastics and Composites 35(4)
Figure 4. Complex viscosity and tan traces as function of temperature at measuring frequency of 1 Hz and temperature ramp of
5 C/min.
250
225 T2
200
Peel force (N/m)
S2 (LT) 2
L2 (LST) 2
(LS) 1 (ST) 2
175
(LS) 2
150 L1 (ST) 1 (LST) 1
S1 (LT) 1
125
T1
100
experimental work. Orthogonal array has been con- orthogonal array allows for the confident estimation
sidered for estimating the best material formulation of average effects, as the factors do not influence each
needed to obtain an optimum peeling force. As the fac- other. Moreover, all the trials undertaken in this array
tors are independent of each other, the use of are unique and do not repeat. A coded design matrix,
Rao et al. 281
with factors and responses at their respective levels, is Table 4. Values used for calculation of LS interaction effect.
listed in Table 3.
Observed Average
L S (N/m) Values (N/m) (N/m)
250
Interaction effects. In order to determine whether any of
the two particular processing parameters interact or
225 not, a graphical tool called interaction graphs was
used (Figures 6 to 8). If the lines in the interaction
Peel force (N/m)
200
plot are parallel, there is no interaction between the
L2 processing parameters, implying that the change in
175
the mean response, from a low to a high level of a
L1
150 factor, does not depend on the level of the other
factor. In contrast, if the lines are non-parallel, an inter-
125
action exists between the factors. The greater the degree
100 of departure from being parallel, the stronger the inter-
1 2 action effect between the two parameters. Interaction
Levels (S)
effects in this study were established by examining the
variation of each factor with respect to the mean or
Figure 6. LS interactions effect plot, L2 ¼ 1 kN, L1 ¼ 0.5 kN, average value and calculated as mentioned in
S1 ¼ 20 mm/min, S2 ¼ 120 mm/min. literature.32
Figure 8. ST interaction plot S1 ¼ 20 mm/min, S2 ¼ 120 mm/ ST interaction effect. In Figure 8, a positive inter-
min, T1 ¼ 25 C, T2 ¼ 65 C. action is seen, with the force required to peel the tape
Rao et al. 283
off the rigid adherend increasing as the temperature changed from a lower level to a higher level. The
increases from T1 to T2. This interaction suggests that values of the interaction factors are listed in Table 6.
if higher deposition rate is required, the layup tempera- Considering interaction factors to estimate overall
ture must be increased accordingly (65 C for 120 mm/ peel force, results in the following equation
min in this case). The interaction is not as severe as the
LT interaction and the contribution of this interaction ymax ¼ y þ ðL contributionÞ þ ðS contributionÞ
produces a 13 N/m increase in the peel force when þ ðT contributionÞ þ ðLS contributionÞ
þ ðLT contributionÞ þ ðST contributionÞ
Table 5. Values used for calculation of LT interaction effect. þ ðLST interactionÞ
L T Observed Values Average
Peel force ¼ 169:65 þ 7:70L2 þ 8:49S2
1 1 118.50 150.10 134.30
þ 42:86T2 4:15ðLSÞ2 þ15:21ðLTÞ2 ð3Þ
1 2 180.12 199.08 189.60
2 1 131.14 107.44 119.29 þ 6:52ðSTÞ2 þ9:68ðLSTÞ2
2 2 214.88 255.96 235.42
where LST is the three factor interaction estimated as
per literature.32
From Figures 5 to 8, it is evident that all factors
Table 6. Values used for calculation of ST interaction effect.
show some contribution to the overall peel force, and
S T Observed Values Average their significance can be determined using half-normal
plots.33 Less significant factors tend to manifest them-
1 1 118.50 131.14 124.82
selves as being near zero, displaying a normal distribu-
1 2 180.12 214.88 197.50
tion centered about the line passing through the origin,
2 1 150.10 107.44 128.77 while significant factors depart away from zero, having
2 2 199.08 255.96 227.52 a normal distribution centered about their respective
larger values.
Table 7. Error in the predicted and measured values considering only two-factor effects.
Factor
Experimental Predicted Absolute
Trial L S T (N/m) (N/m) error (%)
y1 1 1 1 118.5 128.2 8
y2 1 1 2 180.1 170.4 5
y3 1 2 1 150.1 140.4 6
y4 1 2 2 199.1 208.7 5
y5 2 1 1 131.1 121.5 7
y6 2 1 2 214.9 211.9 1
y7 2 2 1 107.4 117.1 9
y8 2 2 2 255.9 246.3 3
In Figure 9, it is clear that temperature plays an i.e. LS and ST. Therefore, including those factors in
important role, deviating significantly from the zero predicting the peel force, using equation (3) results in
line. This is followed by load-speed and speed-tem- 255 N/m.
perature interaction factor effects. The interactions, Correlating the actual experimental results with the
such as lay-up speed, load-temperature interaction predicted results, shown in Figure 10 and Table 7, it
and the three-way interaction between the factors is evident that the values predicted using equation (3)
follow a normal distribution, clustering around zero are within 10% of the experimental data, with the
and not contributing significantly to the overall peel maximum error being 9% for trial y7 in Table 7,
force. However, in the prediction of the peel force, when the compaction load, lay-up speed and lay-up
the lay-up speed factor ‘S’ must be considered, as it temperature factors are set at levels 2, 2 and 1,
is present in both the contributing interaction effects, respectively.
Rao et al. 285
23. Lichtinger R, Hörmann P, Stelzl D, et al. The effects of 29. Stark W. Investigation of the curing behaviour of carbon
heat input on adjacent paths during Automated Fibre fibre epoxy prepreg by dynamic mechanical analysis
Placement. Compos: Part A 2015; 68: 387–397. DMA. Polym Test 2013; 32: 231–239.
24. Hörmann P, Stelzl D, Lichtinger R, et al. On the numer- 30. Lukaszewicz DHJA and Potter K. Through-thickness
ical prediction of radiative heat transfer for thermoset compression response of uncured prepreg during manu-
automated fiber placement. Compos: Part A 2014; 67: facture by automated layup. Proc Inst Mech Eng Part B:
282–288. J Eng Manuf 2012; 226: 193–202.
25. Pitchumani R, Gillespie JW and Lamontia MA. Design and 31. Robert W and Messler J. Joining of Materials and
optimization of a thermoplastic tow-placement process with Structures. United States of America: Elsevier, 2004.
in-situ consolidation. J Compos Mater 1997; 31: 244–275. 32. Lochner RH and Matar JE. Designing for quality: An
26. Aized T and Shirinzadeh B. Robotic fiber placement pro- introduction to the best of Taguchi and Western methods
cess analysis and optimization using response surface of statistical experimental design. Netherlands: Springer,
method. Int J Adv Manuf Technol 2009; 55: 393–404. 1990.
27. Grouve WJB, Warnet LL, Rietman B, et al. Optimization 33. Daniel C. Use of half-normal plots in interpreting
of the tape placement process parameters for carbon–PPS factorial two-level experiments. Technometrics 1959; 1:
composites. Compos: Part A 2013; 50: 44–53. 311–341.
28. Xie M, et al. A new method to characterize the cure state
of epoxy prepreg by dynamic mechanical analysis.
Thermochim Acta 2009; 487: 8–17.
Appendix – 1
1 2 1 2 25 60 1 2 1 2 1 2 1 2