Beruflich Dokumente
Kultur Dokumente
retrospect, such a demonstration was neces- ing resins or adhesives. They create a weak
sary as a public relations exercise to convince interface, with failure occurring at only a frac-
the decision makers to apply the technology, tion of the strength that would have been
where appropriate, on the primary structure developed by grit-blasting. However, the
of large aircraft. Nevertheless, the technology resulting weak bonds cannot be detected by
did not die. The combination of phosphoric conventional ultrasonic nondestructive
acid anodizing for aluminum alloys, a pheno- inspection. Also, most bonded composite
lic-based corrosion-inhibiting primer and structures are so thin that they will not fall
adhesives formulated to resist the absorption apart even if the bond has as little as one-tenth
of water has since been applied very exten- of the strength of a properly processed bond.
sively on the primary structures (wings, Consequently, little has been done about the
fuselages, and tails) of Cessna and SAAB air- problem, in the mistaken belief that doing so
craft with excellent results. would incur an unnecessary expense.
The composites bonding industry could be On the contrary, even if local bond failures
described today as being in the same state that can be tolerated structurally, once they become
metal bonding was in 25 years ago, but with detectable in service they cause expensive
one big difference. The need to properly pre- inspection programs and possibly rework, not
pare the surfaces of composite laminates prior necessarily to add to the strength of the parts,
to bonding is acknowledged only by those but to prevent the flaws from being detected
who have suffered a major problem from hav- and repaired again on subsequent inspections.
ing failed to do so on some previous occasion. Failing to ensure that the glue sticks properly
Ironically, there are many researchers and pro- in the first place is definitely a false economy.
duction engineers worldwide who are aware The delay in solving the corresponding prob-
of the problems and their causes. However, lems on bonded metal structure gave the
there is no highly visible activity like the technology such a bad name in the USA that it
PABST program to draw attention to the was the direct cause of the extensive use of riv-
issues. This chapter cannot be expected to eted structure when bonding would have been
solve these problems, but it is hoped that it lighter, less expensive, and more durable, with
will raise the level of awareness of the subject better than a twentyfold reduction in the inci-
and that its content will help achieve better dence of fatigue cracks and a dramatic
bonded composite structures by providing improvement in damage tolerance. Lest the
standards for inspecting the surfaces before same preference for mechanical fastening
bonding, rather than creating the impossible rather than bonding continue to be followed
situation of trying to detect weak bonds after for thin composite structures, it should be
bonding. What is needed is a method that will noted that the typical fasteners cost about a
ensure a reliable bond every time, using pro- hundred times as much as the simple rivets
cedures to which all manufacturers of used with aluminum structures. Worse, the
composite structures will be willing to adhere. minimum diameter of the specialty fasteners
The first part of this challenge has already for composites is 50% larger than the diameter
been accomplished: light grit-blasting or thor- of comparable rivets. Conventional aluminum
ough mechanical abrasion has been shown to rivets are not used for carbon composite struc-
work. Unfortunately, more often than not, tures because of concern about galvanic
films of adhesive are applied to the surfaces corrosion, between the rivets and the carbon
left by the removal of a peel ply without any fibers, so composite structures that could have
surface treatment. Most, and possibly all, peel been bonded must be locally reinforced to be
plies are known not to create a suitable surface mechanically fastened instead, adding to both
for bonding, at least when used with laminat- the cost and weight.
Historical background 669
There are, then, very great incentives for aerospace industry. It seems significant that,
reliable processing of adhesive bonds in com- today, some factories prohibit the use of bond-
posite structures. There is also no hstory of ing directly to a composite surface created by
properly processed composite bonds failing in the mere removal of a peel ply and that the
service. (This is also true of metal bonding. All automotive racing industry has experienced
in-service failures are the result of environ- the same kind of premature failures with peel-
mental attack, at load levels far too low to ply treatments that the aerospace industry has
have initiated mechanical failure of the bonds. suffered. If what may be called the peel-ply
Testing during the PABST program of coupons bonded joints were twice as strong as they
cut from retired aircraft stored at Davis- actually are, there would be little concern.
Monthan, Arizona, showed that there was no Conversely, if they were only half as strong,
structural deterioration after 20 years.) the problem would have been more widely
There are others that call for improvements recognized long ago and far more done about
in surface preparation for bonding of compos- it. The real problem with peel-ply bonded
ites. Robert Schliekelmann, the famous pioneer joints is that, while quite unreliable in the eyes
of Redux bonding at Fokker, was sufficiently of those closest to the subject, they usually
concerned about the failure to recognize the have sufficient strength to pass ultrasonic
need for proper surface treatment of compos- inspections (because there are no gaps) and are
ites as well as metals to make a plea’ for more sufficiently strong some of the time that the
attention to this issue. Almost a decade ago, joints do not fail prior to delivery of the com-
Douglas engineers prepared an article on the ponent. This makes it difficult to present a
subject2to help the airlines until the repair man- convincing case to non-technically minded
uals could be updated. The article was judged participants in the industry that there is a real
to be of sufficient importance to be reprinted in problem. One non-failure tends to be inter-
the Canadair house journal3.Every experiment preted as complete success.
had to be repeated to create new photographs This chapter cannot possibly provide all the
and every phenomenon was duplicated, answers on the subject of surface preparations
thereby verifying that the problems discussed for composite bonding. It took a $20 million
in the original article really existed. A similar 5 year program to solve the corresponding
concern was expressed in England, where problems for metal bonding. It should also be
Parker and Waghom4 reported on a far more acknowledged that the concerns expressed
comprehensive test program on the effects of here are based exclusively on consistently bad
surface preparation on adhesive bond strength experiences with 180°C (350°F) cured epoxy
for carbon-epoxy laminates. They also con- composites. (The problem may be associated
cluded that ‘composite surfaces must be with the curing of the laminate and the peel
abraded to achieve strong adhesive-bonded ply as much as with the adhesive.) It may also
joints.’ Pocius and Wentz advocated the use of be significant that most, if not all, of these
Scotchbrite pads with embedded abrasive par- problems were associated with nylon peel
ticles as an effective and reliable technique for plies, whose use with a 180°C (350°F) cured-
achieving good composite bonds5. A recent laminate is believed by some polymer
research program6 addressed both thermoset chemists to be undesirable, because the nylon
and thermoplastic composites. Reference 6 also may break down and transfer a thin film of
cites other English research. The problems still low-molecular weight material to the compos-
exist, on both sides of the Atlantic Ocean. ite surface. Photographic evidence of this
A more recent article on this subject7,voiced phenomenon is included here. Such a surface
concerns because of the reintroduction of infe- would be very difficult for an epoxy adhesive
rior techniques throughout much of the to wet, because of too low a surface energy.
670 Surface preparationsfor ensuring that the glue will stick in bonded composite structures
The polymer chemists express less concern peel plies. Not one in-service delamination has
about the use of nylon at 120°C (250"F), but occurred in those composite structures pre-
some adhesion failures of glued joints have pared for bonding by the grit-blast technique.
occurred with those materials also. The domi- It is hoped that this chapter will inspire
nant factor may be the low surface energy those who believe that their structural adhe-
level of the composite cured against a slick sive bonds are both strong and reliable to
peel ply, or one contaminated by a release assess composite surfaces prior to bonding. (A
agent to ensure that it can be removed easily request for information from a major US aero-
without fracturing (damaging) the composite space manufacturer about a peel ply different
laminate. from those used at Douglas resulted in their
It is certain that the criteria by which manu- switching to another peel ply. Microscopic
facturers evaluate their peel plies are precisely examination of the surface to be bonded made
the opposite of those that someone else trying it clear to them that one of their problems
to promote adhesion in a bonded joint would could be tied directly to the choice of a new
follow. Some peel plies are even coated with peel ply. Once alerted, they checked all related
release agents to ensure that they can be procedures and even changed one of their old
removed easily without damaging the basic and trusted peel plies once its true capabilities
laminate. Silicone transfer has been observed had been exposed.)
with some peel plies supposedly totally free of This chapter concentrates exclusively on the
contamination. What is needed to create a issue of surface preparation. Obviously, one
bondable surface is a tear ply that will remove must also follow correct thermal and pressure
a small amount of surface resin over the entire profiles during cure to achieve a properly
area to be bonded. However, peel plies known cross-linked adhesive that has a chance to flow
to be capable of achieving this are almost and wet the surface to which it is to be
invariably shunned because they are so hard bonded. Prebond moisture, in both laminates
to peel off, because they break and inevitably and adhesives, is also known to be a major
lead to local contamination as the process is cause of weak bonds. These and other impor-
restarted, or because there is concern that the tant issues are discussed in References 7 and 8.
interior of the laminate will be damaged. This does not imply that they are any less
It cannot be asserted on purely technical important.
grounds that there are no circumstances under We now present electron-micrographs of
which some peel plies will produce an ade- glue that failed to stick to the composite, com-
quate surface for adhesive bonding. It stands posite surfaces to which the adhesive did not
to reason that some users of this approach stick, similar surfaces to which the glue will
have been spared the problems Douglas and not stick, surfaces to which adhesive is known
others have encountered, or the products to have stuck in the past, and surfaces to
would have all been removed from the mar- which it is hoped it will stick. The effectiveness
ketplace. However, it is likely that still others of light grit-blasting as a reliable surface treat-
have also suffered, but are unaware of the ment has already been established. It was used
cause of their problems. Nevertheless, the on all bonded joints of the Lear Fan all-com-
argument against peel plies is irrefutable on posite aircraft9, creating bonds stronger than
economic grounds. The cost of investigating the surrounding structure even when half the
weak bonds has been far, far greater than overlapping area was disbanded because of
would have been incurred if Douglas had inappropriate solid-rubber tooling used for
insisted that all suppliers and subcontractors the fuselage skin splices. The bond to the grit-
lightly grit-blast or thoroughly hand-sand blasted surface was so strong, where the
composite surfaces following removal of the mating surfaces were brought properly into
The problem - weak bonds 671
contact, that 100%bonding was not necessary. area with a different texture is part of the sur-
Grit-blasting is used today on some of the face ply of unidirectional carbon-epoxy left
most trouble-free composite components on when the remainder separated. The reason
the MD-11 aircraft; specifically, the compo- why this very small segment adhered is
nents made in Japan. However, some unknown. It is almost as if there were one
manufacturers of bonded composite structures small drop of chemical acting like a wetting
will prefer to continue to use peel-ply-only agent on the composite when it was bonded.
surface preparations, no matter how weak or If this should prove to be the case, and
unreliable the resulting bonds are. One must research can identify an agent capable of pro-
find different peel plies that do not release moting the wettability of cured epoxy in a
cleanly, but which take some of the surface laminate by uncured epoxy in an adhesive
resin with them, without damaging the layer, it would be a giant contribution to com-
remainder of the matrix, or find a coupling posites technology. The idea of a coupling
agent to improve the behavior of what are agent, equivalent to the silanes used for
known today to be unsatisfactory peel plies no epoxy-bonding of aluminum alloys during
matter how widely they may be used, or trans- repairs, is apparently feasible for composites
fer the production to factories in which reliable also - at least when the peel ply has not been
surface-treatment practices are followed. coated with a release agent. Coupling agents
would be more likely to work if the basic lam-
inate were not fully cured prior to bonding.
29.3 THE PROBLEM - WEAK BONDS
One would prefer an incomplete initial cure in
No more convincing proof of the existence of order to leave some active chains in the mole-
weak bonds created on peel-ply composite cular structure of the composite to which the
surfaces can be found than in Fig. 29.1. This is glue could bond. The bond cycle would then
a scanning electron microscope (SEM) pho- be relied upon to complete the cure of the
tomicrograph not of a composite surface after laminate and this process might affect the
the peel ply was lifted off, but of the cured choice of adhesive. However, such an out-
adhesive, showing the imprint left by the com- come must be preceded by an
posite laminate after it peeled off! The small acknowledgment that the remainder of
Fig. 29.1 and the large similar surrounding
area imply the existence of a problem.
Figure 29.2 shows the same lack of adhesion
evident in Fig. 29.1 at a small magnification
over a very much larger area, to show how
widespread this phenomenon can be. The tex-
ture of the peel ply is clearly imprinted in the
glue over almost the entire area. The different
texture visible on the left side of the figure is
the peel-ply imprint on the underlying com-
posite laminate. In this area, the adhesive
failed to bond to the lower surface.
Throughout everywhere else shown in Fig.
29.2, the adhesive failed to bond to the other
composite part, the peel-ply imprint of which
is embossed on the adhesive. Figure 29.3
Fig. 29.1 Imprint of peel ply, in background, on shows an adjacent area on the same-part. The
cured adhesive film that failed to adhere (mag x30). thick glue layer (shiny area) at the bottom left
672 Surface preparations for ensuring that the g2ue will stick in bonded composite structures
Fig. 29.9 Lightly grit blasted composite surface, Fig. 29.10 Highly magnified grit-blasted epoxy sur-
retaining imprint of peel ply (mag X50). face, showing no damage to underlying fibers (mag
x1000).
I
Fig. 29.17 Same sample at higher magnification
Fig. 29.15 Peel-ply imprint on epoxy surface to (mag x750).
which the glue will not adhere (mag X38).
Fig. 29.20 Moderately sanded peel-ply imprint on Fig. 29.21 Highly magnified surface, showing how
surface of woven composite laminate, showing how most of the peel-ply imprint remains after hand
sanding does not abrade the entire surface (mag sanding (mag ~1000).
x100).
improvement over the few percent removed in
the sample shown in Fig. 29.18.
While it must be acknowledged that both
grit-blasting and hand-sanding can be over-
done, doing so takes time and effort if one is
using the right abrasives and equipment. A
significant loss of strength from such actions is
far less likely than from either simply remov-
ing a peel ply or sanding the composite
surfaces far too lightly.
To cover the possibility that the bad experi-
ences with peel plies at Douglas might have
been associated exclusively with the break-
down of nylon at too high a curing
temperature, samples were obtained from
Oxford Brookes University, which used a dif-
ferent laminating resin and both a polyester
peel ply and what is probably the same nylon
peel ply. The polyester peel ply was noticeably
more difficult to remove than the nylon peel
ply, but far easier than what were referred to
as tear plies in Reference 2. The surface created Fig. 29.22 Imprint of polyester peel-ply, showing
by removing the polyester peel ply from a more fractured resin than with nylon peel-plies
120°C (250°F) cured carbon-poxy laminate (mag ~ 5 0 ) .
Samples of diferently prepared surfaces for bonding 681
I 1
--- -1
removal of a tear ply that adhered to the the production level in over 20 years. Because
matrix as tenaciously as conventional peel the metal-bond problem was resolved so
plies refuse to adhere. This, in turn, raises quickly, very few panels were involved. There
questions about the feasibility of tear plies as a were no in-service failures because every
surface-preparation technique since the lami- affected panel was identified before delivery
nate is damaged far more than by light and reinforced by rivets.
grit-blasting, as a comparison between Figs. Peel-ply ’surface preparation’ for bonding
29.10 and 29.27 will attest. of composites has not been as thoroughly
explained. There have been many instances of
such weak bonds not being detected until they
29.5 COMMENTARY
had split apart in service, even though there is
In the mid-l980s, a major effort was aimed at no reason to believe that the bonds wore out
the resolution of a problem associated with under mechanical loads. The observed modes
adhesive bonding of aluminum alloy struc- of failure are consistent with a manufacturing
tures. In that case, as in the problem discussed problem. There is little doubt that gentle grit-
here, the glue failed to stick and interfacial fail- blasting is the most reliable method of
ures resulted. However, there remains a very preparing thermoset composite surfaces for
great difference between the two cases. In the bonding. Sanding can work only on fairly fine-
earlier case, every factor associated with the weave cloths in composite laminates.
weak bonds that passed all ultrasonic inspec- Otherwise, it is not possible to sand to the bot-
tions was identified in a matter of weeks. The tom of all depressions left by the peel ply
conditions had been replicated in a laboratory without also sanding significantly into the
and the investigation closed within a few structural fibers. A few other manufacturers
months. Significantly, there were no loose have used tear plies, which leave a completely
ends and the problem has not recurred. The fractured resin surface, rather than the more
primary cause of the problem was condensate commonly used easily removed peel plies dis-
on adhesive film that had been removed from cussed here. More use should be made of tear
storage before it had thawed out. plies. However, if a tear ply were used on a
However, there was a second factor lightweight honeycomb or foam sandwich
involved as well. The first violation of proper panel, there is a good chance that the core
processing procedures would not cause a would fail instead.
defective bond unless the moisture was There can also be no doubt that when prob-
trapped at the interface between the details. lems have occurred as a result of bonding
This condition happened consistently with directly to the surface left by removal of a peel
one kind of bonding tool, while similar parts ply, or one with totally inadequate sanding,
made concurrently on a different kind of tool they are widespread and serious. These prob-
showed no such problems. The second kind of lems have been experienced at many places.
tool permitted complete ventilation of any Yet, if they happened as repeatedly as the
trapped volatiles as well as of any generated defects caused by the combination of the two
during the cure. There were no large area factors cited above, one would have expected
defective bonds. Both tools produced local such a backlash against the procedure that
bond defects where a tool or the parts were out every unreliable peel ply would have been
of contour. These problems were eliminated withdrawn from the marketplace, preventing
by correcting the tools and by better straight- any recurrence of the problems. That has not
ening of the stiffeners. This problem was happened so, presumably some organizations
resolved very quickly, whereas the present are able to bond successfully to composite
problem with peel plies has not been solved at structures by simply removing a peel ply.
684 Surface prqarations for ensuring that the glue will stick in bonded composite structures
Regrettably it seems as if anyone doing this for bonding has been found to be extremely
process successfully has no reason to investi- reliable in service. While one needs to buy
gate why he is successful, so the differences appropriate equipment, the cost of doing so is
between his techniques and those that lead to a small fraction of the typical cost of even one
trouble have remained unidentified. composite detail. With the right equipment,
It seems that there is something else training of technicians is straightforward.
involved as well possibly prebond moisture or There can be no valid argument in favor of
something associated with venting during the not mastering the art of grit-blasting. While
cure or possibly the breakdown and transfer of there may be choices for the surface prepara-
nylon peel plies cured at too high a tempera- tion during initial fabrication, peel plies
ture or the transfer of a release agent on some cannot possibly be stripped off a second time
peel plies. Weak interfacial bonds between the to create a new 'clean' surface during repairs.
adhesive layer and both composite adherends It may be that some manufacturers make
should never have been strong enough to rip a successful bonds to peel-ply composite sur-
properly cured adhesive back and forth from faces already. It may also be possible that
one interface to the other in the manner that is coupling agents may be found to enable others
so evident in Fig. 29.8. And, more significantly, to do the same. However, it is undeniably true
attempts to replicate the weak bonds in the that, for at least two decades, some manufac-
laboratory have been inconclusive and, at turers of composite components who have
times, inconsistent. This merely reinforces the relied on peel plies alone as surface prepara-
assumption that the problem is not yet fully tion for bonding have created weak bonds that
understood without, in any degree, diminish- have fallen apart in service. The cost of unan-
ing the conviction that the problem is serious ticipated repairs and investigations has greatly
and needs to be resolved. exceeded any expected initial cost savings.
It would be helpful if there were a reliable Given the widespread nature of these prob-
a peel-dominated quality-control test for com- lems, it is appropriate to recommend that the
posite surface preparation that was use of peel-ply surface preparation alone be
equivalent to the wedge-crack test used for discontinued unless it can be shown that its
metal bonding. Unfortunately, the experi- use never results in interfacial failures
ments performed to date have been between the composite and adhesive layers. It
bedevilled with extraneous influences that are is quite clear, from experience, that the cost of
not yet understood but have a much greater even one in-service bond separation exceeds
effect on the strength of the coupon than the by far the savings derived during an entire
variations in surface treatment. Neverless the production run by not thoroughly abrading
goal remains. the surface.
Scanning electron microscope images,
such as those presented here, show easily dis-
29.6 CONCLUSIONS
tinguishable differences between the matt
As acknowledged above, more technical infor- rough surface created by mechanical abrasion
mation needs to be uncovered about adhesive and the microscopically smooth furrows
bonding to fiber-polymer composite surfaces associated with peel-ply removal alone. It is
created by simply removing a peel ply. The not difficult to ensure that the surface prepa-
weak bonds associated with this technique are ration is adequate for bonding or painting
a financial burden that is both serious and easy before the bond has been made. Conversely, it
to avoid. is extremely difficult to detect a weak bond
The use of low-pressure grit-blasting as the nondestructively once the error has been
final step in the preparation of these surfaces made.
References 685