Você está na página 1de 13

The Journal of Strain Analysis for Engineering Design http://sdj.sagepub.

com/

Stress analysis of adhesive-bonded lap joints


R D Adams and N A Peppiatt The Journal of Strain Analysis for Engineering Design 1974 9: 185 DOI: 10.1243/03093247V093185 The online version of this article can be found at: http://sdj.sagepub.com/content/9/3/185

Published by:
http://www.sagepublications.com

On behalf of:

Institution of Mechanical Engineers

Additional services and information for The Journal of Strain Analysis for Engineering Design can be found at: Email Alerts: http://sdj.sagepub.com/cgi/alerts Subscriptions: http://sdj.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://sdj.sagepub.com/content/9/3/185.refs.html

>> Version of Record - Jul 1, 1974 What is This?

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS

R. D. mms

Lecturer in Mechanical Engineering University of Bristol Research Assistant, Department o f Mechanical Engineering, University of Bristol

N. A. Pl%PPIAm

Stresses in a standard metal-to-metal adhesivebonded lap joint are analysed by a two-dimensional finiteelement method and comparisons are made with previous analyses. Particular attention is paid to the stresses at the ends of the adhesive layer. Unlike previous work, which assumes the adhesive to have a square edge, the adhesive s p e w is treated as a triangular fillet. The highest stresses exist at the adherend corner within the spew. Linear elastic behaviour is assumed throughout. A rubber model is reported which confirms these results physically. Good agreement was also obtained between some practical results and the finiteelement predictions.

1 INTRODUCTION

Ideally there should be a uniform shear stress in the adhesive layer of an adhesive-bonded lap joint under tensile load in order to give maximum joint efficiency. Unfortunately, this ideal is rarely achieved in practice because of stress concentrations due to three separate factors: differential straining in the adherends - the shear-lag effect; bending induced by the non-axial loading; end effects caused by the free surfaces at the edges of the adhesive layer. Differential straining in the adherends gives rise to a non-uniform shear-stress distribution in the adhesive, the maximum shear stress occurring at the ends of the overlap; the best-known analysis of this problem is that due to Volkersen (l)*. Demarkles (2) modified Volkersens analysis to take into account the effects of shear deformations in the adherends, and Adams and Peppiatt (3) have shown that significant transverse shear stresses arise in the adhesive layer because of Poissons ratio strains associated with the tensile loading of the adherends. The effect of adherend bending, which gives rise to normal stresses across the adhesive layer, has been considered for the single-lap case by Goland and Reissner (4). However, it should be noted that their solution for the normal stress in the adhesive is incorrectly written and has unfortunately been used in this form by many authors. The corrected form is given in Appendix 1. Sneddon (5) and Kuenzi and Stevens (6) have given the correct version but have not illustrated clearly the magnitude of the discrepancy. Fig. 1 shows the two solutions together with the percentage error of Golands and Reissners values for a range of Youngs moduli for adhesives: it can be seen that
The MS. of this paper was received at the Institution o f Mechanical Engineers on 19th September 1973 and accepted for publication on 29th April 1974. 13 *Referencesare given in Appendix 2.
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

their prediction is considerably in error, especially for lowmodulus adhesives. Volkersen (7) has more recently given expressions for the shear and tensile stresses in the adhesive when the effects of internal bending are considered in a double-lap joint. All of these solutions allow for the effects of non-axial loading but none takes into account the shear deformations of the adherends. Coker (8), experimentally (using photoelasticity), and Inghs (9), analytically, have shown that in a plate subjected to shear loading on two opposite sides there are high tensile and compressive stresses at its comers, the magnitude of these being about four times the applied shear stress and the direction being at right-angles to the sides on which the

60 ! O

Original solution

OO

I L--I1

Youngs modulus o f adhesive

Fig 1. Original solution of Gohnd and Reissner and the corrected solution for the maximum transverse tensile stresses in the adhesive hyer of a single-lap joint plotted against the Youngsmodulus of the adhesive
185

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

R. D. ADAMS AND N. A. PEPPIATT

shear load is applied. These transverse direct stresses arise because the direct and shear stresses acting on the free surface must be zero. Previous lap-joint analyses have considered the adhesive layer as having a square edge (as shown in Fig. 2a) and because of the free surface one would expect similar tensile and compressive stresses to occur in the comers of this layer. Furthermore, the earlier analyses (1) and (2) must be in error in the important region at the end of the joint because they predict a maximum shear stress here, whereas the principle of complementary shears indicates that no such shear stress can exist on the free surface, i.e. at right-angles to the predicted shears. Volkersen (7) has taken these effects into account by using the stress-equilibriumrelation

free surface

ao,+a7,,=,
a y

ax

H i g h tensile and compressive stresses are predicted in the comers of the adhesive layer. These stresses are about 5.5 times the average applied shear stress in a typical lap joint 0.5 in (12.7 mm) long bonded with a high-modulus adhesive. These end-effect stresses must be added to the normal stresses induced by bending. In fact, Volkersens analysis (7) gives solutions for shear and transverse tensile stresses which take both bending and end effects into account. Unfortunately, there are a number of errors in the paper as published, mainly in the expressions given for the boundary conditions of the equations. For comparison later, simplified approaches based on this work have been used. Real adhesive joints are, however, formed with a fdet of adhesive spew which is squeezed out under pressure while the joint is being manufactured. This fdet is shown diagrammatically in Fig. 2b. The assumption that the adhesive layer has a square end is thus unlikely to be realistic. Mylonas (lo), using photoelastic techniques, has investigated the stresses induced at the end of an adhesive layer for a number of adhesive edge shapes and has shown that the position of the maximum stress is dependent on the edge shape. The adherends in h i s model were rigid, and none of the edge shapes studied is typical of the normal spew fillet, where the adhesive flows round, and bonds to, the end of the adherend. A full analysis of lapjoint stresses, which should include the effects of bending, adherend shears, and end effects, has not yet been reported but these factors can be considered by using the finiteelement method of stress analysis. Previous authors have used the finite-element method, but have not considered all three factors. Wooley and Carver (1 1) have written a general-purpose computer programme to describe a single lap joint. They obtained good agreement with the original analytical solution of Goland and Reissner (4), but they have not considered the stresses caused by end effects. The model of Harrison and Harrison (12) is that of a square-edged adhesive layer between rigid adherends (Fig. 2a) and so the effects of spew, bending, and differential straining are not considered. The purpose of this work was to investigate analytically by the finite-element method joints with either square edges or spew fdets and to compare the results with the theoretical solutions of previous authors. The models are made up of constant-strain triangular elements and only Hookean adhesives and adherends are considered. This is a realistic assumption for some high-strength adhesives at
186

F i g . 2. Diagrammatic lap joints to show adhesive layers with


(a)square edge and ( b )spew fillet

room temperature. A silicone-rubber model, developed from the models reported by Adams et al. (13), is also introduced to show the effect of the spew filet physically. Finally, some practical consequences of the results are given.
2 FINITE-ELEMENT MODELS OF LAP JOINTS Because of stress gradients across the adhesive layer and because of possible high stress gradients along the adhesive layer, sufficient resolution is only likely to be obtained if a considerable number of elements is used. However, although Adams and Peppiatt (3) showed that the lap-joint problem is three-dimensional, they also showed that the shear and tensile stresses in the direction of the applied load are not significantly influenced by the transverse stresses caused by Poissons ratio strains in the adherends. Since a high matrix bandwidth is inherent in the use of threedimensional elements, we decided to formulate the problem using simple constant-strain twodimensional triangular elements to give stresses at their centroids. This decision was further influenced by the fact that many adhesives are non-linear, and further work would require a finiteelement programme t o take into account plastic deformations. Such a programme is available internally for non-linear analysis using the simple triangular element. As the joint is wide compared with its thickness, the h i s problem is considered as one of plane strain. T assumption should be satisfactory for the adhesive layer, but less so for the adherend. Unless otherwise stated, the lap joints studied are of standard lap-joint testpiece size (0.5 in, 12.7 mm, long by 1 in, 25.4 mm, wide) and are made of 16 gauge aluminium (1.62 mm thick) bonded with a high-strength modified epoxy-resin adhesive of Youngs
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS

modulus 7.0 X lo5 lbf in-' (4.82 GNm-'), shear modulus 2.5 X lo5 lbf in-' (1.72 GNm-'), and a glue-line thickness of 0.010 in (0.25 mm). In the double-lap joints, the centre adherend is twice the thickness of the outer adherends (i.e. 0.128 in, 3.24 mm). The modulus values were obtained from unpublished work by the authors on AF-130 adhesive which was found to have a substantially linear stress-strain relation to failure. Both double-lap and single-lap joints have been treated. The double-lap joint has been considered in two ways as shown in Figs 3a and 3b. The first model uses symmetry about the centre-line of the double lap and lateral restraints are applied along this centre-line. The second model uses the fact, revealed by simple analysis, that, if the adherends are equally thick, the loads in the adherends are equal at the centre of the overlap and the double lap can be represented by a 'half-length' model, with consequent economy of computing effort. Because of bending effects, restraints must t e applied in two ways to obtain the stresses at the end where the adhesive is in tension and at the end where it is in compression. The load is applied uniformly across the adherends at the mid-section of the joint (Fig. 3b). These boundary conditions are exact if the overlap length is long enough for the shear stress at this point to be zero. For the case considered here, this condition is by no means fulfilled, and these boundary conditions are therefore approximate. However, Saint Venant's principle suggests that the stress distribution in the outer quarters of the overlap length, as obtained from this model, was correct. Single-lap results were obtained from the full-length lap model; the constraints used are shown in Fig. 3c. The spew was approximated to a 45" triangular fillet of varying size. To find the effect of the spew size on the stress distribution the half-length double-lap type model was used in order t o reduce computing time. A summary of the models is given in Table 1. One disadvantage of the finite-element method is the time required to prepare the element data for different joint geometries. A programme was therefore written which

computed the geometrical data from the following parameters: adherend length; overlap length; spew size (within the limits 0.015 in, 0.38 mm, to 0.9 times the adherend thickness); adhesive-layer thickness; adherend thickness. Thus a joint of differing geometry only requires the changing of a few data cards and it is equally easy to vary the elastic properties of the adhesive and adherends. This model can be used to obtain results for both single- and double-lap joints by means of suitable nodal constraints. The mesh used for the full-length lap joint with spew is shown in Fig. 4 to three different scales to enable the mesh to be seen in perspective and with clarity.
3 RUBBER MODEL TO SHOW THE EFFECT OF SPEW Previous rubber models for lap joints as reported by Adams et a L (13) show the non-uniform shear-stress distribution in an adhesive along a lap joint, but d o not have sufficient resolution to show deformations caused by the high, stress gradients at the ends of the adhesive layer. Accordingly, a simplified model has been designed for an experimental investigation of end effects. The model is shown under load in Fig. 5 and the undeformed outline is drawn in Fig. 6. The part representing the adhesive layer was cast in silicone rubber and has both a 45" fillet to represent the adhesive spew and a square edge. The silicone rubber was cast in a mould of which the steel adherends formed the sides (except on the two free surfaces). The adherends were treated with a suitable primer before they were cast to ensure a good bond; the 'free' surfaces of the mould were coated with a release agent. Because of the difficulty of marking the silicone rubber with a grid, V-grooves were machined into the surface of the base of the mould so that a raised orthogonal grid of silicone rubber was created directly on the side face of the 'adhesive'. The effects of differential straining and adherend

.
1
I : '

--c

bF

a Full-length double-lap joint. b Half-length double-lap joint.

c Single-lap joint.

Fig 3. Constraintsfor finiteelement models of lap joints


JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974
Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

187

R. D. ADAMS AND N. A. PEPPIATT

Table 1. Table of jiniteelement models


Type of model Half length Half length from adherend Half length from adherend Half length Full length Full length Half length Double-lap Double-lap Single-lap Double-lap Large spew No spew. Square ended No spew. Square ended Large spew. Adherend corners with 0.016 in radius Spew from adherendend
0.074 in (1.9 mm) 0.010 in (0.25 mm) 0.010 in (0.25 mm) 0.020 in (0.5 mm) 0.010 in (0.25 nlm)

Type of joint Double-lap

Spew shape No spew. Square ended

Spew height

Glue-line thickness
0.010 in (0.25 mm)

Number of nodes

Number of elements
153

96
171

0.010 in (0.25 mm)

0.010 in (0.25 mm)

II

lo3
135 231

I
149
432 432 117 256 754 754 202

0.074 in (1.9 mm)

0.010 in (0.25 mm) 0.010 in (0.25 mm)

General purpose full length

Double-lap or single-lap

Generally 0.040 in (1.0 mm) but variable

Variable

~~

562

338

The section between the lines is shown enlarged below

Fig. 4. Finite-element mesh for full-length lap joint with spew


188
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1914 OIMechE 1914
Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS

The black crosses are the finite-element predictions of the intersections of the grid lines of the model.

Fig. 5. Comparison between calculated and experimental displacements of the silicone-rubber model

bending are not represented on this model because the adherends are very much stiffer than the adhesive. Restraints could be appljed to prevent the rubber from bending owing to rotation of the adherend in the plane of the grid face when loading was applied. Comparison with the finite-element results was obtained by comparing the grid deformations for a given displacement at the loading point of the model with the nodal displacements at corresponding points given by the finite-element method for the same displacement. A graph plotter drew the positions of the displaced nodes with vertical crosses and the plot was reproduced on a transparent film. The film was placed over the model so that a direct comparison could be made between the finiteelement deformations and those of the model (see Fig. s). The experimentally determined Youngs modulus of the silicone rubber used in the model was 2841bf in-2 (1.96 MNm-) and the shear modulus was 97 lbf in- (0.67 MNm-). These values were obtained from the initial region of the stress-strain curve where the relation is approximately linear. Similar stress levels were used in the tests on the model.
4 DISCUSSION

Fig. 6 shows the principal-stress pattern obtained by the finite-element analysis. The length and direction of the lines represent respectively the magnitude and direction of the principal stresses at the centroid of each finite element. A bar at the end of the line implies a negative principal stress, i.e. compressive. It is evident that the presence of the fillet causes the stress pattern to differ significantly from the pattern at the end with no fillet. At the points Al and A1 the high tensile and compressive stresses predicted by Inglis ( 9 ) are shown, the absolute magnitude of the largest elemental principal stresses being at least 3.6 times the shear stress in the rubber between the plates. This is of similar size to the value predicted by Inglis who says that the normal stress is more than four times as large as the applied shear stress. It should be noted that the rubber away from the ends of the steel plates is in pure shear, as is shown by the equal and opposite principal stresses in the elements in this region. The stresses in the fillet are predominantly tensile, the maximum stress concentration at this end being at the sharp corner B. The maximum stress is at least 3.5 times the shear stress in the rubber between the plates.
4.2 The square-ended adhesive layer Fig. 7 shows the shear-stress distributions in the adhesive layer of a double-lap joint as predicted by the following six analyses.

OF RESULTS

4.1 Results from rubber model

Fig. 5 shows that close agreement was obtained between the deformed grid of the silicone-rubber model and the nodal displacements predicted by the finite-element analysis. As surface displacements of the model are used for comparison, the finite-element results were obtained with plane-stress stiffnesses. The errors were principally due to the difficulty of applying a restraint in the correct position while still allowing a photograph to be taken of the whole model.
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OlMechE 1974

The simple Volkersen analysis (1). A modification of the Volkersen analysis to take into account the fact that the shear stress must be zero at the ends. This is based on Volkersens more recent work (7). Demarkless modification of Volkersens original analysis to take into account adherend shears (2). 189

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

R. D. ADAMS AND N. A. PEPPIATT

Fig. 6. Principal-stress pattern for rubber model, showing end effects

6th-2

2 Volkersen's analysis (7), allowing for end effects. 3 Analysis of Demarkles (2). 4 Volkersen's analysis (7), allowing for bending. 5 Full-lengthdoublelap finiteelement modeL 6 o Points from half-length finte-element model.

--

- - -- - --

--

'I
1I I
I
1

Fig. 7. Shear-stress distributions from double-lap theories and finite-element models

(4) An analysis allowing for the effects of bending of

the adherends in a double-lap joint derived from the equations given by Volkersen (7). ( 5 ) Finite-element results from the full-length lap model. (6) Finite-element results from the half-length lap model. The finiteelement results should take into account end effects, bending of the adherends, and shear straining of the adherends Comparing results (1) and (2), one can see that, if end effects are neglected, the requirement that shear stress in the adhesive should be zero at the joint ends only affects the shear-stress distribution significantly for about the end
190

4 per cent of the overlap length. Curves (3) and (4) show that the maximum shear stress predicted in the adhesive layer is reduced both by the shear deformation of the adherends and by the reduction of adherend strain at the adhesive-adherend interface caused by the bending of the adherends There is good agreement between the two curves from the finite-element results ( 5 ) and (6) except at the centre of the joint. The variation around the mid-point is due partly to the free surface present at the centre of the half-length lap model where the shear stresses must be zero, and partly to the fact that' the loading was only approximate. However, the good agreement at the ends shows that the half-length lap model is valid in t h i s region for the combination of adhesive and adherend properties considered. Problems do arise if the adhesive and adherend
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS

shear moduli are of the same order, but this is not usual in practice. The stress distributions from the finiteelement results are closest to the solutions taking into account the shear deformations of the adherends and the bending of the adherends, which shows that these solutions give a reasonable representation of the shear-stress distribution within the lap. However, each neglects a different but important aspect of the analysis, so the correlation may be somewhat fortuitous. The tensile stresses across the adhesive layers of a double-lapjoint from analyses (4), (5) and (6), are shown in Fig. 8. There is reasonable agreement between the two finite-element solutions, half-length lap and full-length lap in the outer quarters of the joint. However, there is a marked difference between the finite-element curves and the curve obtained from the analytical solution (4), especially at the end of the joint where the adhesive is in compression: here the finite-element soh tions predict a compressive-stress concentration about half that given by
(4).

suitability of this particular finiteelement method for analysing the stresses in an adhesive-bonded lap joint.
4.3 End effects

The shear-stress distribution in a single-lapjoint is shown in Fig. 9 and the transverse direct-stress distribution in Fig. 10. Finite-element results from a full-length single lap with no spew are compared with the solution of Goland and Reissner which is the only analytical solution which really applies to the single-lap case. There is good agreement between the two for the shear stresses (Fig. 9). In Fig. 10 the general shapes of the normal-stress distribution curves are similar, although the tensile-stress concentration at the ends of the overlap is predicted to be higher by the finiteelement solution. Because of the high stress gradient, the actual stress concentration is difficult to determine. The curves of the finite-element solutions taking spew into account will be discussed later. The good agreement between the theoretical and the finite-element results for both the single- and double-lap joints, and the agreement between the silicone-rubber model and its finite-element representation show the

The stress pattern at the end of a squareedged adhesive layer is shown in Fig. 11. This plot was obtained from the tension end of a half-length double-lap joint model. This is again a plot of principal stresses, the interpretation of which is given in the discussion of the rubber model (section 4.1). The highest tensile stress exists at the comer of the adhesive adjacent to the loaded adherend and represents a stress concentration of at least 10 times the applied shear stress on the joint. It should be noted that, because constant-stress elements are used and the stress gradients are high, it is impossible to determine the stresses just below the surface very accurately without using infinitesimally small elements: the stresses acting on the surface are, of course, zero. The effect of bending of the outer adherends modifies the stress distribution from that obtained in the simple rubber model, the absolute value of the largest principal stress at this end being about four times the absolute value of the largest principal stress at the other side of the adhesive layer (compare points A, and A2 in Fig. 6). The effect of the spew on the stress pattern is shown in Fig. 12, which is at the tension end of a double-lap joint. The spew is represented by a triangular fillet 0.020 in (0.5 mm) high. It can be seen that, because of the predominance of the major principal stress, the adhesive at the ends of the adhesive layer and in the spew fillet is essentially subjected to a tensile load at about 45" to the axis of loading. The highest stresses occur within the spew at the corner of the unloaded adherend, the presence of the 90" comer introducing a stress-concentrating effect. As the maximum stress occurs within the spew and not at or near the adhesive surface, it is unlikely that the approximation to the spew shape by the triangular fdet has a significant effect on the stress distribution. The stress pattern shown in Fig. 12 suggests that the area

4D

:F
32/

.
5
y1 Ln

u -

4 Volkersen's analysis (7), allowing for bending. 5 Full-length doublelap finiteelement model. 6 o Points from half-length finiteelement model

JI I

- - -(
4

l-

' .

-44

-J
Fig. 8. Transverse normal-stress distributions from double-lap theory and finite-element models
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1914 OlMechE 1914
Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

191

R. D. ADAMS AND N. A. PEPPIATT

12r
lo

----x--

Finite-element model, squore end Goland and Reissner solution (colculated for 4.5 kN load Finite-element model with spew

'I
1
I

..
I

Averoge applied shear stress

8 6 7 Distance olong overlap mm

10

11

12

13

Fig. 9. Shear-stress dism'butions from finire-elemenrmodels with and without spew, and from Goland and Ressner (4)

-4L

Fig. 10. nanmerse normal-stressdistribution fiom finite-element models with and without spew, and from Goland and Reissner (4)
of transfer of load between the adherends is effectively lengthened. Fig. 13 shows the average shear-stress distributions across the adhesive layer obtained for joints with varying sizes of spew. The distributions are shown dotted outside the overlap length to show an area of load transfer within the spew. As the loading in the spew is predominantly tensile, a meaningful distribution of shear stress is difficult to obtain. All curves, except the curve for the 0.040 in (1.O mm) spew, were obtained from half-length lap solutions. The stress distribution for the square-ended cases is shown for comparison. The values for the 0.040 in
192

(1.0 mm) spew were obtained from the full-length lap datageneration programme. The maximum shear stress obtained from the largest spew size, which extends completely across the ends of the adherends, is 70 per cent of that obtained for the squareended adhesive layer. A spew of 0.040 in (1.Omm) height gives a shear-stress reduction of 15 per cent. It can be seen that, when the effects of spew are taken into account, the maximum shear stress obtained is considerably less than that predicted by the earlier analytical theories. Figs 9 and 10 compare the distributions of longitudinal
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS


/
/

Fig. 11. Finite-element prediction of the principal-stress pattern at the end of a square-edged adhesive layer

/. I/

occurs in this manner in normal, well-bonded joints. (In adhesives terminology, adhesive failure means the destruction of the bond between adherend and adhesive; cohesive failure means a fracture wholly within the adhesive material.) If, however, the boundary of the adherend and adhesive is weak, the spew is not cracked but is pulled away from the loaded adherend surface by the tensile stresses in the spew. It should be noted that complete removal of the spew by machining would be difficult without machining either of the adherends, so that some spew, similar in size to the glue-line thickness, is likely to be left at the end of any joint. Moreover, machining may initiate cracks in the adhesive, especially if it is one of the more brittle high-tempers ure adhesives. Thus an adhesive layer with a square edge is not only undesirable but is unlikely, and even difficult to obtain in practice.

Fig. 12. Finite-element prediction of the principal-stress pattern at the end of an adhesive layer with 5 mm spew
shear and of transverse direct stress (i.e. across the adhesive layer, not principal) within the overlap length for a single-lapjoint having a 0.040 in (1.0 mm) spew height with the distributions when the adhesive is square-ended. The maximum shear stress predicted is reduced to 70 per cent of that given by the Coland and Reissner theory. For the tensile stress, the analytical prediction lies between the two finite-element curves at the ends of the lap, but the maximum transverse tensile stress when spew is present is reduced to 80 per cent of that given by Coland and Reissner.
4.4 Practical aspects

Reduction of stress concentration One of the results of the finite-element analysis is that the highest principal stresses in the adhesive layer of a lap joint occur in the region of the corner of the unloaded adherend. This suggests that it may be possible to increase joint strength by radiusing the corner of the adherend. A finiteelement model with a 0.016 in (0.4 mm) radius was analysed and the maximum principal stress in the spew was shown to be 40 per cent less than that obtained with a right-angle corner. . 5 in (12.7 mm) lap joints Standard 1 in (25.4mm) by 0 were prepared with L73 aluminiumalloy adherends, half of which had a 0.016 in (0.4 mm) radius hand-reamed on the adherend comer. The surfaces were prepared by a standard etching procedure and bonded with AF130 adhesive. The joints were then tested in tension in accordance with ASTM D1002-64.The strengths obtained are given in Table 2 . No significant increase in strength was obtained, the small improvement (less than 3 per cent) with rounded comers certainly not being of the size predicted by the h i s is that etching the finite-element model. The reason for t adherends produces a radius on the adherend comer, even on the nominally rectangular edges. The finiteelement model was therefore modified so that the adherend had a small chamfer (0.001 in X 0.001 in, 0.025 mm X 0.025 mm) at its comer. The maximum adhesive stress predicted was now 30 per cent less than that obtained when the adherend corner was square. Thus, to obtain a more representative picture of the maximum stresses likely to be obtained in a lap joint, the adherends are now always treated as having a small chamfer on the comer within the spew. Comparison of double- and single-lap results The effect of greater bending in the single-lap joint appreciably affects the magnitude, but not the direction, of the principal stresses within the spew. There is an 80 per cent increase in the magnitude of the maximum principal stresses in the single-lap joint over those in the double-lap joint for the same load applied to the adhesive layer. Tests on a number of single- and double-lap joints made with AF130 adhesive give the ratio of maximum joint loads, i.e. (failure load of a single-lap joint)/(half-failure load of double-lap joint), as 0.52. The finiteelement results for a 0.010 in (0.25mm) bond-line thickness indicate the stress-concentration ratio, 193

Direction of cracks in failed joints It has been observed that in the spew of aluminium to aluminium joints bonded with low-ductility adhesives, cracks are formed approximately at right-angles to the directions of the maximum principal stresses predicted by the finite-element analysis. In general, these cracks run close to the corners of the adherends. The region where cracks are formed in the spew is indicated in Fig. 12. These observations give weight to the view presented here that failure in a lap joint is initiated by the high tensile stresses within the spew. The cohesive failure of the adhesive
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

R. D. ADAMS AND N. A. PEPPIATT

u Square edge.. b 0.25mm spew (equals glueline thickness). c 0.50 m m spew. d 1 m m spew (full-length solution) e Adherend

thickness spew

Distance from centre of overloe


mm

Fig. 13. Influence ojspew size on shear-stress distribution at tension end of double-lap joint

Table 2. Strength of round- and square-cornered joints


Corner Average strength Maximum strength Minimum strength Round (ninejoints) 0.556 tonf (5.53 kN) 0.608 tonf (6.05 kN) 0.486 tonf (4.83 kN) Square (eight joints) 0.540 tonf (5.37 kN) 0.575 tonf (5.72 kN) 0.490 tonf (4.87 kN)

i.e. (maximum principal stress in double-lap joint)/(maximum principal stress in single-lapjoint), as 0.54. The stress-concentration values were obtained from the average of four adjacent elements in and near the chamfer and are based on the increase over the applied average shear stress. The agreement between the results is good, indicating that the principal stresses in the spew are the cause of joint failure. It is unrealistic to compare the earlier analytical theories in this way since these are known not to allow for the effect of spew.
194

Effect of glueline thickness on m x i m u m stresses The strengths of lap joints of epoxy adhesives do not vary significantly over the range of adhesive-layer thicknesses used in practice. A plot of lap-joint strength against glue-line thickness for BSL308 adhesive, as presented by Bennett (14), is shown in Fig. 14. The stresses predicted by most stress analyses, however, are very dependent on glue-line thickness. Generally, if all other joint parameters are constant, the maximum stress predicted is proportional to the reciprocal of the square root of the glue-line thickness, provided the thickness is sufficiently small. Thus, it would be expected from these theories that joint strength increases as the glue-line thickness is increased. It is for t h i s reason that Adams et a1 (13) and Semerdjiev (15), using a shear-failure criterion, have proposed that joints with profded adherends should be used to give a varying glue-line thickness. Fig. 14 shows the strengths predicted by a number of analyses over a range of glue-line thicknesses, based on the maximum strength of the adhesive calculated for the failure load for a joint with an adhesive-layer thickness of 0.005 in
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 OIMechE 1974

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

STRESS ANALYSIS OF ADHESIVE-BONDED LAP JOINTS

01

02

05

04

05

Glue-lint lhicknm mm

1 Demarkles (2) (shear stress). 2 Volkersen (7) end-effect solution (tensile stress). 3 Coland and Reissner (4) (tensile stress). 4 Finiteelement single-lap (principal stress). 5 BSL 308 adhesive.

Fig 14. Influence of glueline thickness on the predicted and actual lapjoint strengths
(0.13 mm). The following analyses are considered.

(1) Demarkles's analysis (2) which gives the maximum

shear stress only. (2) An analysis, based on Volkersen's work (7), for the maximum tensile stress caused by end effects. (3) The analysis of Goland and Reissner (4). The corrected maximum tensile stress is used here, (4) Finite-element analysis of a single-lap joint with spew, considering the maximum principal stresses in the region of the chamfer. The analyses (1)-(3) a l l predict a significant increase in strength by increasing the glue-line thicknesses over the range shown. The least increase is given by Demarkles's shear-stress analysis. On the other hand, the fmiteelement analysis is reasonably close to the curve shown for the BSL308 adhesive, although the analysis still predicts a slight increase in strength as the glue-line thickness is increased. The decrease in strength of the actual adhesive joint, however, as glue-line thickness is increased, may be caused by factors which cannot easily be taken into account by any predictive technique. Niranjan (16) suggests that there are three separate factors, dependent on glue-line thickness, which influence the strength of a lap joint: stress concentration; the porosity of the glue line; the strain rate. The results of the finite-element method suggest that the strain-rate factor is not important because the maximum stress, and hence the maximum strain rate for a given crosshead speed, does not vary significantly with glue-line thickness. It is thus concluded that the decrease in strength of actual joints, as the glueline thickness is increased, is caused by the increase in porosity and the number of microcracks in the adhesive.
5 CONCLUSIONS

It has been shown that there is reasonable agreement between the stress distributions obtained for an adhesivebonded lap joint by the finiteelement method and classical analytical techniques, the best agreement being obtained with those theories making the most realistic assumptions. The finite-element method has been used to give shear, transverse normal, and principal-stress distributions in both single- and double-lapjoints. Most of the classical analytical techniques do not allow for end effects and none can predict the stresses in the adhesive spew. With the assumption that the joint materials are linearly elastic, good agreement with practical results has been obtained. The maximum principal stresses at the ends of the adhesive layer are predicted to be at right-angles to the direction of cracks formed in the spew of failed lap joints. The etching process in the preparation of aluminium adherends leaves a small radius at the adherend comers which relieves part of the stress concentration. The finiteelement method shows that if the adherend is assumed to have a small chamfer on its comer due to the etching process, good agreement between the strength ratios ahd the predicted maximumstress ratios of single- and double-lap joints is obtained. The influence of glue-line thickness on the maximum stress within the joint is more realistically predicted than it w a s by earlier theories. The classical, closed-form analytical results for adhesivebonded lap joints are limited by the assumptions made in order that a solution may be obtained. Moreover, the effects of spew or etching cannot be included because of the complexity of the geometry. However, it is relatively easy to consider these effects by using a fmiteelement technique and it has been shown that it is most important to include them if realistic stress distributions are to be obtained, particularly in the highly stressed areas at the ends of the joint.
APPENDIX 1

The expression given by Goland and Reissner (4) for the tensile stresses across the adhesive layer is
a. = @ C'AI -[(R2
I

h 2 t + Xk'coshXcosX

whereas the corrected expression, given in similar form by Sneddon ( S ) , is

- M' coshXcosh

h x x
C

Similarly, the expression for the maximum transverse tensile stress must be corrected to give
(ao)max

-pt2

The stresses in adhesivebonded lap joints of standard testpiece size have been examined two-dimensionally. Plane strain has been a p m e d and constant-strain triangular finite elements have been used.
JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1914 QIMechE 1974

-7

1 Xz5
r

(sinh2h-sin2X) 2 (sinh2X 4- sin2X)


1

(cosh2X + C O D A ) - Ak' (sinh2h + sin2X)

195

Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

R. D. ADAMS AND N. A. PEPPIATT

APPENDIX 2
REFERENCES
(1) VOLKERSEN, 0. Die Nietkraftverteilung in Zugbeanspruchten Nietverbindungen mit Konstanten Laschenquerschnitten, Luftfarhtforschung 1938 15, 41-7. f the use of a rubber (2) DEMARKLES, L. R. Investigation o analog in the study of stress distribution in riveted and cemented joints, Tech Notes natn. advis Comm Aeronaut., Wash 3413 1955. (3) ADAMS, R D. and PEPPIAIT, N. A. Effect of Poissons ratio strains in adherends on stresses of an idealized lap joint, J. Strain Analysis 1973 8,134-9. (4) GOLAND, M. and REISSNER, E. Stresses in cemented joints, J appl Mech, nuns A m SOC mech Engrs 1944 66, A17-A27. (5) SNEDDON, 1. The distribution of stress in adhesive joints, Adhesives (ed. Eley, D.) 1961 chapter 9 (Oxford University

press).
(6) KUENZI, E. W. and STEVENS, C. H .Determination of the mechanical properties of adhesives for use in the design of

bonded joints, U.S. Forest Products Laboratory Report FPL-011,1963.

(7) VOLKERSEN, 0. Recherches sur la thiorie des assemblages colles, Construction mktalliaue 1965 (No. 4). 3-13. (8) COKER, E . C. An optical determination of the variation of stress in a thin rectangular plate subjected to shear, h c R. Soc Series A 1912 86,291-319. (9) INGLIS, C. E. Stress distribution in a rectangular plate having two opposing edges sheared in opposite directions, h c R. Soc Series A 1923 103,598-610. (10) MYLONAS. C. Experiments on composite models with application to cemented joints, h c . SOCexpl Stress Analysis 1954 12,129-42. (11) WOOLEY, G . R and CARVER, D. R Stress concentration factors for bonded lap joints, J. Aircr. 1971 8,817-20. (12) HARRISON, N. L. and HARRISON, W. J. The stresses in an adhesive layer, J. Adhesion 1972 3, 195-212. (13) ADAMS, R D., CHAMBERS, S. It, DEL STROTHER, P. J. A . , and PEPPIATT, N. A. Rubber model for adhesive lap joints,J. Strain Analysis 1973 8,52-7. (14) BENNETT, W. F. on-destructive testing of adhesive bonding, Paper presented at the 9th Annual Conference on Non-Destructive Testing, Loughborough University, 1972. (15) SEMERDJIEV, S. Metal to metal adhesive bonding, 1970 (Business Books, London). (16) NIRANJAN, V. Bonded joints a review for engineers, 1970 UTIAS Review No. 28 (University of Toronto).

196

JOURNAL OF STRAIN ANALYSIS VOL 9 NO 3 1974 QIMechE 1974


Downloaded from sdj.sagepub.com at UNIVERSITAETBIBLIOTHEK on April 23, 2013

Você também pode gostar