Download as pdf or txt
Download as pdf or txt
You are on page 1of 22

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258158283

The experimental static mechanical performance of ironed repaired GFRP–


honeycomb sandwich beams

Article in Journal of Sandwich Structures & Materials · November 2012


DOI: 10.1177/1099636212460538

CITATIONS READS

4 2,052

3 authors, including:

Benjamin Kirollos

12 PUBLICATIONS 98 CITATIONS

SEE PROFILE

All content following this page was uploaded by Benjamin Kirollos on 27 January 2019.

The user has requested enhancement of the downloaded file.


Article
Journal of Sandwich Structures and Materials
14(6) 694–714
The experimental static ! The Author(s) 2012
Reprints and permissions:
mechanical performance sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1099636212460538

of ironed repaired jsm.sagepub.com

GFRP–honeycomb
sandwich beams
Benjamin William Mounir Kirollos1,2,
Richard Trede1 and Peter Lampen1

Abstract
Damaged glass fibre reinforced plastic–honeycomb core sandwich beams are repaired
using uncured glass fibre reinforced plastic fabrics and a handheld iron. The effect of iron
temperature, application time and pressure on the effectiveness of repair is investigated
by measuring the failure load and flexural stiffness of the repaired beams using third span
four-point bending tests. Repairs are tested in compression and tension. A repair pro-
cess is suggested which consistently recovers 95% of the compressive strength and 77%
of the tensile strength of the damaged beam. The repair is shown to have little effect on
beam flexural stiffness.

Keywords
Honeycomb sandwich, repair, iron, strength, stiffness, four-point bending, experimental

Introduction
Composite materials have become a mainstay of modern aviation. Their strength
and stiffness to weight ratio makes them extremely attractive to airline companies,
which are striving to reduce aircraft weight and increase fuel efficiency. The Boeing
787 Dreamliner is a prime example of the use of composite materials in aircraft,
with over 50% of the primary structure (including wings and fuselage) made from
composite materials [1]. The use of composite materials also extends to the aircraft

1
Mühlenberg Interiors GmbH & Co KG, Hamburg, Germany
2
Department of Engineering Science, University of Oxford, Oxford, UK

Corresponding author:
Benjamin William Mounir Kirollos, University of Oxford, Oxford, UK.
Email: ben.kirollos@eng.ox.ac.uk
Kirollos et al. 695

interior. Galleys, stowages, partitions and other similar cabin components are
manufactured to be as light and as stiff as possible. Though their application is
less demanding than the airframe, aircraft interiors must still satisfy strenuous
crashworthiness and fire-safety requirements [2].
By far, the most popular composite material used for aircraft interiors is the
honeycomb sandwich panel [3]. Typically, this consists of a honeycomb core made
from aramid fibre paper faced with glass fibre reinforced plastic (GFRP) thermoset
skins (Figure 1), or for stiffness-critical applications carbon fibre reinforced plastic
skins. The skins provide the stiffness and strength of the panel whilst the honey-
comb core acts as the lightweight means to separate the skins and transmit shear
forces. The role of the skins and core is similar to that of the flanges and web of
an I-beam.
Damage to the panels, either during manufacture or from misuse during service
life, can reduce the structural and aesthetic quality of the panel. Panels are usually
too costly to replace entirely and so any damage is repaired locally. One common
method of repairing damaged honeycomb cores for aircraft interior applications is
to fill the affected area of the sandwich panel with a suitable edge filler material.
GFRP skin damage, on the other hand, is commonly repaired by gluing a pre-
cured patch made from the same skin material over the damaged area. The repair
patch is then sanded and surfaced with a decorative material to cover surface flaws.
Although this method of skin repair is effective in recovering most of the original

Figure 1. Layup and orientation of layers of the honeycomb sandwich panel; t ¼ thickness.
696 Journal of Sandwich Structures and Materials 14(6)

aesthetic and structural properties of the panel, it has the disadvantage of being
time consuming and results in relatively thick repairs. The adhesive takes time to
prepare and the layers that form the repair patch must be pre-cured.

Aim
The aim of the study was to develop a novel repair method that would recover the
original strength and stiffness of a damaged honeycomb–GFRP sandwich panel,
whilst being quicker, simpler and more cost-effective to implement than current
practices. This would significantly cut down on manufacturing and repair costs.
A time of repair of no more than a few minutes was targeted.
The strategy presented in this article uses the pre-impregnated phenolic resin
within the uncured GFRP fabrics as the adhesive which bonds the repair to the
panel. The same principle is used during manufacture of the original panels
(‘GFRP–honeycomb sandwich panels’ section). This eliminates the need for any
external adhesive. In the proposed strategy, the GFRP fabrics are cured onto the
damaged panel using the heat of a standard handheld iron pressed by hand. The
use of a household iron greatly reduces cost and complexity. The flat surface of
most aircraft interior panels is conducive to the use of an iron, whereas curved
airframe surfaces would require a more complicated application method, e.g.
vacuum moulding [4]. Although the use of pre-impregnated resin as the repair
adhesive is not new [4], the authors believe this to be the first reported use of
heat and pressure being applied by a handheld iron.
The repair conditions which best recovered the original structural properties
were sought. Three main factors affecting repair strength were investigated: the
temperature of the iron; the time for which the heat was applied; the force with
which the iron was applied. The effect of the surface preparation of the damaged
area was also considered. The strength and stiffness of repair was quantified using
four-point bending tests.
The paper consists of five main parts:

1. Strength and stiffness of a set of undamaged panels cut into beams is measured
(seven samples).
2. Repairs made at five temperatures (85 C, 110 C, 120 C, 130 C and 155 C) over
five lengths of time (30, 60, 90, 120 and 150 s) at constant load (147 N) are tested
in compression. A first approximation of the optimal repair temperature and
time is made (25 samples).
3. Repairs made at six different loads (29, 59, 88, 118, 147 and 176 N) at constant
temperature and time (130 C–90 s) are tested in compression (six samples).
4. Three temperature–time combinations (130 C–90 s, 155 C–90 s and
130 C–150 s) at a constant load (147 N) are investigated in further detail.
Many repeat tests are undertaken to better quantify the effectiveness of the
three repair methods (48 samples).
Kirollos et al. 697

5. The optimum time–temperature–load combination is identified (130 C–


90 s–147 N) and is tested in tension (six samples).

The GFRP–honeycomb sandwich panel


The honeycomb sandwich panels used for this study were manufactured by
Mühlenberg Interiors GmbH, Hamburg. The sandwich panels were symmetric
and were composed of a honeycomb core faced on both sides by two layers of
bi-directional GFRP. The core was an aeronautical quality EuroComposites
Nomex honeycomb core of density 48 kg/m3, cell size 3.2 mm and nominal thick-
ness 14 mm. The skins consisted of an inner layer of Gurit PHG 600-68-50 and an
outer layer of Gurit PHG 600-44-50. Both fabrics have a marginally higher
Young’s modulus in direction 1 and are orientated such that direction 1 is aligned
with the honeycomb core transverse direction (W). The layup and orientations of
the core and skins are shown in Figure 1.
The panels are manufactured using flat press moulding, in which the lower
uncured GFRP, honeycomb core and upper uncured GFRP are stacked onto a
heated steel caul plate [3]. A second heated caul plate in a hydraulic plate is lowered
onto the layup which holds the stack under a pressure of at least 0.07 MPa for
approximately 40 min at 140 C.
Preliminary tests showed that the sandwich panel was stronger and stiffer in
bending when the core transverse direction (W) was aligned with the bending axis.
Therefore, to ensure that the repair would be subject to the most intense condi-
tions, the Mühlenberg panels were cut into test sections of length 400 mm and
width 50 mm such that the longest dimension of the beam was aligned with the
honeycomb ribbon direction (L). The width was chosen to maximise the number of
samples cut from a single panel whilst easily satisfying the ASTM standard four-
point bending guideline that the width should be at least double the thickness of the
specimen [5].
This study focuses on the repair of damaged skins since they are the main source
of strength and stiffness of the panel. As such, this study does not consider the
repair of the honeycomb core. The damage was simulated by a 1.5 mm wide gap
through the top skins (both PHG 600-44-50 and PHG 600-68-50) at the centre of
the beam, leaving the core and lower laminate in tact. The centre of the gap was
aligned with the same repeating structure in the underlying honeycomb
core (Figure 2).

Repair procedure
An overlap repair was preferred to a scarf repair due to its ease of application.
It has been shown that matching the stiffness of the repair patch to the stiffness of
the parent laminate is important for maximising repair effectiveness [4]. Therefore,
the repair patch consisted of an inner layer of Gurit PHG 600-68-50 and an outer
698 Journal of Sandwich Structures and Materials 14(6)

Figure 2. Test sample dimensions and position of the cut relative to repeating core
structure.

layer of Gurit PHG 600-44-50 (Figure 3). The PHG 600-68-50 and PHG 600-44-50
fabrics were cut into identical 80  50 mm2 patches. The patches were cut such that
the principle direction of the fabric (direction 1) was aligned with the longest
dimension of the patch. This was to ensure that the repair patch had identical
orientation to the underlying laminate.
All materials were cut from new, since preliminary tests demonstrated that
exposure of an uncured fabric to standard room conditions for several days
resulted in a loss of tackiness and a subsequent stiffening of the uncured repair
fabrics. To keep peel stresses to a minimum, it is normally recommended that the
ratio of overlap length to repair patch thickness should be at least 30 in supported
lap repairs [6]. Preliminary tests showed a maximum repair patch thickness (after
the proposed repair process) of 0.6 mm, which justified the overlap length of
39.2 mm used for all subsequent tests.
Initial tests showed that sanding the surface of the damaged beam before
the repair patch was applied increased the reliability of the repair. Repairs
sanded with grade P40 paper had a lower chance of failure by repair patch delam-
ination than P60 sanded-repairs or repairs without sanding. All subsequent
damaged samples were therefore sanded with P40 over the surface area where
the repair patch would be applied. After sanding, the area was cleaned with
a dry cloth to remove dust. Additional preparation of the damaged area was
deemed unnecessary (e.g. the use of acetone to clean the damaged area after
Kirollos et al. 699

Figure 3. Position, layup and dimensions of repair.

Figure 4. Repair method illustrating direction of resin flow.

sanding). This was justified by later tests, which showed that repairs made at the
optimal temperature–time–load conditions did not delaminate from the parent
beam; this suggested that the quality of the interface between the repair patch
and parent beam was sufficient.
The repair patch was centred over the damage and the rougher sides of the PHG
600-88-50 and PHG 600-44-50 fabrics were faced towards the parent beam. The cut
was not filled. A 120  120 mm2 piece of silicone-layered paper was centred over the
top of the repair patch and wrapped around the parent beam. The paper was
secured to the beam with masking tape. The oversized silicone-layered paper pre-
vented the pre-impregnated resin flowing from the fabrics to the iron, but allowed
clearance for excess resin to flow out of the repair patch (Figure 4). Securing the
700 Journal of Sandwich Structures and Materials 14(6)

Figure 5. Temperature at six positions on the Philips Trocken-Bügeleisen when set halfway
between the ‘oo’ and ‘ooo’ settings.
Area-weighted time averaged temperature ¼ 130 C. Area in grey shows the part of the iron used
for the repair.

silicone-layered paper prevented the repair patch from moving during the repair
process, which was important since initial tests showed that movement during the
process was detrimental to the final repair strength.
The heat was applied using a steamless iron (Philips Trocken-Bügeleisen). The
temperature of the iron was monitored using a handheld thermocouple. The
iron temperature varied with position and time, with a repeating period of
approximately 2 min. For a given position and setting, the temperature could
vary by 15 C over a 2-min period. To ensure repairs were subjected to the same
heating profile, the repair process was initiated when the indicator light switched
off. The same area of the iron base was used for all repairs and is shown in Figure 5.
The iron was not moved during use. All subsequent iron temperatures quoted in
the article are area-weighted time averaged values over the area of the iron used
for repair.
The iron was pressed onto the repair by hand and the force measured by a
weighing scale placed under the sample. Contraction of the repair was noted
after cooling, causing the parent beam to bend upwards and an initial central
deflection of the repaired beam of up to 2.5 mm. In general, a full-scale panel
would be wider and therefore stiffer in bending than the beams tested here; the
same size repair patch (80  50 mm2) would cause significantly less deflection in a
full-scale panel. So the samples had negligible initial deflection and the parent beam
was pre-displaced during the repair by 1 mm in the opposite direction and clamped.
This resulted in a level repaired beam after unclamping and cooling. Figure 6 shows
a damaged beam and a typical repair.
Kirollos et al. 701

Figure 6. A damaged and typically repaired beam.

Figure 7. Four-point bending setup with third span loading.

Four-point bending test method


Four-point bending tests were used to determine the beam failure load and stiff-
ness. A Zwick/Roell Z050 static load test machine was configured for third span
loading and linked to a PC for data acquisition. The span between the lower sup-
ports was set to 300 mm (Figure 7) so that the load points were outside the repaired
zone. Rubber pads (20 mm wide) were placed under the loading points to spread
the load and prevent local skin indentation and core crushing. Load and deflection
data were recorded up to failure using testXpert software. The deflection of the
load test machine was negligible in comparison with the deflection of the sample
and was therefore discounted.
Each loading roller excerpts a load of P/2 onto the specimen. The load
recorded is the sum of the load through both rollers, P. The flexural stiffness D
is calculated as [7]
1
D¼ ð1Þ
ð1296=23PL3 Þ  ð216=23L2 UÞ
where L is the span between the supports and  the central deflection;
/P is evaluated in the linear region of the force–displacement curve. U is the
shear rigidity and is given by [5]

Gðd þ cÞ2 b
U¼ ð2Þ
4c
702 Journal of Sandwich Structures and Materials 14(6)

Table 1. Test results for undamaged beams

Failure load per Flexural stiffness per


unit width (N/mm) unit width (MN mm)

Maximum 26.0 1.06


Minimum 22.4 1.14
Mean/median 24.7/24.8 1.11/1.11
SD 1.4 0.03

where G is the core shear modulus [8] and b, c and d the width, sandwich thickness
and core thickness, respectively.

Undamaged beam results


Seven identical undamaged sandwich beams were tested to failure using the four-
point bending test. The failure load and force–displacement data were recorded
and the flexural stiffness calculated using equations (1) and (2). The results are
presented in Table 1.
Six of the specimens failed by compressive laminate fracture, four of which
occurred in the vicinity of the loading points. The one remaining beam failed by
tensile laminate fracture. It was concluded that local stress concentrations under
the loading rollers did not cause premature failure, since three specimens failed
away from the loading zone within a SD of the mean ultimate failure load. This
conclusion was supported by the absence of local core crushing under the loading
points. Therefore, the specimen width of 50 mm was deemed sufficient for this
study, although verging on the lower limit of acceptable width for four-point
bending tests.
Tests were also undertaken to examine how direct heating of the undamaged
beam affected failure strength. Heat was applied directly with an iron and without
silicone-coated paper over the central 50  80 mm2 of the beam for 90 s at 130 C.
The heated side of the beam was placed in compression and tested to failure. The
results from seven samples showed a mean decrease in failure load of 1.7 N/mm:
a 7% reduction. Failure occurred within the heated zone, between the two loading
points and approximately 30 mm from a loading roller. However, it was concluded
that the additional layers of repair fabrics and silicone-coated paper during an
actual repair would reduce the heat transferred to the parent laminate. As such,
it was concluded that the effect would be negligible.
Damaged but unrepaired beams were also tested. Due to the absence of the top
skins, failure by core crushing occurred immediately. The maximum load-carrying
capacity of the damaged beams was approximately 30% of the undamaged equiva-
lent, clearly showing the need for repair.
Kirollos et al. 703

Preliminary investigations
First, repairs were made using a wide range of temperatures and times at constant
load, in order to narrow down the optimum repair conditions. The temperature
range was necessarily limited to the operating range of the iron to keep the repair
process as simple as possible. The time of repair was capped at 150 s; longer repair
times were shown to have little added benefit to the repair effectiveness.
Second, the relationship between applied loading and repair strength was inves-
tigated. Repair time and temperature was kept constant. The applied loading was
capped at 176 N so the repair could be carried out comfortably by hand. This
equates to a pressure of 0.44 bar, which is of the same order as the manufacturer’s
recommended pressure of at least 0.7 bar when joining the PHG600 series fabrics to
a honeycomb core [9].

Repair time and temperature


Twenty five specimens were repaired using five temperatures (85 C, 110 C, 120 C,
130 C and 155 C) over five lengths of time (30, 60, 90, 120 and 150 s) at constant
load (147 N). The repaired beams were tested in compression until failure and the
ultimate failure load recorded. Trends are shown in Figure 8.
The repairs made at temperatures 120 C were not as effective as repairs made
at higher temperatures. The maximum failure load of repairs made at 120 C or
below was 16.3 N/mm. In comparison, repairs made at higher temperatures
attained maximum failure loads of 24.3 N/mm (130 C–150 s) and 23.0 N/mm
(155 C–90 s) which approach the strength of an undamaged beam (24.8 N/mm).
The repair effectiveness tended to increase with repair time for all temperatures up
to approximately 120 s. Over 120 s, the repair effectiveness appeared to plateau for
repair temperatures 130 C. A trend was not obvious for repairs made at 155 C
over 120 s.
During the high-temperature repairs, gaseous release from the repair fabrics
was observed. This is due to boiling of the water and volatile content of the fab-
rics (organic solvents and other ingredients). The volatile content of the PHG600-
44-50 and PHG600-68-50 can account for up to 8% of the fabric weight [9].
It was also noted that the high-temperature repairs often had potmarks on their
surface, which may have been due to rapid gaseous release. Initial indications
showed that heavily potmarked repairs were less effective than repairs with a
smooth surface.
The final colour of the repair varied significantly with the repair conditions. Low
temperatures (85–110 C) produced transparent repair patches that retained their
uncured tackiness; medium temperatures (120 C) resulted in white repairs; high
temperatures (130–155 C) produced light to dark brown repairs. This may be an
indication of the final chemical and structural composition of the repair patch. The
final repair colour was marked as a possible indicator of the repair effectiveness.
Repairs made at higher temperatures for longer times appeared most promising
and were the subject of further tests.
704 Journal of Sandwich Structures and Materials 14(6)

Figure 8. Failure load per unit width of repaired beams for varying repair times and tem-
peratures (repaired side in compression).
Logarithmic trends are shown. n ¼ 5 for each repair temperature. A reliable trend could not be
discerned for repairs made at 155 C.

Repair load
Repairs made at six applied loads: 29, 59, 88, 118, 147 and 176 N. The repair
temperature and time was kept constant at 130 C and 90 s. The repaired beams
were tested in compression until failure and the ultimate failure load recorded
(Figure 9) and flexural stiffness calculated (Figure 10).
The results showed a clear increase in repair effectiveness with applied load.
Repairs made at 147 and 176 N recovered 95% and 98%, respectively, of the
median strength of the undamaged beams (24.8 N/mm). These repairs returned
the damaged beam to at least the 25th percentile of the failure load of undamaged
beams. In terms of flexural stiffness, beams repaired with an applied load of 120 N
or more were stiffer on average than the undamaged beams; the beam repaired at
176 N was 7% stiffer on average than the undamaged samples. In contrast, the
lower repair loads (<120 N) decreased the stiffness of the beam by up to 4%.
When choosing the best repair load, a number of factors were considered. Whilst
the repair ought to be as strong as possible, a low repair load is preferable as it will
be applied by hand. It was also considered that achieving a repaired beam stiffness
close to stiffness of an undamaged beam was beneficial for the quality of the
repair [4]. Based on these conditions, an applied force of 147 N was chosen for
all subsequent tests.
Kirollos et al. 705

Figure 9. Failure load per unit width for varying applied repair force (repair in compression).
The force was applied for 90 s at 130 C. For comparison, the 25th and 75th percentiles of the set
of undamaged beams are shown by the lower and upper limits of the greyed area. The central
grey line is the median failure load of the undamaged beams (24.8 N/mm). Error bars are an
approximation based on the failure load range of 17 tested beams repaired at 130 C for 90 s at
147 N (Figure 11).

Detailed repaired beam results (repaired side in compression)


From the initial investigations, the three most promising repair conditions
were identified and chosen for repeat tests: 130 C–90 s–147 N (17 samples),
155 C–90 s–147 N (18 samples) and 130 C–150 s–147 N (13 samples). All repairs
were tested in compression using the method detailed. Boxplots of ultimate failure
load and flexural stiffness are shown in Figures 11 and 12.

Failure load
Repairs made at 130 C and 90 s recovered on average 95% of the median undam-
aged beam strength, with a median failure load of 23.6 N/mm. Significantly, 25%
of repairs made at these conditions failed above the median undamaged beam
strength. Indeed the strongest repaired beam was manufactured using this repair
method and was 6% stronger than an average undamaged beam. The increase in
strength can be attributed to the increase in second moment of area of the beam,
caused by the added thickness of the repair. The failure load SD of beams
706 Journal of Sandwich Structures and Materials 14(6)

Figure 10. Flexural stiffness per unit width for varying applied repair force (repair in
compression).
The force was applied for 90 s at 130 C. For comparison, the 25th and 75th percentiles of the set
of undamaged beams are shown by the lower and upper limits of the greyed area. The central
grey line is the median flexural stiffness of the undamaged beams (1.11 MN mm). Error bars are an
approximation based on the flexural stiffness range of 17 tested beams repaired at 130 C for 90 s
at 147 N (Figure 11).

manufactured at 130 C for 90 s was 1.7 N/mm, cf 1.4 N/mm SD of the undamaged
beams. This similarity in SD gives confidence in the reliability of this repair
method.
Similarly, repairs made at 130 C for 150 s also recovered on average 95% of the
beam strength. However, the effectiveness of this repair method was less consistent,
with a failure load SD of 2.2 N/mm. This was mirrored in the 155 C–90 s repair
method, which had a similar consistency of SD ¼ 2.3 N/mm. Repairs made at
155 C for 90 s produced the weakest repairs, on average recovering 93% of the
failure load.

Flexural stiffness
Figure 12 shows the calculated flexural stiffness for the three repair conditions. All
repair methods showed negligible difference in flexural stiffness compared to the
undamaged beams. Repairs made at 130 C for both time periods were on average
2% stiffer. In contrast, repairs made at 155 C for 90 s were 1% less stiff on average
Kirollos et al. 707

Figure 11. Boxplot of failure load per unit width for three repair time/temperature combin-
ations at 147 N (repaired side in compression).
Outliers are denoted ‘+’. The lower and upper limits of the greyed region are the 25th and 75th
percentiles of the undamaged beams. The central grey line represents the undamaged median
failure load.

than the undamaged beams; they also proved to have the greatest SD (0.05 N mm).
It should be noted, however, that all repair methods had comparable SD
(0.03–0.05 N mm) to the undamaged beam set (0.03 N mm). Repairs made at
130 C for 150 s were consistently stiffest. It was concluded that the stiffness of all
three types of repair were acceptable for the operational requirements of the
material.

Failure loci
The mode and position of failure was a good indicator of the effectiveness of a
given repair method. Those failed outside the repaired zone were regarded as suc-
cessful since the parent beam and not the repair was the limiting factor in the
strength of the panel. Table 2 summarises the various failure modes of the different
repair methods.
Of the repairs made at 130 C for 90 s, 82% failed outside the repaired zone. This
was seen as a great success for that particular repair method. Approximately, half
of the 130 C–150 s repairs failed outside the repaired zone, whilst those repaired
708 Journal of Sandwich Structures and Materials 14(6)

Figure 12. Boxplot of flexural stiffness per unit width for three repair time/temperature
combinations at 147 N (repaired side in compression).
The lower and upper limits of the greyed region are the 25th and 75th percentiles of the
undamaged beams. The central grey line represents the undamaged median stiffness.

at 155 C for 90 s had limited success with only 23% failing outside of the repair.
Failures outside of the repaired zone were distributed between fracture of the
parent laminate in compression (Figure 13) and fracture of the parent laminate
in tension with an associated ripping of the honeycomb core structure (Figure 14).
For all repair methods, a greater percentage of failures occurred on the tensile side
of the beam than the compressive side. In contrast, the undamaged beams failed
predominantly on the compressive side. The difference in failure location may be
explained by the stiffening of the compressive laminate due to the repair patch,
preventing local buckling of the compressive laminate.
Three main failure methods of the repaired zone (in compression) were identi-
fied: fracture of the repaired patch directly above the parent laminate damage
(Figure 15); delamination of the repair patch from the parent laminate, either of
one entire side (Figure 16(a)), the middle portion (Figure 16(b)) or in a V-shape
(Figure 16(c)); and a combination of both fracture and delamination. Fracture
occurred more often than delamination for all three time/temperature combin-
ations. During the initial tests in which much lower temperatures were tested,
the most prevalent mode of failure was delamination. The reduced incidence of
delamination was seen as a good indication that the bond between the repair patch
and parent laminates was sufficient to transmit the required shear forces. The
fracture of the repair directly above the damage is probably due to lack of support
for the repair by the honeycomb core at this point.
Kirollos et al. 709

Table 2. Summary table of repair failure modes

Repair in Failure outside Failure of


Repair compression/ of repaired repaired
strategy tension zone (%) zone (%)

Undamaged – 100 (6,1)a –


130 C–90 s Compression 82 (6,8)a 18 (1,2,0)b
155 C–90 s Compression 23 (1,2)a 77 (3,5,2)b
130 C–150 s Compression 56 (4,6)a 44 (2,3,3)b
130 C–90 s Tension 67 (0,4)a 33 (2,0,0)b
a
Values in parentheses refer to the number of samples which failed by fracture of the
parent laminate in compression and tension, respectively.
b
Values in parentheses refer to the number of samples which failed by repair delamin-
ation, repair fracture and a combination of repair delamination and fracture,
respectively.

Figure 13. Compressive failure of parent laminate.


All failures of this kind occurred within the shaded region.

Figure 14. Tensile failure of parent laminate.


All failures of this kind occurred within the shaded region.
710 Journal of Sandwich Structures and Materials 14(6)

Figure 15. Compressive failure of the repair by fracture.


The fracture forms directly over the imposed parent laminate damage.

Discussion
All three time/temperature combinations were successful in recovering the majority
of the undamaged beam strength. In some trials, the repaired beams were stronger
than the undamaged beams. Specimens repaired at 130 C for 90 s were consistently
strongest, though only marginally. Repairs using this method failed outside the
repair zone in the majority of cases, which clearly differentiated it from the other
methods in the absence of a distinct preference in terms of flexural stiffness or
failure load. Accordingly, the repair conditions 130 C for 90 s at 147 N were
chosen for further static tests in which the repaired side would be tested in tension.

Surface appearance
Interestingly, visual inspection of the repair patches could signal the effectiveness of
the repair. The most successful repairs were light brown in colour (Figure 17).
Darker brown repairs often failed by fracture or delamination, whilst white or
transparent repairs were too tacky and failed prematurely. Often, potmarks
could be seen in repairs made at 155 C and was accompanied during the repair
process by gaseous release. Whether the rapid gaseous release (water, organic solv-
ents and other ingredients) was a reason for the moderately lower strength of
samples made at 155 C is possible. An even spread of resin out of the sides of
the repair patch demonstrated that even pressure and temperature was applied over
the repair patch during manufacture and was a necessary sign of an effective repair.
These visual cues would be useful for in-line repair quality control. Unintrusive
microstructural analysis of the repair would be useful to further qualify the nature
of each cure system. Such technologies were unavailable to the authors.

Repaired beam results (repaired side in tension)


Six damaged samples were repaired with a temperature of 130 C for 90 s, at a load
of 147 N. The repaired side was tested in tension by positioning the repair on the
lower surface during four-point bending (upside down in Figure 7).
Kirollos et al. 711

Figure 16. Failure by delamination of the repair patch: (a) complete delamination of one
side; (b) delamination centred above the parent laminate damage; and (c) V-shaped
delamination.

The SD in failure load was 1.8 N/mm, which was very similar to the SD of
1.7 N/mm of the same repairs in compression. The average failure load was
19.2 N/mm (both median and mean), corresponding to a strength recovery of
only 77%. This was also observed in the work of Mahdi et al., where it was
demonstrated that overlap repairs were considerably more effective in compression
than in tension. It is interesting to note, however, that four out of six failures
occurred outside the repaired zone (Figure 18).
Mahdi et al. postulated that the reduction in failure load may be due to the
increase in stress and strain concentrations resulting from the stress transfer at
the end of the overlap. This agrees with the observations in this study, in which
the limiting stress concentrations appear within the parent laminate. A solution
712 Journal of Sandwich Structures and Materials 14(6)

Effective
repairs

(a) RGB (8bit) = 184:91:67 (b) RGB (8bit) = 190:114:98

Ineffective
repairs

(c) RGB (8bit) = 156:45:34 (d) RGB (8bit) = 178:162:147

Figure 17. Surface appearance of effectual (a, b) and ineffectual (c, d) repairs.
Repairs shown in (a) and (b) are evenly light brown in colour, have few potmarks and have an even
spread of resin from the sides; (c) is unevenly dark brown and heavily potmarked, with an uneven
spread of resin; and (d) is unevenly white and still slightly tacky to the touch.

Figure 18. Tensile failure of parent laminate.


All failures of this kind occurred within the shaded region.

may be to eliminate the step-change in laminate thickness to form a smooth tran-


sition by sanding the repair patch edges. This is more akin to a scarf repair which is
shown in the literature to be effective in tension [4]. Unfortunately, scarf repairs
have been shown to be less effective in compression.
The flexural stiffness of the repaired beams in tension was 1.15 MN mm, an
increase of 4% on the undamaged beams. It is to be expected that the repaired
beam would be stiffer than the same undamaged beam because of the additional
thickness. It is unclear, however, why the same repair would be stiffer in tension
more than compression.
Kirollos et al. 713

Conclusions
This study was successful in showing the viability of a simple time- and cost-saving
technique for repairing damaged GFRP–honeycomb core sandwich panels.
It was shown that the repair effectiveness varies significantly with the applied
time, temperature and pressure of the iron. Clear advantage was gained from
high pressures, long cure times and medium to high temperatures. The repair
conditions 130 C–90 s–147 N were deemed to maximise repair strength whilst redu-
cing by as much as possible the time and pressure of the repair. These conditions
are therefore suggested for this kind of repair. It should be noted that the proposed
procedure in its current form should be limited to flat, non-critical components,
due to the individual human factor associated with each repair. It is not intended to
compete with vacuum bagging in the repair of critical components and/or curved
surfaces.
It was demonstrated that a damaged beam, repaired using the suggested time/
temperature/pressure conditions, recovers on average 95% of its undamaged
strength when the repair is subjected to compressive forces. The SD in failure load
of 17 samples, repaired at the suggested conditions and tested in compression, was
1.7 N/mm. Since this was comparable to the undamaged beam SD of 1.4 N/mm, it is
concluded that this repair strategy is consistently effective in compression.
The same repair was weaker when subjected to tensile forces; on average, the
repair in tension recovered 77% of the undamaged beam strength. The decrease
was attributed to stress and strain concentrations due to the step change in beam
thickness at the end of the repair. A method to improve the tensile properties of the
repair is suggested by the removal of the step change to smooth the transition
between the repair and parent laminate.
Repairs made at the suggested conditions were shown to increase the stiffness of
the beam by up to 4%. For the usual applications of GFRP–honeycomb sandwich
panels, a 4% difference in stiffness is insignificant and therefore it was concluded
that the repairs satisfy stiffness requirements.
The mode and position of failure depended on the exact conditions of repair.
Eighty two percent of repairs made using the suggested conditions failed outside
the repaired zone, which was regarded as a success for the repair method. Fracture
of the repair patch over the cut was the main cause of failure within the repaired
zone, possibly due to lack of support of the repair laminate above the damage.
Failures of this kind could be reduced by filling the damaged space with GFRP
material.
Visual inspection of the repair was a good indicator of repair effectiveness,
a useful property for industrial application. Light brown repairs with few potmarks
and even resin flow were most likely to result in an effective repair. Naturally, these
visual cues are specific to the materials used in this study. Further improvements to
the method could include applying a dead load to the repair for a set time after
heating and using a bespoke iron with improved temperature control. Research
into the cure temperature and load gradients across the repair patch is also
suggested.
714 Journal of Sandwich Structures and Materials 14(6)

Funding
This study received no specific grant from any funding agency in the public, commercial, or
not-for-profit sectors.

Acknowledgement
Many thanks go to Mühlenberg Interiors GmbH for supplying the materials, testing appa-
ratus and advice.

References
1. Hale J. Boeing 787 from the Ground Up, AERO, QTR_04.06, 2006.
2. General Aviation Manufacturers Association (GAMA). Acceptable practices document,
cabin interior monument structural substantiation methods. Washington DC: GAMA,
20 May 2009.
3. Black S. Advanced materials for aircraft interiors. High-performance Composites
November 2006; 24.
4. Mahdi S, Kinlock AJ, Matthews FL, et al. The static mechanical performance of repaired
composite sandwich beams: part I – experimental characterisation. J Sandwich Struct
Mater 2003; 5(2): 179–201.
5. American Society of Testing and Materials (ASTM). International standard test method
for flexural properties of sandwich constructions, ASTM C393-00, West Conshohocken,
PA: ASTM, 10 January 2000.
6. Hart-Smith LJ. Design of adhesively bonded joints. In: Matthews FL (ed.) Joining fibre-
reinforced plastics. London: Elsevier Applied Science, 1987, pp.271–311.
7. Giroux C and Shao Y. Flexural and shear rigidity of composite sheet piles. J Compos
Constr, ASCE 2003; 7(4): 348–355.
8. Euro-Composites Data Sheet. EC536-13e, www.euro-composites.com.
9. Gurit Data Sheet. PHG 600-44-50 and PHG 600-68-50, www.gurit.com.

View publication stats

You might also like