Engineering Failure Analysis 14 (2007) 321–348

www.elsevier.com/locate/engfailanal

Failure prevention in bonded joints on primary load

bearing structures
Gary Savage *

BAR Honda Formula 1 Racing Team, Operations Centre, Brackley, Northants NN13 7BD, UK

Received 20 November 2005; accepted 20 November 2005

Available online 16 May 2006

Abstract

Adhesive bonding is a particularly effective method of assembling complex structures, especially those made from dif-
ferent materials. Provided the joint is well designed, the adhesive bond ought to be one of the strongest aspects of the struc-
ture and most certainly should not be the life limiting factor. This of course pre-supposes that the joint has been correctly
executed. The major factors determining the integrity of an adhesive bond are selection of the most appropriate adhesive,
joint design, preparation of the bonding surfaces, strict quality control in production and condition monitoring in service.
Adhesives have become increasingly important in assembling many of the multi-material structures which make up a con-
temporary Formula 1 racing car. Structural fibre reinforced composite materials were introduced into Formula 1 motor
racing in 1980. Since then the cars have become increasingly dependent on these materials such that the weight of a
state-of-the-art chassis may consist of up to 80% of carbon fire reinforced epoxy resins and the appropriate adhesives
to facilitate fabrication. One might argue, therefore, that a modern Formula 1 chassis consists of a series of ‘‘plastic’’ moul-
dings held together with glue! Many of these structures are very highly stressed and required to operate in aggressive envi-
ronments, particularly high temperatures. The consequences of failure of ‘‘Class A’’ structures can be catastrophic to the
operation of the vehicle and impinge on the safety of the driver. Although increasingly better understood, the science and
engineering of adhesives is very much in its infancy. Consequently, the design and operation of bonded components tends
to be a constantly evolving, semi-quantitative process that combines fracture and finite element analysis with practical
experience. In recent years BAR Honda have been very successful using and developing adhesive technology. The process
of ‘‘total quality management’’ (TQM) encompasses the whole operation from R&D and design through materials pro-
curement, component manufacture and condition monitoring to ensure successful exploitation up to the point of with-
drawal from service.
Ó 2006 Published by Elsevier Ltd.

Keywords: Adhesive bonding; Total quality management; Condition monitoring; Best practice

*
Tel.: +44 1280 844 000; fax: +44 1280 844 001.
E-mail address: gsavage@barhondaf1.com.

1350-6307/$ - see front matter Ó 2006 Published by Elsevier Ltd.

doi:10.1016/j.engfailanal.2005.11.013
322 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

1. Introduction

Adhesives are being increasingly employed in the assembly of complex components within the Formula 1
industry. In particular they are used to replace or augment more traditional joining techniques such as weld-
ing, mechanical fastenings and interference fits, etc. The driving force for the use of adhesives is the widespread
use of composite materials, although this has precipitated an increasing number of other applications [1,2].
Adhesives may be quickly applied, possess excellent properties and are very cost effective. The many advan-
tages of adhesives include:

1. The ability to join dissimilar materials.

2. A continuous bond is formed such that stronger and stiffer structures are often produced (Fig. 1).
3. A more uniform stress distribution is achieved on loading, avoiding local stress concentrations
(Fig. 2).
4. Reduced weight and part count.
5. Small areas may be bonded accurately and large areas may be bonded without inducing stresses.
6. Little, if any, finishing is required.
7. With careful joint design, excellent fatigue resistance and vibration damping can be achieved.

It is very rare for the failure of a bonded joint to involve adhesive strength. Failures are generally due to
poor design, inadequate preparation and poor production procedures, therefore, the successful exploitation of
adhesives necessitates a thorough knowledge of all aspects of the materials science of the joint as a complete
system. A large variety of products are available which may be engineered to suit individual applications. Nev-
ertheless, the selection of adhesives can be overwhelmingly complicated due to this enormous diversity.
Among the limitations of adhesives which must be considered are the following:

Fig. 1. Stiffening with adhesively bonded joints.

Fig. 2. Stress distribution in a loaded joint.

 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 323

1. Durability is unpredictable and must be determined experimentally for each application.

2. Surface preparation is critical to the success of the operation and complete wetting of the substrate must be
ensured.
3. Identification of defects by NDT requires a great deal of skill and experience and given the present level of
understanding the tolerance to defects is impossible to predict in anything but a semi-quantitative manner.
4. Increasing the service temperature decreases the bond strength.
5. Short-term handleability is poor and bonded structures are often difficult to dismantle for in-service repair.
6. Environmental resistance depends on the integrity of the adhesive.
7. New and unfamiliar production and exploitation controls must be developed and implemented.

That aside, even materials which are traditionally difficult to join can be bonded with adhesives, although
some substrates may give rise to lower bond strengths or limited durability.

2. Defining the adhesive bond

Adhesion is defined as the holding together of two surfaces by interfacial forces that will resist separation [3].
Good adhesion requires very close contact. For an adhesive to bond it is required to flow and wet the surface
of the substrate. The surface energy (surface tension for liquids) of the adhesive must therefore be lower than
that of the substrate to ensure good wetting, the first stage in forming an adhesive bond. The second stage is
the generation of intrinsic adhesion forces across the interface. The various forces which may be acting are
known as the mechanisms of adhesion. There are three primary mechanisms of adhesion proposed (Fig. 3).

1. Mechanical interlocking or ‘‘keying’’ occurs when cured adhesive becomes trapped in the irregular surface of
the substrate. It is, however, possible to show that the attainment of good adhesion is possible between opti-
cally smooth surfaces [4,5]. The frequently observed increases in measured joint strengths with increasing
substrate surface roughness are considered, therefore, attributable to other factors such as the removal
of weak surface material, improved wetting and a larger exposed surface area for bonding.
2. Adhesive and substrate may diffuse across the interface. Similarly the adhesive may diffuse into the surface
of porous substrates and cure. The interdiffusion of the polymer chains of the adhesive requires the adhesive
and substrate to be mutually soluble. Such conditions are thought to occur in the solvent welding of poly-
mers but are not considered capable of contributing significantly to the intrinsic strength of the bond. Inter-
diffusion can also be promoted in polymer/metal bonds via pores and other surface defects in the
substrates, however, this effect tends to enhance the bond by promoting adsorption of the adhesive rather
than by the diffusion process.

Fig. 3. Mechanisms of adhesion.

324 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

3. Intimate contact between adhesive and substrate result in adsorption, electrostatic and molecular attraction
processes established between the atoms and molecules in the surfaces of the adhesive and substrate. The
most common of these are van der Waal’s forces and, to a lesser extent, hydrogen bonding. These are
referred to as ‘‘secondary bonds’’. Additionally, chemical bonds may sometimes be formed across the inter-
face. The establishment of ionic, covalent, or metallic bonds is known as ‘‘chemisorption’’ or ‘‘primary’’
bonds. The terms primary and secondary are fairly arbitrary descriptions of the relative strengths of their
attractive forces.

Under specific circumstances, any or all of the mechanisms may be responsible for the adhesive strength of
a particular joint. For the majority of engineering applications, however, it is the adsorption mechanism that
dominates. Providing, therefore, there is intimate and continuous molecular contact at the interface of adhe-
sive and substrate, a bond will be established due to the interatomic forces between the materials.
Adhesive strength is described as the force required to pull the adhesive cleanly away from the surface of the
substrate. A cured adhesive, in common with any other material, can also be characterised by its internal
strength. Similarly, the third factor influencing the strength of the bond is the internal strength of the sub-
strate(s). The term cohesive strength of adhesives and substrates is used to differentiate from adhesion
(Fig. 4). Preferred design practice is to ensure that the life-limiting factor of the joint is the cohesive strength
of one or other of the substrates as this can be more accurately defined and guaranteed. During the execution
of a bond, adhesives typically follow a flow phase when they are applied, spread and wet the surface, followed
by a hardening (cure) phase during which their cohesive strength develops.
Structural bonding is the term used to define a bond where it performs a load bearing function. This allows
the forces within a structure to be transmitted from one member to another through the joint. The purpose
of the adhesive is to facilitate this load transfer. Despite an ever increasing amount of research aimed at defin-
ing the properties of the adhesively bonded joint, the majority of the data is qualitative at best, necessitating
extensive testing and prototype evaluation.

3. The use of adhesives in Formula 1 car construction

There are three main types of adhesive in widespread use within Formula 1.

3.1. Epoxies

Epoxy resins are characterised by the possession of two or more 1, 2 epoxy functional groups per molecule
(Fig. 5). The three membered rings react in such a way as to ensure no volatile formation and very little shrink-
age during polymerisation. Epoxy adhesives consist of an epoxy resin plus a hardener. They allow great ver-
satility in formulation as there are many resins and many different hardeners. The adhesives can be used to
join most materials provided the correct procedures are followed religiously. Epoxies form very strong bonds

Fig. 4. Forces operating in an adhesive bond.

 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 325

Fig. 5. Chemical structure of a typical epoxy.

with a wide number of substrates and are particularly useful in joining metal/metal and metal/polymer struc-
tures. They do, however, possess low peel strengths (which can be as much as one or two orders of magnitude
lower than their shear strengths) and low flexibility. Epoxy adhesives are available in one and two-part thixo-
tropic pastes and film form. Some (though by no means all) two part adhesives may be cured at room tem-
perature whereas the single part and film systems tend to be heat cured. As a general rule of thumb, the
higher the curing temperature of the epoxy the better its durability at elevated temperatures will be, although
this may be at the expense of low temperature brittleness.
The first major application of bonded structures on Formula 1 racing cars was in aluminium skinned, alu-
minium honeycomb materials which were used to fabricate chassis in the late 1970s. The ‘‘tubs’’ were formed
from pre-bonded sheeting which was routed, folded and riveted into the appropriate shape (Fig. 6). The var-
ious teams involved later pre-formed the skins prior to bonding to the core using an epoxy film adhesive. This
technology was superseded in 1980 by the use of fibre reinforced composite skins [6]. Composite technology
was introduced simultaneously by the McLaren and Lotus teams. The original carbon composite chassis dif-
fered very little in outward appearance from the soon to be obsolete aluminium tubs with the exception that
the skins were black rather than silver (Fig. 7). The Lotus chassis followed the previous ‘‘cut and fold’’ meth-
odology simply by replacing the pre-bonded aluminium skins with a hybrid composite of carbon and Kevlar
reinforced epoxy. The McLaren chassis on the other hand were a considerable departure from tradition. They
consisted of a pseudo-monolithic arrangement laid up over a ‘‘male’’ mould or mandrel using unidirectional
(UD) preimpregnated carbon fibre tape (‘‘prepreg’’). The mandrel, made of cast and machined aluminium
alloy, was dismantled for removal through the cockpit opening following an autoclave cure of the composite.
A three stage cure was required: one for the inner composite skin, a second to cure the epoxy film adhesive
which attached the honeycomb core and a third for a further adhesive layer and the structure’s outer skin.
The film adhesives have improved considerably with the passage of time. Whilst the shear strength of the bond
is the primary design consideration, a number of secondary properties are of importance to ensure the overall
tolerance of the structure to the localised impact damage, to which they are periodically subjected. Particular
advances have been made in the toughness and peel strength of the cured adhesives (Fig. 8) which have sig-
nificantly contributed to the integrity of the various composite structures.
Whilst the ‘‘male moulded’’ method provided a means of producing all of the external panels of the chassis
in one piece, it placed severe restrictions on the shapes which could be moulded. A major advance, which
remains the technique in use today, was to fabricate the chassis inside female composite tooling with a

Fig. 6. Aluminium honeycomb composite chassis made by the ‘‘cut and fold’’ method.
326 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 7. McLaren MP4/1, the first carbon fibre reinforced composite F1 chassis.

Fig. 8. Testing the peel strength of a honeycomb structure.

combination of UD and fabric reinforced prepregs. This method of manufacture necessitates a join in the
main shell. The two halves are bonded together using a thixotropic paste adhesive. Female moulding makes
far more efficient use of the available aerodynamic envelope since only a minimum of secondary bodywork is
needed to cover it. It also provides an opportunity to optimise the geometry and therefore improve its struc-
tural efficiency. The traditional approach to bonding the chassis joins the two halves top and bottom rather
like a kayak (Fig. 9). In the past some teams have joined their chassis ‘‘front to rear’’ with the bond about the
dash bulkhead, whilst others have and indeed still do, produce the majority of the tub in one piece, bonding on
the floor panel to complete the structure. The choice of joint geometry is more a matter of preferred ergonom-
ics for a particular team rather than for any performance benefit. There are advantages and disadvantages
associated with each configuration but, properly executed, the adhesive joint will be one of the strongest
aspects of the chassis irrespective of the philosophy adopted. Formula 1 teams aim to use composite materials
and bonded structures wherever they offer advantages in mechanical properties or a reduction in weight and/
or complexity of design. One of the most exciting applications of adhesives was in the production of composite
suspension members. These were originally introduced by McLaren [7] following an extensive period of testing
and development but are now used by all of the teams (Fig. 10). The most recent innovation on an F1 car is
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 327

Fig. 9. Bonding an F1 chassis.

Fig. 10. Bonded composite suspension components.

that of a composite gearbox [8]. This device consists of a monolithic carbon fibre reinforced epoxy ‘‘maincase’’
into which are bonded a number of metallic and composite bulkheads to carry the ‘‘internals’’ (gear cluster
and selectors, etc.) and ‘‘inserts’’ such as suspension and engine mounts. This device illustrates very graphically
the strength and versatility of adhesives, being a multi-material structure capable of transmitting in excess of
900 bhp whilst reacting to severe suspension loads, all at temperatures of up to (and sporadically exceeding)
150 °C (Fig. 11).

3.2. Anaerobics

Anaerobic adhesives are single-component materials which cure (or solidify) at room temperature when
deprived of contact with oxygen [9]. The curing component in the liquid remains inactive for as long as it
is in contact with atmospheric oxygen. If the adhesive is deprived of oxygen within the mating parts of a bond
the curing occurs rapidly. The capillary action of the low viscosity liquid adhesive allows it to fill any gaps in
328 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 11. Composite gearbox.

the joint. They are often referred to as ‘‘locking compounds’’, being used to secure, seal and retain turned,
threaded or similarly close fitting parts. Anaerobic adhesives are based on synthetic acrylic resins which are
characterised by a very high shear strength, good temperature and solvent resistance. Manufacturers such
as Loctite produce a large range of these adhesives. They are generally all cured in the same manner with
the various differences designed to improve their usefulness in more specific applications. Examples of this
would include variations in strength to allow assemblies to be dismantled, different temperature operating
regimes, differences in viscosity to suit joint design, resistance to aggressive environments and the addition
of modifiers such as toughening agents, etc.
Anaerobic adhesives used alone are often stronger than mechanical fixings and will also substantially
improve bond strength when used to augment existing methods. When a bolted assembly is distorted the com-
ponents move relative to one another and this can occur when parts are heavily loaded or subjected to vibra-
tion. Should such a relative movement occur, a thread will loosen unless an effective locking system is
employed. It is now commonplace to use a number of anaerobic adhesives to lock and seal threaded compo-
nents and retain cylindrical parts. The advantages of thread locking adhesives are that they are vibration
proof, single part, easy to apply, seal and lock in any position allowing for a controlled torque and, therefore,
eliminate over tightening. They can be obtained in grades of different strength to enable disassembly when
required and facilitate any fastener becoming a locking fastener. Retaining adhesives can be used to fit bushes
and bearings using a slip rather than a press fit and can, therefore, prevent damage to the softer bearing hous-
ing (Fig. 12). The adhesive seals and prevents relative movement at the joint thus eliminating fretting.
The simultaneous locking and sealing of steel studs into the titanium inserts of a composite gearbox
(Fig. 13) obviates the use of mechanical fasteners. In order to work, mechanical devices are required to expand
within the inserts as the studs are tightened in. This method would generate stresses capable of inducing crack-
ing thus reducing the life of the box and could, in extreme cases, cause a catastrophic failure during service.
Application of a retaining adhesive to secure a steel bearing into a metal matrix composite steering arm per-
mits the bearing to be slip rather than interference fitted. It can, therefore, be removed and replaced when
worn out without damaging the bearing housing. Prior to the adhesive, too tight an interference could distort
the bearing and pressing it out could increase the size of its housing rendering the whole component useless.
Bearings in the suspension rockers (Fig. 14) have traditionally been held in place with an interference fit of
0.025 mm. Under certain loading conditions it is possible for the push rods to generate out of plane loading
sufficient to exceed the strength of the fit. Combining the interference fit of these components with an adhesive
provides a reserve factor far in excess of any loads which could possibly be generated by the car under normal
operating conditions.
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 329

Fig. 12. Soft housings are easily damaged by interference fit bearings.

Fig. 13. Steel studs bonded into titanium gearbox insert.

Fig. 14. Augmenting interference fit with an adhesive.

3.3. Cyanoacrylates

Cyanoacrylate adhesives, known colloquially as ‘‘super glues’’, cure through reaction with the alkalinity, in
the form of moisture, held on the surfaces to be bonded [9]. As they have limited gap filling ability they require
close fitting joints and generally cure in a few seconds. This speed of action is of particular value since it allows
the joining of intricate parts without the need for complex jigs and fixtures. Whilst not considered to form very
330 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

strong bonds compared to epoxies for example, cyanoacrylates are particularly suited to small polymeric com-
ponents and elastomers. Once again, the various manufacturers supply a large range of products with subtle
variations for specific applications. This includes variations in viscosity to aid processability and blends con-
taining rubbers and/or polyurethanes to overcome brittleness. Cyanoacrylates are extensively used to make
rapid modifications to wind tunnel models during the development of the cars. They are also particularly use-
ful in making custom elastomeric seals.

4. Joint design

When considering an adhesively bonded joint it is important to appreciate that the optimum design will not
be the same as for traditional fixing methods such as welding or mechanical fasteners. The characteristics of
adhesives dictate that their mechanical response to compression and shear loading regimes is far superior than
their tensile response. Peel and cleavage strengths of an adhesive may be up to two orders of magnitude lower
than the shear strength and such loading must be avoided at all costs (Fig. 15). It is paramount that a com-
ponent be designed such that the bond will be strengthened by the geometry of the final artefact rather than be
weakened or destroyed by it. The strength and durability of a joint is a complex function of the stress concen-
trations set up by the applied loads and operating conditions. In a simple lap joint made from thin metal sheets
there are two types of stress: shear and peel. The shear stress varies along the length of the joint with concen-
trations at the ends. The peel stress acts at right angles to the lap joint and is also maximised at the ends
(Fig. 16). The peel stress tends to distort the joint and consequently weaken it. Similarly any deflections of
the structure under load increase the peel component and may lead to premature failure. Unsupported lap
joints are one of the weakest configurations and are seldom used in practise. A number of possible bond geom-
etries, both good and bad, are illustrated in Fig. 17.
Such configurations can be applied, either in isolation or combination, to more complex geometries. The
chassis in Fig. 9 for example uses a ‘‘tongue and groove’’ joint, which is in fact an example of the ‘‘double
lap joint’’. The best results are clearly achieved by designing the joint such that the forces experienced tend
to compress the adhesive or induce shear loads whilst taking steps to eliminate or at least minimise peel
and cleavage forces, particularly at the vulnerable edges of the joint. Fig. 18 shows a simple but effective ‘‘dou-
ble strap joint’’ used to join two carbon composite tubes using a third, aluminium, tube as the strap. Fig. 19
illustrates the practical application of this joint to produce a bespoke racing kayak paddle. The joint not only
facilitates assembly of the paddle, it enables both its length and the angle of the blades to be set exactly to the
athlete’s preference using an inexpensive assembly jig. It should be noted that the kayak itself is produced
using the same joint to assemble the top and bottom sections. In this case two thin carbon fibre composite
mouldings are butted together and strapped either side with a thin Kevlar composite tape.

Fig. 15. Types of loading endured by adhesive joints.

 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 331

Fig. 16. Stress distribution in a simple lap shear joint.

Fig. 17. Basic joint geometries – in practice two or more may be used in conjunction.

Fig. 18. Double strap joint.

The performance of structural adhesive bonds is observed to depend significantly upon the thickness of the
adhesive layer, known as the ‘‘glue gap’’. Too thin a glue gap will generally result in poor adhesion due to non-
uniform wetting. As the joint is stressed it is possible to induce a bending moment which will cause the joint to
effectively ‘‘rotate’’ increasing the peel component. Good joint design will minimise this effect to some extent
by stabilising the deflection of the joint under load, but thicker bond lines will have the effect of increasing the
peel stress component. Furthermore, thick glue gaps, whilst attractive in terms of ease of assembly, are not
practical due to the impossibility of executing them without inducing intolerable levels of flaws and porosity.
332 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 19. Bonded racing kayak paddle.

Bonds are observed to be most efficient in the thickness range 0.1–0.3 mm. The BAR Honda standard is
0.25 ± 0.05 mm. Experience has shown that optimum wetting and relief of the stress concentrations at the
edges is achieved by lightly tapering the bonding faces and, wherever possible, machining them to the required
tolerance. It is advisable to jig the assembly in such a way as to maintain a uniform bond line. It goes without
saying that the design should maximise the area over which the load is distributed. Consideration must also be
given to practical aspects such as ease and reproducibility of assembly. Non-stressed components may be
designed such that the adhesive bonds are self jigging but structural bonds must be assembled on purpose built
jigs to ensure correct alignment throughout the curing process.
Estimation of the theoretical strength of a joint is relatively straightforward using data which may be
obtained from the adhesive manufacturer or (preferably) determined experimentally (Fig. 20). A number of
tests are described in various testing ‘‘standards’’ and proposed in academic literature (it should be noted that,
as with composites, there are no universally accepted standards for the testing of adhesive joints). Due to the
ease of specimen preparation the easiest test to carry out uses the single lap joint configuration. It is advisable,
however, to use the double lap shear joint as the results are less prone to variability due to deflection of the
coupons under load. To a first approximation, the load capacity of a joint is simply the product of the bonded
area and the shear strength of the adhesive. In practical applications, a number of variables must be factored
in. These include substrates and surface finish, bond thickness (clearance), temperature, joint geometry and
environment, etc. Calculations must be verified by testing in the laboratory. This programme is particularly
important in the eventual success of the application since it enables the identification of potential problems
at an early stage. A variety of tests need to be considered in order to determine any possible problems and
achieve the following objectives:

1. Compare the properties and suitability of a group of adhesives.

2. To act as a quality check for a batch of adhesives.
3. Verify the effectiveness and repeatability of the preparation techniques to be used.
4. Measure the environmental effects (temperature, fluids, moisture, etc.).
5. Provide quantitative data on the joint and adhesive.
6. Analyse failure mechanisms.

Some important criteria of an adhesive joint may be determined by visual observation of the parts following
a failure during a test. It is important to determine whether adhesive or cohesive failure has precipitated over-
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 333

Fig. 20. Evaluating the strength of an adhesive joint in a double lap shear test.

Fig. 21. Failure modes in adhesive bonds.

all joint failure, or if the bonded parts (substrates) have been destroyed (Fig. 21). In the case of adhesive failure
the adhesive is observed to be completely separated from the face of one substrate. This is the ‘‘worst case’’
scenario in that the weakest aspect of the joint is the boundary layer between the bonded parts and the adhe-
sive. This means that the joint is prone to failure without warning and at a load far lower than predicted. In
such circumstances the causes are either that the material is unsuitable for bonding or the bonding surface was
contaminated and therefore incorrectly prepared. In both cases the strength can be increased by devoting more
attention to the pre-treatment of the surface. Cohesive failure occurs when the adhesive itself breaks. The fail-
ure is characterised by remains of the adhesive being found on both substrates. This occurs because the adhe-
sive is over stressed through external action such as temperature, ageing or off-axis loading due to
misalignment or deflection, etc. It can be remedied by design changes in the bonding geometry, tightening
of production procedures, and/or selection of an adhesive which is more suitable to the fine detail of the
334 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

application. The preferred failure mode is substrate failure. In such instances the bond is not the life-limiting
factor of the joint and should not therefore influence its long-term durability. It is recommended that all
stressed joints are designed such that ultimate failure occurs in the substrate.

5. Surface preparation

It is widely accepted that pre-treatment of an adhesive joint is necessary to attain optimum performance.
The surface thermodynamics of the majority of materials used on the race cars are extremely favourable to the
adhesive bonding process and those few which are not may be treated with a primer with a modicum of suc-
cess. This presupposes of course that the bonding surfaces are free from contaminants. The primary aim of
surface preparation is the removal of weak boundary layers. Metal substrates for example are often covered
with a weak oxide scale which must be removed prior to bonding. Polymeric materials often contain plasti-
cizers which can migrate to the surface and the processing of composite structures creates problems due to
the use of mould release agents. Other contaminants include dust, dirt, grease and oils. Surface pre-treatment
further serves to optimise the degree of contact between the adhesive and substrates. This can be achieved
mechanically to increase the effective bonding area and may be augmented by a chemical modification of
the substrates’ surface layers.
The types of surface treatment available may be divided into three major categories, mechanical, chemical
and energetic, each of which may be further subdivided into particular techniques as described in Table 1.

5.1. Mechanical preparation

Mechanical abrasion is the most widely used surface preparation methodology, being suitable for almost all
materials. Abrasion acts to remove the weak boundary layers and change the topography of the substrate
therefore increasing the bondable surface area on a microscopic scale. It is further postulated that the rough-
ened surface aids in the wetting action of the adhesive and provides a modicum of mechanical ‘‘keying’’ to
augment the adhesive force although, as previously discussed, there is a degree of doubt as to whether this
actually occurs.
The simplest form of mechanical pre-treatment is the use of silicon carbide paper to abrade the bonding
surfaces. This can be carried out dry or in conjunction with an appropriate solvent. The quality of the sub-
strate surface obtained depends upon the grit size and the skill of the operator. Anyone who has repaired a
puncture on a bicycle will be well aware of this method! Whilst this is a simple technique it is extremely
crude and difficult to control especially in producing a uniform finish. The debris initially removed can eas-
ily be re-deposited. Furthermore, it is possible for mechanical abrasion to result in damage to ‘‘fragile’’ sub-
strates (particularly polymers and composites) and impair the integrity of the joint. Grit blasting involves
abrasion of the substrate using a gas propelled silica or alumina powder. The variables associated with this
technique are grit size, gas pressure, exposure time, angle of blast and distance between the blast nozzle and
the surface being treated. Nevertheless, a simple controllable process can be written and followed in order to
provide good quality reproducible results. Cryo- and soda-blasting are two similar but less aggressive tech-
niques. In cryoblasting solid carbon dioxide pellets are used instead of alumina grit. Soda-blasting was orig-
inally developed for the preparation of aircraft exteriors prior to painting. It uses a suspension of sodium
bicarbonate in water as the abrasive agent. ‘‘Wet blasting’’ is a combination of soda- and grit blasting in

Table 1
Surface preparation techniques commonly employed in adhesive bonding
Mechanical Chemical Energetic
Silicon carbide abrasion Solvent degreasing Plasma
Grit blasting Detergent washing Corona discharge
Cryoblasting Etching Flame
Soda/wet blasting Anodising Excimer laser
Peel ply Primer
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 335

which a suspension of alumina grit in water is employed. This method is far less aggressive than dry grit
blasting, is easily controllable and has the added advantage of cleaning any dirt and debris from the sub-
strate in the same operation. Ingress of water into the interior of the bonding surface is a potential problem,
although research suggests that this does not occur to any appreciable degree [10]. Nevertheless, a simple
drying operation in an oven will ensure there is no water contamination. Some substrates are, however,
quickly corroded by water, most notably ferrous alloys and magnesium, such that any preparation must
be carried out ‘‘dry’’ to avoid the production (rather than removal) of deleterious contamination. The effec-
tiveness of the pre-treatment can be measured very simply using the ‘‘water-break test’’. If a few drops of
distilled water are applied to the substrate, they will spread and wet the surface if it has been properly pre-
pared (Fig. 22). It should be noted that certain plastics may not be wetted by water but will be wetted by
the adhesive, and certain substrates will be damaged by water such that this technique cannot be used. In
those circumstances the quality of preparation must be verified by a more sophisticated methodology such
as a ‘‘surface roughness’’ measuring device.

5.2. Chemical compatibility

Bonding surfaces are invariably contaminated with dirt, oils and greases, etc., all of which will conspire to
destroy the integrity of any adhesive bond. It is, therefore, necessary to produce bonding surfaces which are
scrupulously clean using solvents. Degreasing in its simplest form can be performed using a lint-free cloth to
apply the solvent to the bonding surfaces. The cloth must be applied so that the surface is wiped in one direc-
tion only to prevent surface debris from being re-deposited. Ceramic, metal and some polymeric materials can
be very thoroughly cleaned by a solvent vapour degrease in a purpose built tank. The solvent is boiled in a
chamber whereupon it condenses on the cooler substrate, dissolving the oil and grease before dripping back
into the heating tank. Small components may be immersed in a solvent in an ultrasonic bath, the agitation of
the bath accelerating the process. The most efficient degreasing solvents are hydrocarbons (paraffins, for exam-
ple), ketones, particularly acetone, and alcohols such as methanol, ethanol and isopropanol. Chlorinated sol-
vents are also potent degreasers but are increasingly difficult to obtain due to environmental legislation. Great
care must be taken in the choice of solvent as experience teaches that certain substances are not suitable with
some adhesives, epoxies with ketones for example (Fig. 23). The majority of solvents are organic based and
may attack certain substrates, in particular amorphous polymers. In such cases it may be possible to use

Fig. 22. Water break test.

336 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 23. Solvent damage to adhesive bond.

detergents dissolved in water, alkaline or acidic solutions to clean the parts. These operations are typically car-
ried out between 40 and 70 °C. It should be noted that one of the major advantages of the wet-blasting prep-
aration technique is that the surface agitation induced by the high pressure water jet, which carries the
abrasive medium, acts as a very effective cleaning and degreasing agent. Having said that, it is advisable to
degrease the substrates prior to blasting to prevent contamination of the equipment as this could necessitate
a time consuming strip-down and cleaning operation.
Metal bonding surfaces rarely consist of pure metal but rather are a combination of oxides, sulphides, chlo-
rides and other atmospheric contaminants. Chemical etching can be employed to remove the weak scale and
replace it with a thin oxide layer which is mechanically and chemically compatible with the adhesive. Thermo-
dynamics dictates that different etches be developed and applied to different substrates. Aluminium alloys for
example bond extremely well following a chromic acid etch. Stainless steels are generally treated with a sul-
phuric acid based etchant, copper-based alloys with nitric acid and titanium and its alloys with alkaline per-
oxide. Chemical etching has been experimentally proven to increase the bond strength by up to double and can
significantly increase the lifetime of the bond by passivating the various surfaces and therefore preventing the
setting up of an electrolytic corrosion cell. It is best used as a supplement to mechanical pre-treatment. Etched
aluminium and titanium substrates are sometimes subsequently anodised. The purpose of this procedure is to
deposit a porous oxide layer on top of the oxide formed during etching. This enables the adhesive to penetrate
the pores to form a stronger bond, although it can be very time consuming and is not appropriate for some of
the chemically formed oxides which are metastable and must be bonded quickly (within 2 h) to ensure bene-
ficial effects.
Many polymeric materials, in particular polyolefines and PTFE, require pre-treatment of the bonding sur-
faces. Such plastics may be prepared for bonding by the application of a primer. Whilst not as effective as ion-
ization techniques such as corona discharge, the use of primers is very simple involving just the spraying or
brushing on of as thin a layer as possible to the substrates. After a short drying time (typically 10–60 s) the
joint is executed as normal. High bond strengths can be achieved with many polymers and others (most nota-
bly fluoropolymers) may be joined with a limited degree of success whereas without the primer they are impos-
sible to bond. It must be noted that a primer should not generally be used with materials which can easily be
bonded without this form of pre-treatment. Furthermore, the primer is used as an addition to the accepted
preparation procedures (cleaning, grit blasting, etc.) not an alternative. Primers often contain components
which enhance the environmental resistance and thermal stability of the bond. They are sometimes used as
the final stage of a multistage process for bonding metals as they can act as a medium which can bond readily
to both substrate and adhesive. They are particularly useful when using high viscosity adhesives which do not
flow readily over the substrates, and in preparing ‘‘difficult’’ materials such as copper-based alloys. The primer
can also be used to protect the prepared surfaces and thus increase the time available for executing the bond or
enable storage of pre-prepared parts.
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 337

5.3. Energetic preparation

Energetic pre-treatments include plasma, corona discharge, flame, and Excimer lasers. They all act to pro-
duce a change in the surface texture of the substrate due to the interaction with highly energetic species. Such
methodologies have been applied to metals, composites and plastics with varying degrees of success.

6. Production considerations

Process quality control is paramount in the exploitation of adhesives as once the joint has been made it is
difficult to inspect and in many cases cannot be reworked. This is particularly true of highly stressed, load
bearing bonds but is similarly good practise in those applications where the consequences of failure are less
catastrophic. In short there are two ways to perform a task; correctly or incorrectly. There must be no half
measures or short cuts! Adhesive products are used for a wide variety of bonding, thread locking, retaining,
gasketing and sealing applications. For many of these tasks it is sufficient to dispense product directly from its
container onto the surfaces to be joined. Other cases require very precise mixing and application. It is very easy
to execute a poor bond whereas the production of complex structures of dissimilar materials requires very
careful processing. Far from being de-skilling with respect to more traditional joining technologies, such as
welding, bonding must be carried out by well trained, highly motivated and skilled staff. This is something
which must be considered when setting up an operation. Bonding must be specified in a design on its technical
merits, it is not a cheap and easy option. The operator must wear gloves and ‘‘fit for purpose’’ clothing at all
times (lint free lab coat, etc.). Under no circumstances must the bonding surfaces be touched with a foreign
object except for any tools with which the adhesive is applied. It goes without saying that the tools must be as
thoroughly clean and degreased as the substrates. Great care must be taken to ensure that adhesives are used
sparingly. Insufficient adhesive results in poor joint performance and durability whereas an excess hampers
assembly and necessitates time consuming cleaning operations. This is best achieved by the installation of spe-
cialist dispensing equipment for the adhesives. The equipment will not only help to ensure accuracy and repro-
ducibility of the operation, but will also save money by reducing waste. Where appropriate, the adhesives must
be mixed thoroughly in the correct proportions, and the cure carried out correctly according to the manufac-
turers’ instructions; never take short cuts ‘‘you cannot hurry chemistry!’’ These latter two factors may be mon-
itored using thermochemical analysis techniques such as Differential Scanning Calorimetry (DSC) as part of
the quality control procedure [11,12]. DSC provides data on the thermodynamics and kinetics of the cure
which can be quickly compared with that from a ‘‘master batch’’ to ascertain that all is well (or otherwise).
When effecting primary structural bonds, it is standard procedure at BAR Honda to carry out a DSC analysis
on the adhesive prior to cure and a subsequent check on samples cured along side the component(s). This
enables both mixing and cures to be verified and documented.
Bonding of components whether they are polymeric, composite or metallic has increased in both quantity
and complexity in the construction of Formula 1 vehicles over the last decade and a half. The use of adhesives
within the industry was originally the exclusive domain of composite materials and all of the work carried out
by that specialist department. Introduction of bonded metallic and multi-material structures has resulted in
these techniques becoming interdepartmental involving additionally the joining of fabrications and monolithic
(machined) components. The manufacturing process requires all of the constituent parts to be scheduled to
arrive at the right place at the right time given the restrictions of the bonding process. Efficient management
of this production is vital to prevent bottlenecks from occurring. The time elapsed between preparation and
assembly must be kept to a minimum in order to minimise contamination. The working environment must be
controlled in terms of temperature and relative humidity coupled with general standards of cleanliness. The
volume of work within our team has developed to the point where it now commands a purpose built facility
culminating in a ‘‘clean room’’ environment for the final assembly of components. BAR Honda operates a
three stage process for bonding; Stage 1 takes place in the ‘‘dirty’’ area (anyone who has visited a Formula
1 factory will probably not understand this concept as they are kept like hospitals!) where any final stage mod-
ifications are carried out and parts are trimmed to fit. Any dust and dirt from this operation is removed before
the route moves onto the second stage. Stage 2 is carried out in a ‘‘semi-clean’’ environment and involves the
dry-fitting, assembly and inspection of the various structures. The final stage consists of bonding and curing in
338 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

the clean room. In addition to climate control, such a unit requires the specialist equipment needed for clean-
ing and pre-treatment of the substrates and dispensing of adhesives. It must furthermore be large enough for
assembly and storage of components during curing which may require the use of programmable ovens.
The number of adhesives and substrates used will increase with the passage of time owing to the very spe-
cific nature of many of the joints. Bonding procedures are developed in the R&D laboratory during the pro-
totyping phase of a new component or subassembly. They tend to be periodically modified and improved as
the technology develops, newer applications are introduced and feedback from operators is received. It is
important, therefore, to formalise these operations in order to achieve the following objectives:

1. To simplify the description of bonded joints on engineering drawings.

2. To remove any ambiguities as to how the bonds are to be prepared, components assembled and adhesives
cured.
3. As an aid to quality control in production.

The BAR Honda ‘‘system’’ calls up two types of procedure on drawings:

1. ‘‘Bonding Preparations’’ are the operations which must be carried out in order to prepare the various com-
ponents for bonding.
2. ‘‘Adhesive Cures’’ are procedures which govern mixing and application of the adhesive, assembly of the
components on the jig(s) and curing of the bond.

A simple list of instructions can then be added to an assembly drawing according to an accepted and doc-
umented format. It is here one reaps the benefits of a prototyping stage in the design process as it allows an
efficient manufacturing sequence to be derived and prevents any production problems at a later date.
The parts which make up a Formula 1 car are precision components designed to accuracies of a fraction of
a millimetre. In order to ensure this accuracy it is generally necessary to assemble, support and cure parts on
purpose built jigs and fixtures. Jig design is a very demanding operation (indeed the BAR Honda team employ
an engineer fully dedicated to this operation) requiring an understanding of both the function and engineering
of the artefact and a thorough knowledge of the industrialisation of its manufacture. Feedback from the pro-
duction staff is of vital importance if the operation is to be efficient. Considerations such as ease of assembly
and cleaning of excess adhesive, etc., are just as important as the accuracy of the final part. Fig. 24 shows just
such a jig used to produce rear suspension wishbones. Note how the jig is designed to have clear access all

Fig. 24. Composite wishbone assembled on bonding jig prior to curing.

 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 339

round and uses a carbon composite base plate to eliminate expansion during the heat curing process. It is fur-
ther configured to allow multiple (in this case two) items to be assembled on the same jig and has standing
support blocks to allow optimum use of the limited space within the curing oven. Optimum bond line thick-
ness is assured by using ‘‘glue rails’’ or ‘‘shim wire’’. Glue rails are raised lines (Fig. 25) machined or moulded
onto one of the substrates to produce an offset of the correct depth into which the adhesive collects. Shim wire
performs the same function using a metal wire of the correct diameter to set the gap between the substrates. In
addition to the formal engineering drawings, a production handbook is produced for each bonded assembly.
This document details all of the information gleaned during the prototype stage (build sequence, trim details,
etc.) and is continually updated to improve the process for present and subsequent designs.

7. Testing and durability of adhesive joints

Structural joints are required to perform under a combination of service conditions, which include both
static and dynamically applied loads and exposure to hostile environments, particularly temperature, moisture
and solvents. It is, therefore, paramount to design a bonded structure capable of operating under such con-
ditions and to develop processes to guarantee integrity over the entirety of their service life. The initial stage in
this operation is to carry out simple quasi-static mechanical testing to guarantee that the performance of the
component corresponds to the design calculations (Fig. 26). Testing to failure is used to determine safety fac-
tors and ensure that the bond is not the limiting factor. A further ‘‘proof’’ test may be developed in which a
component is loaded to a value beyond that it is postulated to endure in service but not so high as to result in
failure. Such a test may be carried out on all components when new and periodically through their service life
in order to pass them ‘‘fit for service’’. Similarly, a measure of its global stiffness is a useful check on the integ-
rity of the joint (or indeed any other part of the structure). Comparison of these data with the initial value and
batch average for identical components is an excellent form of condition monitoring since any deterioration in
parts will be manifest as increased compliance. Any measure of reduced stiffness is, therefore, indicative as a
problem and demands corrective action.
Adhesives generally exhibit superior fatigue properties in comparison with competitive joining technolo-
gies. This is a consequence of a more even distribution of stress. Nevertheless, the consequences of failure
are such that a great deal of work is necessary to minimise the probability of such an event. When evaluating
the fatigue performance of a joint one must consider a wide range of variables [13] (stress amplitude, mean
stress, frequency, waveform, ambient and internal temperature of the system, etc.). There are, therefore,
potentially more unknowns to consider than with a more ‘‘traditional’’ joining technique. The complex fatigue
conditions must be approximated to much simpler laboratory test conditions in which many of the parameters
are held constant. This presents a formidable task in that it is vital to identify those service conditions which
directly influence the life of a particular joint. Failure to do this will render the test useless and any results not

Fig. 25. Methods of setting the glue gap.

340 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 26. Quasi-static testing of a bonded composite wishbone.

merely of no value, but may present too optimistic (or pessimistic) a prediction of the durability of a particular
component.
A number of loading methods and specimen geometries have been reported in the literature [14,15] for fati-
gue testing of adhesive joints. In practical applications however, the life limiting factor of many joints is the
stress concentration set up by the geometry of the design. Consequently, there is a tendency to test full-scale
structures rather than standard test pieces [16]. Whilst this approach provides the designer with confidence
about a particular item, it does not generate basic materials data apposite to other applications. Formula
1engineers, as they strive for improved performance, often use the very latest materials in novel applications,
often with only limited property data available. Despite great advances in computer stress analysis and mate-
rials science, our ability to produce components more often than not leads to our ability to fully understand
them. The finite element analysis carried out during the design phase, whilst being extremely accurate, assumes
a perfect bond and is unable to take into account any deterioration due to fatigue, etc. The dynamic perfor-
mance and long-term durability of such structures, particularly those involving adhesive joints with their
inherent variability, can only be demonstrated by means of experiments with complete components and sim-
ulation of actual operating conditions. The validity of accelerated durability tests depends on how closely the
extrapolation follows service conditions and as such relies very strongly on the experience of the team’s tech-
nical staff. Durability testing may be carried out using calculated loads or service data collected from the test
track to drive servohydraulic actuators which stress the components on purpose built jigs [17,18]. The digital
operation of the equipment enables the programming of safety factors with relative ease. Fig. 27 shows a test
facility developed in partnership between BAR Honda and Instron. The 12-axes are able to apply all six
degrees of freedom (3 forces: vertical, longitudinal and cornering and 3 moments: braking, camber and steer)
to a car’s suspension system. It provides, therefore, a very accurate assessment of the durability of the bonded
components of which it consists. The effect of increased temperature is to increase the rate of any deteriora-
tion. This can be modelled by incorporating a heating chamber into the test rig. The time constraints imposed
due to the nature of the sport make more long-term environmental effects such as moisture uptake difficult to
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 341

Fig. 27. 12-axis test rig used to test a complete suspension system.

simulate. Given time one would carry out tests using relative humidity cabinets, etc. Instead, environmental
testing tends to be an ‘‘over-kill’’ such as a 24 h condition exposure to, for example, hydraulic fluid prior
to the durability tests. One has to be very careful when interpreting the results from this type of test. In
the interests of expediency there is a risk of producing misleading results. Too harsh an environment compared
to reality may result in premature failure (and subsequent anxiety) over a mechanism which would never occur
in practice. Similarly, if the environment is not simulated in such a way as to at least provide subjective results,
there is likely to be a degree of over-confidence in the component. There is a tendency, therefore, for bonds to
be over designed but this is obviously preferable to a failure.

8. Fatigue mechanisms

Fatigue is generally considered to involve initiation of a crack and its subsequent propagation. Most stud-
ies, however, tend to neglect the initiation phase and focus upon growth of a crack of microscopic proportions,
since it is this portion of the failure mechanism which actually determines the fatigue life. This is an acceptable
practice since the initiation of a crack is difficult if not impossible to model, and any calculated lifetime will be
somewhat conservative which is generally preferred by designers. In practical systems cracks tend to nucleate
as a result of defects in the structure. These are introduced either in manufacture or localised damage during
service. Studies on carbon fibre composites bonded using epoxy adhesives show fatigue failure to occur due to
crack growth within the adhesive [19]. A similar process is observed when bonding composite to metal
(Fig. 28). The concept of fatigue crack growth in adhesive joints is a great concern, but little work has been
reported to enable a fundamental understanding of the mechanisms involved. It is, therefore, very difficult to
establish design criteria and develop quantitative methodologies for predicting service lives. The nature of the
damage which accumulates at the crack tip, and causes the loading/unloading cycle to be so deleterious, has
not been identified. The affects of changing the many loading and environmental parameters upon the rate of

Fig. 28. Fatigue failure in metal to composite joint.

342 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

crack growth is all but unknown. Sub-critical crack growth is often observed in real systems but it is impos-
sible to quantify the rate of propagation because the defective area increases the compliance of the system,
consequently changing the balance of the various stress components.
Despite a lack of hard evidence, a number of studies have suggested an endurance or fatigue limit in adhe-
sive systems (analogous to that observed with ferrous alloys) of 35–45% of ultimate strength [20]. Experimen-
tal evidence suggests that applying a design safety factor of 3 or more to a joint will all but eliminate purely
mechanical fatigue problems (this does not of course negate environmental degradation and the effect of
defects). Regardless of anecdotal evidence, the work of Lewis et al. is considered to over-simplify the problem
and make far too many assumptions such that any agreement between calculated life and general design prac-
tice is far more likely to be a fortunate coincidence and is certainly not an established phenomenon [21].
A further complication in understanding the fatigue behaviour of bonded joints arises from the nature of
the adhesives themselves. A perfectly elastic material will remain at the temperature of the test environment
throughout a fatigue test since no energy is dissipated within the specimen. Polymeric materials, however,
are viscoelastic and exhibit mechanical hysteresis even at relatively low applied strains. Under cyclic loading
some of this deformation energy will be dissipated as heat during each loading cycle. As a consequence, the
adhesive’s temperature will rise, until the heat generated per cycle is equal to the heat dissipated by conduction
and radiation. It is theoretically possible for the adhesive to heat up quite quickly to relatively high temper-
atures. The affect, if any, is difficult to predict; the elevated temperature may soften the adhesive causing blunt-
ing of the crack tip, hence act as a toughening mechanism, or may result in a loss of strength by weakening the
adhesive. Thermal effects resulting from fatigue loading may well be very significant in determining the long-
term performance of a joint under load, but are frequently far too complex to interpret. This then is yet
another factor making the fatigue response of adhesives difficult to quantify.

9. Defects in bonded structures

A number of possible sources of weakness have been identified within adhesively bonded joints. These
occur within the adhesive itself and at the adhesive/substrate interface. They may propagate under load
and can lead to degradation of the component due to in-service degradation. The strength of the joint depends
on how successfully the loads are transmitted by each part of the structure. Any weak link will, therefore, lead
to premature failure. The most common defects found in bonded joints are shown schematically in Fig. 29
[22]. Porosity results from trapped gasses (perhaps during mixing) and volatiles associated with the curing pro-
cess. Voids may be formed by the coalescence of porosity, entrapment of air during application of the adhesive
or by insufficient adhesive being applied. Incorrect cure is caused by the presence of contaminants, bad for-
mulation, or poor mixing of the adhesive. It may occur locally in small pockets but is more likely to occur
throughout the whole of the bond line. Cracks within the adhesive are generally associated with curing and

Fig. 29. Potential defects in adhesive bonds [22].

 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 343

Fig. 30. Defects in honeycomb structures.

Fig. 31. ‘‘Peel ply’’.

thermal shrinkage during manufacture or handling and impact in use. This is a particular concern when using
high temperature adhesives which can sometimes be quite brittle at room temperature.
A weak or non-existent bond generally results from a contaminated substrate surface or if the adhesive is
used too long after mixing. The resulting total lack of bond, or disbond of zero volume at the interface, can be
detected using NDT techniques albeit with a degree of difficulty. Their significance on the performance of the
joint is however very difficult to predict. One of the major uses of structural adhesives in Formula 1 is in the
bonding of carbon composite skins to both aluminium and nomex (polyaramid) honeycombs in chassis and
bodywork [2,6]. The defects particular to this type of structure are illustrated in Fig. 30. A skin-core disbond
results from a lack of adhesive or a gap between skin and core caused by damage to one or both components,
such as a locally crushed core or defective skin for example. A complete skin-core disbond can also arise when
the fillet of adhesive is not properly formed. Specific bonding problems can arise with the ‘‘peel ply’’ (Fig. 31)
which is frequently used when joining composite materials [23]. In practice these consumables are not removed
until just prior to bonding and are employed in the mistaken belief that the roughened surface which results is
favourable for bonding. It is not uncommon for the release agents used in formulating the peel ply to migrate
into the surface of the composite substrate which may lead to a weakening of the subsequent bond. Hot cured
epoxy adhesives offer an advantage in such applications in that they can dissolve some of the surface contam-
ination. In the long-term, dispensing with peel plies in favour of a mechanical preparation technique offers
many advantages.
As previously discussed, bonded joints are particularly susceptible to attack by aggressive chemicals and
moisture. The mechanical properties of the adhesive may be progressively degraded by solvent or other dam-
age, but a far more severe problem is degradation of the bond interface. Not only is the bond strength reduced,
it may lead to weakening of the substrate(s) due to corrosion. Chemical etching of metallic substrates is par-
ticularly useful in preventing or, at the very least, slowing this process by reducing the electrochemical
344 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

potential. The nature of the loading regime also affects the durability of the joint. Adhesive bonds, like com-
posite materials, rely on a perfect stress transfer between their constituents. High frequency (shock) loading
may result in local stress concentrations, especially in the presence of flaws and other inhomogeneities, which
exceed the bond strength and lead to a progressive, or in the worst case, catastrophic failure. Defects within
bonded joints can be detected by a number of techniques, but there is insufficient data and understanding to
allow a quantitative analysis of their affect on durability. Clearly the presence of defects can only have a neg-
ative influence on the service life of a bonded component. Any lifing predictions will, therefore, have to be
qualitative based upon experience and laboratory testing.

10. Non-destructive testing

The most widely employed NDT technique used for adhesives within the Formula 1 industry is the ‘‘coin
tap’’ test to check the integrity of honeycomb structures. The region of the component to be tested is tapped
with a coin, the inspector listening to the resulting sound. Defective areas sound duller than defect-free areas.
This has been developed into an automated device using a small hammer with a force transducer incorporated
into the head which can be used to provide an assessment of the integrity of the bond albeit somewhat crude.
Ultrasonic testing may be carried out either with a single transducer in pulse-echo mode or with two transduc-
ers in through-transmission mode. Defects in multi-material bonded structures, particularly those involving com-
posites, may be identified using low energy X-rays (Fig. 32). The team is currently investigating new equipment

Fig. 32. X-ray photograph used to search for defects in bonded suspension members.
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 345

and software which can produce a 3D image from the orthogonal shots shown in Fig. 32 which will provide
accurate representation of defective areas.
The primary weapons in the team’s NDT armoury are a combination of mechanical proof testing and ultra-
sonic. Provided the joint has been designed with a large enough reserve factor, the service stress levels will be
considerably lower than those demanded during pre-service proof testing. In the event of a poor bond being
produced it ought to be self evident from the mechanical test results. Similarly, periodic retests during the ser-
vice life of the component provide a fair estimate of the durability of the bonded joints. One aims to apply as
high a proof load as possible without damaging the components in any way. All BAR Honda composite parts
(particularly the bonds) are scanned using an ultrasonic flaw detector connected to a hand held probe. The
flaw detector is calibrated to the probe being used and the thickness range required. The probe is manually
scanned across the test area using a gel couplant to provide a sound path for the compression wave (around
5 MHz) emitted from the probe face and the reflected signal to be received by the probe. The signal is then
displayed onto the flaw detector screen. The length of time taken for the reflection to be received is translated
into a measurement, corresponding to thickness. The effectiveness of the ultrasonic technique may be illus-
trated using a rear upper wishbone outboard end, inspecting a carbon ‘‘doubler’’ bonded to the carbon lam-
inate. A component with a good bonded joint will show a total through thickness signal at 80% full screen
height and a smaller signature signal at the glue interface as in Fig. 33. A glue void/disbond will not show
the total thickness. Rather the glue interface signal will become high, breaking the screen gate thus reading
the depth of the defect as in Fig. 34. By carefully scanning the whole of the component, the skilled operator
may build up a detailed map of any defects within the structure. The ultrasonic technique, in stark contrast
with proof testing which must be carried out in the laboratory, is ‘‘air portable’’ and can thus be transferred
to the circuit. It is extremely useful for in the field evaluation of parts following ‘‘incidents’’.
Following an ultrasonic scan, if a defect area is detected, the team employ a further NDT procedure.
‘‘Spott-video-inspection’’ is used to confirm the ultrasonic findings and to carry out a more detailed investi-
gation of the fault. This technique uses thermal imaging to record results from gently heating the test piece
using a diode laser. A high quality thermal imaging camera captures data that is processed for interrogation
and future reference. The diode laser uses a scanning head to cover the area of interest with a blanket of heat.
A scan setup and the image recorded from the suspect wishbone are shown in Fig. 35. The darker shaded area
confirms a small disbond; this is considered as an acceptable discontinuity which will be recorded in the com-
ponent’s life history data and closely monitored.
Defects may be readily identified and measured but their effect on the durability of the joint cannot be pre-
dicted (unless of course they are extremely large). The only recourse is to instigate a ‘‘closed loop’’ research
programme into the production schedule. All bonded joints will contain flaws to a greater or lesser extent.
Once those defects have been identified a degree of common sense and experience is required in order to decide
which components are to be rejected. Those items are then put into the durability testing programme in the

Fig. 33. Ultrasonic scan of acceptable bond.

346 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

Fig. 34. Ultrasonic scan of defective bond.

Fig. 35. ‘‘Spott-video’’ NDT analysis.

laboratory whence they are used to determine the level of defect which can be tolerated and thus progressively
re-define the pass/fail criterion. One observation that becomes very obvious from this process is that, contrary
to popular belief, adhesives are remarkably tolerant to defects.

11. Total quality management

There are clearly a number of distinct advantages offered by the extensive use of adhesive bonding. Nev-
ertheless, there are also a significant number of problems which arise from their seemingly unpredictable
long-term behaviour. The inability to accurately quantify the durability of adhesives arises from a relatively
poor understanding of their materials’ science, particularly with respect to the affect of environmental vari-
ables. Many of those difficulties can, however, be circumvented by the introduction of Total quality manage-
ment (TQM). TQM is a process of controlling all aspects of the technology from inception through
exploitation to eventual obsolescence. Central to this is full traceability of components and operations. A
‘‘zero defects’’ operation is employed in which each component is documented throughout the entirety of
its manufacture, testing and service life. All aspects of manufacture must be fully documented and follow pro-
cedures developed during the R&D phase of component design. Complex assembly operations are broken
down into more manageable segments which can be efficiently managed. Each stage of the operation must
be verified using non-destructive testing (NDT) techniques in order to ensure that all finished components
 G. Savage / Engineering Failure Analysis 14 (2007) 321–348 347

meet specification. Once passed for exploitation, a condition monitoring programme must be employed to illu-
minate any potential problems before they become serious.
In-service component failure represents the most serious problem we can possibly encounter in the running of
the race and test cars. Aside from the obvious concerns of driver safety and effects on operational logistics, its
implications may also be manifest in the need to redesign and replace broken parts. The ‘‘lifing’’ process operated
by the team seeks to provide a system to ensure, to the best of our skill and ability, the total quality management of
all Class A components on the cars in order to maximize safety and reliability. A Class A component is defined as
one that should it fail would cause the driver to lose control of the vehicle. All of those components subjected to
statutory FIA safety tests and requirements are also defined in the Class A category. ‘‘Lifed’’ parts are those com-
ponents whose integrity is deemed critical to the safety and/or operation of the car. These parts must be fully con-
dition monitored through the entirety of their service history and their service life limited to that which can be
guaranteed through calculation and testing. The TQM process operated by BAR Honda for such pieces is:

1. All aspects of manufacture and exploitation must be fully traceable (a process known as ‘‘lifing’’).
2. A representative sample of parts must be fully tested in the laboratory in order to prove that they meet
design criteria for load bearing and other properties.
3. A representative sample of parts must be fully tested in the laboratory in order to prove that they are capa-
ble of surviving a simulation of a season’s operation without deterioration in any way.
4. A car set of parts must complete a minimum of 1000 km of trouble free operation on a test car.
5. Following a service interval of 2500 km (an arbitrary figure chosen to approximate to 2 tests or 3 races)
parts must be returned to the factory for full proof and NDT inspection.
6. Only those components that show no deterioration may be returned to service.

TQM procedures are developed during the prototyping phase of a new component or subassembly. They
tend to be periodically modified and improved as the technology develops, newer more specialised applications
are introduced and through feedback from operators. It is important understand the concept of a ‘‘process’’ in
this context: A process is a series of operations which is developed such that it represents the ‘‘best practice’’
for the task involved. It will, therefore, have no end (unlike a programme or project) since it will constantly
evolve along with the technology.

12. Conclusion

There has been traditional reluctance amongst engineers to use adhesives primarily due to the fear of cata-
strophic failure. The widespread application of bonded joints in performance critical applications on Formula
1 racing cars shows just how useful adhesives can be. It would not be unreasonable to suggest that components
manufactured using bonded joints are generally superior to those assembled using more traditional joining tech-
nologies. Indeed, many of the assemblies’ common on contemporary F1 cars simply could not be made any other
way. There is however a very significant ‘‘price of conformity’’ which must be paid. The lack of numerical design
and durability data demand a very sophisticated TQM process is in operation if the integrity of highly stressed
joints is to be guaranteed. Production controls are paramount if the advantages of adhesives are to be successfully
exploited. Attention to detail in manufacture is vital to ensure enhanced mechanical performance, improved
durability, and increased service life of components. The integrity of adhesive joints can only be guaranteed
by observing a ‘‘zero defects’’ approach to quality control. Whilst this may well be acceptable in Formula 1
and ‘‘high end’’ aeronautical applications, it will preclude the technology from more cost sensitive theatres of
engineering. What is required is a far greater understanding of the materials science of adhesives and the need
to consider the design, production, testing and service of the joint as integral components of a complete system.

References

[1] Savage GM. Metal Mater 1991;7(10):617.

[2] Savage GM. J STA 2000;140:18.
[3] Kinloch AJ. Adhesion and adhesives. Chapman & Hall; 1987.
348 G. Savage / Engineering Failure Analysis 14 (2007) 321–348

[4] Tabour D, Winterton RHS. Proc R Soc 1969;A312:435.

[5] Israelachvili JN, Tabor D. Proc R Soc 1972;A331:19.
[6] Savage GM. Metal Mater 1992;8(3):147.
[7] Savage GM. Race Tech 1995;1:1.
[8] Armstrong-Wilson C. Racecar Eng 2004;14(9):32.
[9] Loctite World-wide Design Handbook. Loctite Corporation; 1995.
[10] BAR Honda Formula 1 proprietary data; 2000.
[11] Cox P, Savage GM. Automot Eng 1992;2:14.
[12] Utschick H. Methods of thermal analysis. Landsberg (Germany): Ecomed Verlagsgesellschaft; 1997.
[13] Andrews EH. In: Brown WA, editor. Testing of polymers IV. New York (USA): Interscience; 1969. p. 237.
[14] Imanaka M, Kishimoto W, Okita K, Nakayama H, Shirato M. J Comp Mat 1984;18:412.
[15] Jones TB, Williams NT. In: Proc. SAE int. cong. and exposition, Detroit (USA); 1986. p. 1.
[16] Hart-Smith LJ. In: Kinloch AJ, editor. Developments in adhesives-2. London: App. Sci. Pub.; 1981. p. 1.
[17] Savage GM. Proc STA 1996;125(May/June):33.
[18] Savage GM, May S. J STA 1999;136(May/Jun):35.
[19] Liechti KM, Lin C. In: Structural adhesives in engineering. London: IMechE; 1986. p. 83.
[20] Lewis AF, Kinmouth RA, Kreahling RP. J Adhesives 1972;3:249.
[21] Wake HC, Allen KW, Dean SM. In: Hepburn C, Reynolds RJW, editors. Elastomers criteria for engineering design. London: App.
Sci. Pub.; 1979. p. 311.
[22] Adams RD, Cawley P. NDT Int 1988;21(4):208.
[23] The composites materials handbook – Mil 17. Materials usage, design and analysis, vol. 3. Lancaster (PA): Techonic Publishing Co.;
1999. p. 1–28.