You are on page 1of 10

International Journal of Adhesion & Adhesives 22 (2002) 235244

Effect of surface roughness on the strength of cleavage joints


M. Shahid1, S.A. Hashim*
Department of Mechanical Engineering, University of Glasgow, James Watt Building, Glasgow G12 8QQ, UK
Accepted 27 September 2001

Abstract

In this paper experimental and numerical results on the effect of surface roughness on the cleavage strength of standard steel/steel
cleavage specimens have been presented. Cleavage joints with different surface roughnesses were mechanically tested and examined.
In numerical modelling, the effect was idealised by small butt joints with macro-roughness to determine the maximum normal tensile
stresses at the interface between adhesive and adherend. Good agreements were found between the numerical and experimental
results.
It was also found that average roughness (Ra ) appears to have a linear relationship with average cleavage strength. Average
cleavage strength is also found to be a function of linear prole length (Rlo ). r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: A. Epoxy; B. Surface roughness; B. Steel; Cleavage; Adhesive; Linear prole length

1. Introduction stress applied [9]. Of possible positive effects of surface


roughness [911], increase in surface area results in
Roughness of adherend surfaces has frequently been increasing intermolecular bonds and keying for mechan-
used as a design parameter for adhesive joints. A ical adhesion. This in turn can divert the failure path
number of researchers have examined its effect on the away from the interface into the bulk of the adhesive.
strength and durability of adhesive joints using various However, the actual microscopic distribution of stress at
adherends and adhesives [17]. There is, however, no the rough interface is complex.
published quantitative data, which relates surface This study examines the inuence of surface rough-
roughness parameters to the strength of cleavage joints. ness of a steel adherend on cleavage strength. An
Some form of substrate pre-treatment is almost attempt has been made to relate the surface roughness
always necessary to achieve a satisfactory level of bond parameters Ra and Rlo to cleavage strength. Appendix A
strength. Almost all treatment methods do bring some illustrates the two parameters as dened by ISO 4287
degree of change in surface roughness but grit-blasting is 1984.
usually considered as one of the most effective methods To produce varying degrees of surface roughness,
to control the desired level of surface roughness and steel specimens were diamond polished and grit-blasted
joint strength. Grit-blasting does not only remove weak with four sizes of alumna grit. Pre-treated surfaces were
boundary layers but can also alter the chemical examined with Talysurf Form Series 2 surface proler to
characteristics of the adherends [7]. The authors earlier measure surface roughness parameters like Ra ; Rlo and
work on steel cleavage specimens showed the effective- root mean square slope, Rdq :
ness of grit-blasting over diamond polishing in achieving Surface roughness was also modelled numerically,
improved cleavage strength [8]. using nite element analysis to compare the level of
The relationship between roughness and adhesion is stress at the at and rough surface.
not very simple. Optimum surface prole varies from
one adhesive to another, and depends upon the type of
2. Experimental programme
*Corresponding author. Tel.: +44-141-330-4319; fax: +44-141-330-
4343.
E-mail address: shashim@mech.gla.ac.uk (S.A. Hashim). The materials used in this study were mild steel to
1
Presently at the Department of Engineering, University of Cam- British Standard BS4360 grade 43A and a structural
bridge. epoxy adhesive, Araldite 420A/B (Redux 420A/B).

0143-7496/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 7 4 9 6 ( 0 1 ) 0 0 0 5 9 - 8
236 M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244

Nomenclature Ra Average roughenss, mm


Rlo Linear prole length, mm
S22 Normal tensile stress along y-axis, MPa R2lo Effective surface area for bonding, mm2
S11 Normal tensile stress along xy-axis, MPa Rdq Root mean square slope, 1

Standard cleavage specimens were made to British off limit of 0.8 mm. A sampling length of 15 mm was
Standard BS5350: C1: 1986 [12] with bonded dimensions taken in each case. Measured values of Ra ; Rlo ; and Rdq
of 25 mm  25 mm. Conguration of the specimen is are given in Table 1. Ra values of the various pre-treated
shown in Fig. 1. substrates were found to be in line with those measured
Adherends surfaces were either prepared by grit- by Gilibert and Verchery [2].
blasting or diamond polishing. The grit-blasting was The specimens were bonded about 24 h after grit-
performed on acetone degreased specimens using blasting and were degreased with acetone before
Saftigrits alumna grits from Guyson Corporation.
Four grades, 120/180, 40/60, 30/40 and 24/30 mesh were Table 1
used to produce different levels of surface roughness. Surface roughnesses
The process was performed at a pressure of approxi- Average Average Average root
mately 550 kPa, at right angle to the surfaces and at a roughness prole length mean square
distance of about 5 cm from the nozzle for the duration (mm) (mm) slope (1)
of about 30 s [2]. Polishing was performed using coarse Surface nish Ra * Rlo Rdq
sandpaper rst followed by ner sandpaper and then
diamond polishing with oil-wetted 1 mm diamond paste. Grit-blasted
surface
Afterwards, the polished specimens were washed with
120/180 0.9870.05 13.4470.01 12.5270.59
soap and water, degreased with acetone and then dried 40/60 2.9770.18 13.6970.04 22.7070.56
with hot air. 30/40 4.2370.25 13.8470.04 24.2470.89
After pre-treatment, the surface roughness of adher- 24/30 6.3170.28 13.9570.07 25.9670.82
ends was measured using Taylor Hobsons Form Polished 0.0470.02 12.7970.01 0.1370.02
surface
Talysurf Series 2 50i surface proler with a 2 mm
diameter stylus tip. IS0-2CR lter was used with a cut- *RaFCentre line average (CLA).

15 mm

25 mm
25 mm

Mild steel Adhesive


Fig. 1. Standard cleavage specimen.
M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244 237

20

15

Normal stress, MPa


10

0
0 5 10 15 20 25

-5
Distance along adhesive line, mm

Fig. 2. Stress distribution in a typical steel/steel cleavage joint [8].

bonding. A manual dispensing/mixing gun was used


with an appropriate mixing nozzle to dispense adhesive
in the correct ratio. Adhesive was applied and spread
onto the bonding surfaces with a spatula. An adhesive
thickness of 0.5 mm was applied to all specimens. This
was produced by attaching two wire spacers to the
adherends near the front and rear ends of the joints. The
specimens were then bonded and clamped using a
specially designed jig and were cured for 2 h at 701C.
The cured specimens were removed from the jig and the
adhesive llets were removed manually with a razor
blade.
The specimens were tested destructively on a Llyod
10000 L tensile testing machine using standard testing
xtures. All tests were carried out under monotonic
loading at room temperature with a cross-head speed of
0.5 mm/min. A minimum of ve specimens for each (a) (b)
surface condition was tested to achieve an average Fig. 3. Idealisation of cleavage joint as a series of butt joints (a) full
result. After each test the failure load was recorded and model, (b) partial butt joint model.
fractured surfaces were examined to determine whether
the failure was adhesion or cohesion. In some cases the
failure is referred to as adhesion/cohesion, especially
when no clear pattern is apparent. In view of the shape of the actual roughness prole of a
grit-blasted specimen (Fig. 4) the roughness of the butt
joint was idealised into convex and concave shapes, as
3. Numerical modelling shown in Fig. 5. The surface of the upper adherend was
idealised as at shape (polished). This was meant to give
Elastic stress distribution in a typical steel/steel an easy comparison between the stresses at polished and
cleavage joint was previously studied [8] and is shown at rougher surfaces. It was modelled in 2-D eight-noded
in Fig. 2. It was realised that such a stress prole might reduced integration quadrilateral shell elements using a
approximately be represented as a histogram by assum- PATRAN pre-processor and an ABAQUS processor
ing that a cleavage joint is made of a series of small (Fig. 6). The adhesive thickness (0.5 mm) was modelled
independent butt joints. A similar approach was recently with ve elements through thickness. Fine mesh
used for a lap shear joint by Owen and Lee-Sullivan [13]. elements were applied to the adhesive region at the
To study the effect of surface roughness on macro-level edges to account for the high stress gradients. Cong-
each bar in the histogram was then considered as uration and meshing of the numerical models are shown
representative of a portion of surface roughness (Fig. 3). in Fig. 7. Elastic isotropic properties were considered for
238 M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244

Fig. 4. 3D and 2D surface proles of grit-blasted steel surface measured by Talysurf.

adhesive interface with the upper at steel surface (site


1-1), the centre of the adhesive line (site 2-2) and the
lower triangular steel adherend surface (site 3-3). The
maximum adhesive normal/tensile stress (S22 ) nearer the
edge was taken as a comparative failure criterion. In
case of the lower triangular surfaces, values of normal
tensile stress were transformed with reference to the
slope angle (y) as shown in Fig. 8. The angle was taken
as equal to the average of the root mean square slope
Rdq for various specimens, which is 22.71 in the case of
40/60 grit-blasted specimens. To avoid mathematical
singularity problems at the free tension edge of the joint,
stresses at the edge nodes were ignored.

4. Results and discussion


Fig. 5. Schematic diagrams of roughness models.

The roughness measurements and the results from


mechanical testing are presented in Tables 1 and 2,
both adhesive and steel. A nominal distributed load of respectively. The average cleavage strength from experi-
1 kN was applied at the top of the butt joint and the ments (Table 2) is calculated by dividing the failure load
lower adherend was constrained in the three axes. by the bonded area. The numerical results for the butt
Three possible failure sites within the adhesive line models are given in Fig. 9. The grit-blasted specimens
were considered as shown in Fig. 8. These were the produced higher strength than the polished ones, which
M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244 239

Fig. 6. Numerical model.

Table 2
Cleavage strength of different surface nishes

Surface nish Average Coeff. of Possible


strength (N/ variation (%) failure
mm2) initiation

Polished 15.8 2.9 Adhesive


Grit-blasted
120/180 16.4 4.0 Adhesive/
cohesive
40/60 17.0 3.2 Adhesive/
cohesive
24/30 17.5 1.9 Adhesive/
cohesive
Fig. 7. Possible failure sites in roughness models. 24/30 18.3 1.3 Adhesive/
cohesive

S22 is in line with the ndings of Jenning [3]. A similar trend


can also be seen from the numerical results. The details
are discussed in the following sections.

S22Cos2 4.1. Effect of Ra

Fig. 10 shows the relationship between the average


cleavage strength and the Ra value of the adherend
surfaces. It can be seen that cleavage strength appears to
increases linearly with the Ra value. The increase in
cleavage strength may be attributed to an increase in
surface area by forming of mini scarf joints on adherend
surfaces at micro level. This nding is in line with that of
Sargent [4] who reports, without reporting the level of
increase, an increase in peel strength of aluminium test
specimens with increasing surface roughness. However,
he did not nd any correlation between peel strength
Fig. 8. Transformation of normal stress on the rough surface. and any features of the oxide or interfacial region. On
240 M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244

1
(a) Site 1-1 Site 2-2 Site 3-3

Normalised normal stress, MPa


0.8

0.6

0.4

0.2

0
0 1 2 3 45
Distance along adhesive line, mm

1
(b) Site 1-1 Site 2-2 Site 3-3
Normalised normal stress, MPa

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5
Distance along adhesive line, mm
Fig. 9. Normal stress distribution in adhesive line (see Fig. 7) (a) Model 1, (b) Model 2 (stresses at site 3-3=S22 cos2 y).

20

19
Cleavage strength, MPa

18

17

16

15
0 1 2 3 4 5 6 7

Average roughness,Ra, m

Fig. 10. Graph showing variation of cleavage strength with average roughness, Ra :

the other hand, Harris and Beever [14], Thery et al. [15] surface roughness by mechanical treatment. These
and Critchlow and Brewis [1] found no appreciable contrasting ndings may be due to the fact that each
change in joint strength with increasing adherend researcher used a different set of adherend, adhesive and
M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244 241

20

19

Cleavage strength, MPa


18

17

16

15
160 165 170 175 180 185 190 195 200

Effective area, Rlo2, m2

Fig. 11. Variation of cleavage strength with effective area of bonding (experimental results).

joint geometry. Moreover, the overall effect of grit-


blasting is not limited to the removal of contamination
or to an increase in surface area. This also relates to
changes in the surface chemistry of adherends [7] and to
inherent drawbacks of surface roughness, such as void
formations and reduced wetting [17]. Sometimes these
might act as antagonists.

4.2. Effect of R2lo

R2lo was considered as a measure of the effective


surface area available for bonding and this parameter
was noted in each case while keeping the same
evaluation length, lter, etc. These values were then
compared with cleavage strengths as shown in Fig. 11. It
can be seen that the experimental cleavage strength
increases as the effective surface area increases. It is
however, realistic to believe that this increase in strength
combined with an increase in surface area shall be
Fig. 12. Fractured surface of grit-blasted steel cleavage specimen
limited by bulk adhesive strength i.e. until failure showing failure initiation from the edge.
becomes cohesive within the adhesive.
Due to concentrated loading at one end of the joint
and the complex geometry of the surface consisting of
hills and valleys of various shapes, it is very difcult to
look into the contribution of different portions of the corresponding stresses at sites 2-2 and 3-3. As expected,
overall surface. This is especially important because the adhesive stresses nearer the edges are signicantly higher
initial few millimetres of the adherends surface con- than other locations causing failure to initiate from the
tribute signicantly towards the total joint strength. edges (Fig. 12). This is also consistent with experimental
observations where stress whitening is more prominent
4.3. FE Analysis at the edges. As an average of both models, stresses at
site 1-1 (at top surface) are higher by about 30% than
Fig. 9 shows normalised critical failure stresses at the at site 3-3 (rough lower surface). This is a good
two adherends interfaces and the middle site of the representation as in practice the roughness proles are
adhesive layer. It can be seen that in both models made of a continuous series of concave and convex
stresses at site 1-1 (Fig. 7) are higher than the shapes. The experimental results (Table 2) show only
242 M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244

Fig. 13. Fractured surface of grit-blasted cleavage joint showing stress whitening in the initial joint area.

16% difference in average cleavage strength between the


polished and rough (24/30 mesh) conditions. Although
these experimental conditions are assumed to represent
sites 1-1 and 3-3 of the numerical models, respectively,
the scatter between the numerical and experimental
results is somewhat high. Besides the model representa-
tion problems, scatter is also possibly due to the lack of
wetting in bonded joints and this could be more critical
in the case of cleavage where stresses are highly
concentrated at the edge. Surface prole can lead to
the trapping of air beneath the adhesive and poor lling
of crevices. These voids can lead to stress concentrations
and hence lower joint strength [16]. Depending on the
nature of the roughness and adhesive, the surface may
not be wetted properly, and the adhesive may even start
curing before going deeper into the pore. Hitchcock et al.
[17] reports that increasing roughness usually reduces
the wettability of the surfaces with the exception of very
low viscosity adhesive and signicantly rough surfaces.
Sargent [4] also analysed the contribution of increased
area and surface roughness on the force required for
detaching adhesives and adherends. By considering a
surface composed of spherical depressions, he found an
increase of 15% in the normal tensile force required
detaching the adhesive from the adherend surfaces. Fig. 14. Fractured surface of polished steel cleavage specimen.

4.4. Loci of failure


initialised from one of the corners of the adherend, a
Upon visual and light microscopic examination, it point of theoretical singularity. This is in line with the
appeared that the specimens prepared by grit-blasting ndings of Crocombe et al. [18]. On the other hand, in
steel adherends with coarser grit have shown signi- the case of polished specimens, the fractured surfaces of
cantly higher crazing (stress whitening zones due to the joints showed no sign of crazing, and bare steel and
micro-cracking) in the initial area of the joint compared adhesive regions were clearly visible showing an
to those prepared with the ner grit (Fig. 13). In both apparent adhesion failure (Fig. 14). It is difcult to
cases, however, failure was near the interface region and conrm, without using more sophisticated techniques
apparently in a mixed adhesion/cohesion mode. From such as electron microscopy and X-ray diffraction, if the
the intensity of whitening it appeared that the failure bare portions of fractured surfaces are completely free
M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244 243

from adhesive residues or not. However, the results from may be classied as adhesive (adhesion) failure.
the numerical analysis clearly indicate adhesive failure This mode seems to be independent of level of
mode and hence it may be said that failure initiation in adherend surface roughness.
all these cases was adhesive failure (adhesion). These
observations may only apply to specic cases including
loading conditions and type of adhesive and adherend. Appendix A
It can be seen from Fig. 9 that the stresses near the
edges of model 2 are signicantly higher than those in Rlo is dened in ISO 4287 1984. It is the measured
the centre or edges of model 1. However, if a void defect length of the prole surface within the evaluation length,
nds its place at the convex top of model 1 then failure ln i.e. the length obtained if the prole, within the
may initiate from the centre. One way to verify this is by evaluation length, was to be drawn out into a straight
conducting experiments based on the models dimen- line [19]. Mathematically it is represented as follows:
sions. s
Z ln  2
dy
Rlo 1 dx:
0 dx
5. Conclusions
A graphical representation of Rlo is given in Fig. 15.
From the experimental and numerical work under- Rdq is dened in ISO 4287 1997 para 4.4.1. It is the
taken in this study it can be concluded that: root mean square value of the ordinate slope dz=dx
within the sampling length. The mathematical represen-
1. Normal tensile stresses in the case of rough steel tation for this is
surfaces are lower than those in polished ones. This
difference could be as high as 30%. s
Z
1 L
2. A simplied numerical butt model with macro- Rdq yX  y% 2 dx;
roughness provides reasonable degree of correlation L 0
and representation of adhesion in the standard
cleavage joint. where y is the slope of the prole at any given point and
3. Cleavage strength appears to increase with the Z ln
roughness level and prole area of adherends 1
y% yX dx:
surface. ln 0
4. Mode of failure initiation in steel cleavage joints
under quasi-static loading and ambient conditions Graphically, this is explained in Fig. 16.

Fig. 15. Graphical representations of linear prole length, Rlo [19].


244 M. Shahid, S.A. Hashim / International Journal of Adhesion & Adhesives 22 (2002) 235244

Fig. 16. Graphical representation of Rdq [19].

References [10] De Bruyne NA. Aero Research Technical Notes, Bulletin No.168.
Cambridge: Aero Research Ltd., 1958.
[1] Critchlow GW, Brewis DM. Inuence of surface macroroughness [11] Packham DE. Roughness of surfaces. In: Packham DE, editor.
on the durability of epoxidealuminium joints. Int J Adhesion Handbook of adhesion. New York: Longman Group (FE) Ltd.,
Adhesives 1995;15(3):1736. 1992.
[2] Gilibert Y, Verchery G. Inuence of surface roughness on [12] BSI, Determination of cleavage strength of adhesive bonds,
mechanical properties of joints. In: Mittal KL, editor. Adhesive BS5350: Part C1:1986 (1986).
joints formation, characteristics, and testing. New York: Plenum [13] Owens JP, Lee-Sullivan P. Stiffness behaviour due to fracture in
Press, 1982. adhesively bonded composite-to-aluminium joints I. Theoretical
[3] Jennings CW. Surface roughness and bond strength of adhesive. model. Int J Adhesion Adhesives 2000;20:3945.
Am Chem Soc Div Org Chem 1971;31(2):18492. [14] Harris AF, Beevers A. Grit blasting of surfaces for adhesive
[4] Sargent JP. Adherend surface morphology and its inuence on the bonding. Int J Adhesion Adhesives 1999;19:44552.
peel strength of adhesive joints bonded with modied phenolic [15] Thery S, Legros A, Balladon P. Study of parameters
and epoxy structural adhesives. Int J Adhesion Adhesives inuencing the mechanical behaviour of and damage to steel-
1994;14(1):2130. polymer interfaces. In: Baptiste D, editor. Mechanics and
[5] Katona TR, Batterman SC. Surface roughness effects on the mechanisms of damage in composites and multi-materials,
stress analysis of adhesive joints. Int J Adhesion Adhesives ESIS11. London: Mechanical Engineering Publications, 1991.
1983;3(2):8591. p. 33950.
[6] Matsui K. Size-effects on average ultimate shear stresses of [16] Wake WC. In: Eley DD, editor. Adhesion. London: Oxford
adhesive-bonded rectangular or tubular lap joint under tension- University Press, 1961. p. 191.
shear. J Adhesion 1990;10(2):819. [17] Hitchcock SJ, Caroll NT, Nicholas MG. Some effects of substrate
[7] Harris AF, Beevers A. Grit blasting of surfaces for adhesive roughness on wettability. J Mater Sci 1981;16:71432.
bonding. Conf Proc Structural Adhesives in Engineering V. [18] Crocombe AD, Bigwood DA, Richardson G. Analysing structur-
Bristol: Institute of Materials, 1998. al adhesive joints for failure. Int J Adhesion Adhesives
[8] Shahid M, Hashim SA. Cleavage strength of steel/composite 1990;10(3):16778.
cleavage joints. J Adhesion 2000;73/4:36584. [19] ISO 4287. Help le of Ultra Software. Taylor and Hobson,
[9] Sykes JM. Surface treatments for steel. In: Brewis DM, editor. 1984.
Surface analysis and pretreatment of plastics and metals. London:
Applied Science Publishers, 1982. p. 15374.

You might also like