Journal of Adhesion Science and Technology 25 (2011) 819836

brill.nl/jast

Single Lap Joints with Rounded Adherend Corners: Stress and Strain Analysis
X. Zhao a , R. D. Adams a,b and L. F. M. da Silva c,
a

Department of Mechanical Engineering, University of Bristol, Queens Building, University Walk, Bristol BS8 1TR, UK b Department of Engineering Science, University of Oxford, Parks Road, Oxford, UK c Departamento de Engenharia Mecnica, Faculdade de Engenharia, Universidade do Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal Received in nal form 3 June 2010

Abstract One of the major difculties in designing adhesive lap joints is the stress singularity present at the adherend corners at the ends of the overlap. One way to overcome this problem is to assume that the corners have a certain degree of rounding. The objective of the present study was to better understand the effect of the change in the geometry of the adherend corners on the stress distribution and, therefore, on the joint strength. Various degrees of rounding were studied and two different types of adhesives were used, one very brittle and another which could sustain a large plastic deformation. The study gives a detailed stress and strain distribution around the rounded adherends using the nite element method. The major nding is that the stresses or strains in the adhesive layer of a joint with rounded adherend corners are nite. In real joints, adherends generally have small rounded corners. Consequently, the model with small radius corners may be used to represent real adherends. Koninklijke Brill NV, Leiden, 2011 Keywords Single lap joint, epoxy, nite element stress analysis, adherend rounding

1. Introduction It is well known that there are stress singularities at the re-entrant corners at the ends of the overlap of adhesive lap joints [1]. Although analytical methods can predict stresses inside the adhesive layer accurately for most of the overlap, they fail to give the stresses at the ends of the overlap accurately. Analytical and experimental analyses show that it is at the ends of the overlap where failure starts. As a result,
* To whom correspondence should be addressed. Tel.: +351225081706; Fax: +351225081445; e-mail: lucas@fe.up.pt

Koninklijke Brill NV, Leiden, 2011

DOI:10.1163/016942410X520871

820

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

the stresses at the ends of the overlap are critical for joint failure. These stresses are affected by the material properties of both the adherends and the adhesive. It has been shown [24] that the stiffer the adherends, the lower are the stresses for square-ended joints. Also, the stresses are reduced if the adhesive is more exible [5]. The critical stresses may also be affected by the local geometry at the ends of the overlap [6, 7]. In general, adhesively-bonded joints have some adhesive squeezed out from the adhesive layer, forming llets at the ends of the overlap. Because these llets are near the critical regions in the joints, it has been shown that they affect the local stresses near the ends of the overlap signicantly [6, 818]. This effect can only be analysed by the nite element (FE) method. The bondline thickness has been shown to have only a little effect on the joint strength in the range of 0.10.4 mm [1, 19]. Another factor affecting the critical stresses at the ends of the overlap is the geometry of the adherend corners. If the corners are sharp, the stresses are singular (innite). In that case, a fracture mechanics approach is more appropriate such as that proposed by Groth [20] based on the strength of the singularity. However, it has been shown that stresses are signicantly reduced by rounding the adherend corners. Adams and Harris [6] demonstrated theoretically and experimentally that the strength of single lap joints with rounded adherends with a toughened adhesive increased substantially compared with joints with sharp adherend corners. Furthermore, the adherend corners are usually not sharp in practice. There is, in general, a small amount of rounding at the adherend corner due to the production process. This may affect the stress distribution in the region of the adherend corner and, therefore, the joint strength, because stresses in this area are very sensitive to the change in the geometry of the adherend corners. In order to have a better understanding of the effect of the change in the geometry of the adherend corners on the stresses and, therefore, on joint strength, this paper is a further study on the effect of rounded adherend corners on the stresses and strains in a single lap joint. 2. Geometry and Material Properties 2.1. Geometry The geometry of the single lap joints analysed in this study is shown in Fig. 1, where the width of the joints is 25 mm. The adhesive thickness was 0.25 mm and the overlap length was 25 mm. The adherend length outside the overlap was 75 mm. Four geometries of adherend corners were analysed: sharp, small radius, medium radius and large radius. Sharp corners exist only in theory. In practice, the radius varies but it is not larger than the large radius studied here. 2.2. Adherends All the adherends used were aluminium alloys. As can be seen from Fig. 2, the stressstrain curve is almost linear after yielding. For modelling adherend plastic deformation, a straight line was drawn beyond yielding, which was believed to be

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

821

Figure 1. Aluminium/epoxy single lap joints with different degrees of rounding. (a) Joint geometry (all dimensions in mm). (b) Sharp corner. (c) Small radius. (d) Medium radius. (e) Large radius.

accurate enough to represent the stress hardening part of the aluminium for the current purpose. The Youngs modulus determined from the stressstrain curve is close to 68 GPa. To verify this test, a dynamic test was also carried out to measure the Youngs modulus. Dynamic tests to measure Youngs modulus and Poissons ratio have been widely used and shown to be accurate [21, 22]. This dynamic test gave a slightly higher value of Youngs modulus of 72.3 GPa, so an intermediate value of 70 GPa was used in the analysis. A value of 0.33 was used for Poissons ratio [23]. The main properties of the aluminium alloy used are summarised in Table 1. 2.3. Adhesives Two adhesives were used in the present work: Ciba-Geigy MY750 (a diglycidyl ether of bisphenol A) with hardener HY906 (anhydride) and the same adhesive but rubber-toughened with carboxyl terminated butadiene nitrile (CTBN) rubber. These adhesives are designated as MY750 and CTBN, respectively, in this work and Fig. 3 shows their uniaxial tensile behaviour. It should be noted that adhesive MY750 with hardener HY906 is brittle, so it has been treated as a linear elastic material, while CTBN is a ductile adhesive. The main properties are summarised in Table 1.

822

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 2. Tensile stressstrain curve for aluminium adherends. Table 1. Material mechanical properties Adhesive MY750/HY906 Youngs modulus (GPa) Poissons ratio Yield stress (MPa) Failure stress (MPa) Failure strain (%) 2.8 0.4 84 3 Adhesive CTBN 2.5 0.37 40 62 14.8 Aluminium 70 0.33 250 323

3. Analysis Methods In this study, the FE method was used for both linear and non-linear analyses. To simulate the test conditions, both ends of the joints were constrained so as not to rotate, as shown in Fig. 1(a). Provided the adherends are long enough, these boundary conditions have little effects on the stress distribution within the adhesive with large displacement models. A 2D plane strain model was used with 8-noded isoparametric elements for both the adherends and the adhesive. Two steps were involved in analysing the corner areas to obtain accurate stress or strain distributions. The rst run was based on a quite coarse mesh for the whole joint, followed by a ner mesh around the adherend corner with the displacements from the rst analysis

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

823

Figure 3. Tensile stressstrain curves for adhesives MY750 and CTBN.

used as boundary conditions. In the linear analysis, a large displacement model was used with both adherends and adhesive being linearly elastic. In the elasticplastic analysis, both the adherends and the adhesive were allowed to yield as well as to experience large displacements. Analysis was performed for the four different degrees of adherend rounding shown in Fig. 1(de). In the analyses with a coarse mesh, six uniform elements were used across the adhesive layer thickness and across each of the adherends. In the subsequent rened analyses, ten graded elements were generated across the adhesive layer. To avoid large aspect ratios in the adhesive layer, 50 elements were used along the overlap in the rened analysis. However, such a number of elements along the overlap caused the aspect ratios in the adherends to be high. Fortunately, there was no yield in the adherends near the corners so that the large aspect ratios were still acceptable with the 8-noded isoparametric elements used. The rened mesh for the stress analysis in the sharp adherend corner region is shown in Fig. 4. Similar meshes were used for the other rounded adherend corners. Incremental solutions with a series of loads were performed for both linearly-elastic and elasticplastic analyses. In the elasticplastic analysis, a modied von Mises [24] yield criterion was included, which takes into account the effect of hydrostatic stress components on the yielding of the adhesive. 4. Results 4.1. Linear Elastic Analyses with Adhesive MY750 The peel (yy ), shear (xy ), longitudinal (xx ) and maximum principal stresses in the adhesive were plotted at the Gauss points close to the unloaded adherend at a distance of 0.001 mm from the adherend as shown in Figs 58 for the four types

824

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 4. A typical ne mesh around the adherend corner.

of joints. It may be seen that almost no difference exists for most of the overlap, except near the adherend corners for joints with a small radius and sharp adherend corners. This is because the stresses are not dened at the sharp corners and the small rounding is very localised to the adherend corner. The stresses should, in theory, be innite at the sharp corners [6, 25]. However, stresses from this analysis were nite because the mesh used here was of nite size. To verify the existence of the stress singularity, Fig. 9 shows a logarithmic plot of the stresses against the logarithmic distance from the adherend corner across the adhesive layer thickness. The almost linear relationship in the gure indicates the singular nature of the distribution of the stresses. However, it should be noted that the stresses in the nal element adjacent to the singular point are not reliable because a singular element was not used there. The singular nature of the stresses covers almost the whole adhesive layer as seen from Fig. 9. Another feature of the stresses with the small rounded corner is that a larger area with high stresses exists than with sharp corners in which the stresses are high only in the very small area close to the sharp corner (see Figs 58). The stresses around the adherend corner for the small radius

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

825

Figure 5. Peel stresses in adhesive MY750 with a 20 kN applied load, close to the unloaded adherend.

are nite and much smoother than those around the sharp corner. Furthermore, the positions with the highest stresses move inside the overlap for the small radius adherend corner. Very smooth stress variations exist for the large degrees of rounding and the area with the highest stresses is much larger than that with small or sharp radius corners. This clearly shows that the stress concentration is not so severe for the large degrees of rounding. It should be noted that the longitudinal stresses are signicant (Fig. 7). In fact, the longitudinal stresses for all the radii and for the sharp adherend corners are much higher than either the peel or the shear stresses plotted at the same positions. In lap joints with llets, the longitudinal stresses near the ends of the overlap are as important as the peel stresses, because the llets transfer part of the load and this results in large longitudinal stresses near the adherend ends. The rounding of the adherend corners enhances the longitudinal stresses as can be seen from Fig. 7, which shows that the longitudinal stress increases dramatically in the region of the rounded adherend corner but remains almost constant in the middle of the overlap. The stress state in the llet is predominantly in tension at an angle of approximately 45 to the applied load. Consequently, the longitudinal stresses play an important role in the failure of joints with llets and rounded corners. The maximum principal stresses at the Gauss points close to the unloaded adherend are plotted in Fig. 8. It may be seen that the distribution of the maximum principal stresses is similar to that of the other stresses, especially the longitudinal stresses, plotted at the same positions and these stresses increase sharply inside the rounded

826

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 6. Shear stresses in adhesive MY750 with a 20 kN applied load, close to the unloaded adherend.

adherend corners, which is a typical feature of the longitudinal stresses plotted at the same positions. This also means that the longitudinal stresses contribute significantly to the maximum principal stresses. The maximum principal stresses inside the llets are plotted in Fig. 10. It may be seen that stresses inside the llets near the llet edges hardly change for different degrees of rounding. It should be noted that the stresses near the edge of the llet (point E in Fig. 10) are high, where a weak singularity exists. The stresses at the edges of the llet increase sharply and the magnitudes are difcult to determine when the geometry consists of sharp edges. It may be seen that the stresses at the edges of the llet are higher than those near the unloaded adherend corners for medium and large radius adherends. Therefore, for large radius adherends, the llet edges may be important for failure predictions. In reality, there is usually a wetting meniscus which smoothes the 45 angle (rounded edge) at the edge of the llet to zero and effectively removes the stress concentration [10]. The peel stresses at the Gauss points close to the loaded adherend are plotted in Fig. 11. It may be seen that all the stresses peak in the region opposite to the unloaded adherend corner, although there is no re-entrant corner close to the loaded adherend. Figure 11 shows that there is little difference in the peel stresses between sharp and small radius adherend corners. This is easy to understand because no re-entrant corner exists close to the loaded adherend and the change in adherend

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

827

Figure 7. Longitudinal stresses in adhesive MY750 with a 20 kN applied load, close to the unloaded adherend.

geometry is a signicant distance away from this region. It may be seen by comparing Figs 11 and 5 that the peel stresses near the loaded adherend are much lower than those near the unloaded adherend. This means that the area around the unloaded adherend corner is more important than the adjacent area close to the loaded adherend for joint strength predictions for elastic adherends. The other stresses have similar features. To examine further the nature of the maximum principal stresses, their directions are drawn in Fig. 12 for the small radius adherend corner. It may be seen that the directions of the maximum principal stresses are approximately normal to the interface between the unloaded adherend and the adhesive. In comparison, the directions of the maximum principal stresses around the unloaded adherend corner for a joint with sharp adherend corners are oriented about 45 to the interfaces, as is well documented in the literature [8]. In general, adhesives are weakest in tension. At the interface close to the small radius adherend corner, the maximum principal stresses act approximately normal to the interface. This direction is the worst as far as the joint strength is concerned. In a joint with sharp adherend corners, however, the tensile forces acting at the interface are either peel or longitudinal stresses, whose magnitudes are much smaller than those of the maximum principal stresses. As a result, the direction of the maximum principal stresses for a joint with sharp corners is less critical than that for a joint with small radius corners, although the

828

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 8. Maximum principal stresses in adhesive MY750 with a 20 kN applied load, close to the unloaded adherend.

magnitudes of the former stresses are much larger than the latter. As a result, the joint strength does not improve much in contrast to the reduction in the peak values of stresses for the rounded adherend corners. 4.2. ElasticPlastic Analyses A B-spline curve tting procedure was used to represent the behaviour of the materials. For the CTBN rubber toughened epoxy, the modied von Mises yield criterion [24] was used to take into account the hydrostatic stress components, which can be expressed as:
2 J1 (S 1) + J1 (S 1)2 + 12J2 2 = YT , 2S

(1)

where S is ratio of the yield stress in compression to the yield stress in tension, YT is the yield stress derived from the uniaxial tensile test on the material, and J1 = x + y + z , 1 2 J1 3
2

J2 =

+ y

J1 3

+ z

J1 3

2 2 2 2 + xy + xz + yz ,

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

829

Figure 9. Stresses from the sharp corner across the adhesive thickness for adhesive MY750 with a 20 kN applied load.

where i (i = x, y, z) and j (j = xy, xz, yz) are stress components. For the high strength aluminium alloy, the von Mises yield function f ({ }) = 3J2 YT = 0 (2)

was used, where J2 is dened as above. A load of 20 kN was applied throughout the analysis. Again, the analysis involves two steps, a coarse mesh analysis followed by a rened mesh analysis. The results are presented below. First, the plastic energy density is plotted close to the unloaded adherend as shown in Fig. 13. A similar trend of the plastic energy density exists as that of the stresses in the linear elastic analysis. A high plastic energy density was apparent around the corner; this should, in theory, be innite for a sharp corner as discussed by Adams and Harris [6]. The plastic energy density with the small radius corner was nite and the position of the peak value moved inside the overlap, occurring after the edge of the rounding (see Fig. 13). As for the stresses in the linear elastic analysis, the area with a large plastic energy density for the small rounded corners was much larger than that in the sharp corner. For joints with medium and large radius corners, the magnitudes of the plastic energy density were much reduced and the positions of the peak values moved further inside the overlap, occurring after the edge of the rounding (see Fig. 13). Again, the larger the rounding, the larger the

830

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 10. Maximum principal stress comparison inside the llet for adhesive MY750 with a 20 kN applied load.

area in which the plastic energy density reached its peak values and the further the position moved inside the overlap. The plastic energy density inside the llet is shown in Fig. 14. It can be seen that the majority of the llet did not yield much with applied load. However, yielding happened at the edge of the llet (point E in Fig. 14) where there was a stress concentration in the stress analysis for linear elastic materials. This shows that the regions around the unloaded adherend corners and the llet edges are important areas to examine for failure analysis. However, as discussed above, there is usually a wetting meniscus which smoothes the 45 angle (rounded edge) at the edge of the llet to zero and effectively removes the stress concentration [10]. Strains were also studied to analyse the failure mechanisms. Because the strain distribution is very similar to that of the plastic energy density, no plot with regards to the strain magnitudes has been made here. However, the directions of the maximum principal strains with a small rounded corner are presented in Fig. 15, which gives more information on the failure modes. It may be seen that, as for the directions of the maximum principal stresses in the linear elastic analyses, the directions of the maximum principal strains were approximately normal to the interface between the adherend and the adhesive for small radius corners. It can also be seen that the magnitudes of the maximum principal strains close to the unloaded adherend were much larger than those at other places in the adhesive layer. The

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

831

Figure 11. Peel stresses in the adhesive MY750 with a 20 kN applied load, close to the loaded adherend.

magnitudes of the other principal strains in the same plane were also signicant, but they were in compression. This means that the shear strains at these points were very signicant compared with elongational strains. As discussed in the linear elastic analysis, stresses and strains are the highest near the interface so the joint is most likely to fail there. The action of the maximum principal strains is mainly responsible for the joint failure. The directions of the maximum principal strains with sharp adherend corners were almost the same as those with small radius corners and were approximately normal to the interface between the adhesive and adherend. It should be noted that the maximum principal strains were not actually normal to the interface between the adherend and adhesive for joints bonded with sharp adherend corners. The direction of the maximum principal stresses for a joint with sharp corners is less critical than that for a joint with small radius corners; however, the magnitude of the peak value of the maximum principal strain with sharp adherend corners is much higher than that with small radius corners. As a result, like the case of the elastic analysis, the joint strength may not improve much in contrast to the reduction in the peak values of stresses for the rounded adherend corners. Finally, the shear and peel strains across the adhesive layer are shown in Fig. 16. It may be seen from the gure that the shear strains are much larger than the peel strains close to the unloaded adherend. The strains remain reasonably constant for the majority of the adhesive layer. However, the shear strains peak sharply close to

832

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 12. Pattern of the principal stresses in the adhesive MY750 for a 20 kN applied load around the small (0.25 mm) rounded adherend corner.

the unloaded adherend and are much larger than the peel strains. The change in peel strains is not as sharp as that in the shear strains. 5. Conclusions The following conclusions can be drawn from the above analyses. 1. The stresses or strains in the adhesive layer of a joint with rounded adherend corners are nite and can be determined. In real joints, adherends generally have small rounded corners. Consequently, the model with a small radius corner may be used to represent real adherends. However, the magnitude of the rounding of the adherend needs to be determined. 2. Rounding the adherend corners reduces the magnitudes of the stresses or strains around these corners. The larger the rounding, the larger the reduction in the magnitudes of the stresses. Furthermore, the peak values of the stresses or strains move inside the overlap for joints with large rounded adherend corners: the larger the rounding, the further inside the overlap the peak values move. 3. Rounding of the adherend corners worsens the stress conditions around the adherend corners because the direction of the maximum principal stress is normal to the adherend, although the peak values of the stresses and strains are much

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

833

Figure 13. Plastic energy density in the CTBN adhesive with a 20 kN applied load, close to the unloaded adherend.

reduced. Increasing the radius of the adherend corners results in a large area of the interface between the adherend and the adhesive which experiences large stresses or strains acting approximately normal to the interface between the adherend and the adhesive. It also reduces the parallel portion (no rounding) between adherends which concentrates the load. For example, a 3.2 mm rounding with a 25 mm overlap leaves a 19 mm parallel portion. 4. Rounding the adherend corners is restricted in that the stresses or strains at the edges of the llets do not change much for different degrees of rounding. For a large degree of rounding, the edges of the llets may become the critical points for joint failure. 5. The stress or strain concentrations are very much localised to the unloaded adherend corners, while the stresses and strains away from the corners are quite smooth. This means that the strength does not change much for joints with sharp and small radius adherend corners. 6. Finally, the longitudinal stresses (or strains) along the overlap, which are generally ignored by closed-form solutions, are more signicant than the peel stresses (or strains) around the rounded adherend corners and they should be taken into account in failure studies. However, it is the maximum principal stresses which are the most responsible for joint failure because they take into account all the

834

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

Figure 14. Plastic energy density inside the llet for the CTBN adhesive with a 20 kN applied load.

Figure 15. Pattern of the principal strains in the CTBN adhesive with a 20 kN applied load around the small rounded adherend corner.

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

835

Figure 16. Shear strain and peel strain for the CTBN adhesive with a 20 kN applied load across the adhesive layer.

stress components. Shear strains for the elasticplastic analyses are much larger than either peel or longitudinal strains around the rounded adherend corners. References
1. R. D. Adams, J. Comyn and W. C. Wake, Structural Adhesive Joints in Engineering, 2nd edn. Chapman & Hall, London (1997). 2. O. Volkersen, Luftfahrtforschung 15, 41 (1938). 3. M. Goland and E. Reissner, J. Appl. Mech. 66, A17 (1944). 4. L. J. Hart-Smith, NASA Contract Report, NASA CR-112236 (1973). 5. L. F. M. da Silva and R. D. Adams, Int. J. Adhesion Adhesives 27, 362 (2007). 6. R. D. Adams and J. A. Harris, Int. J. Adhesion Adhesives 7, 69 (1987). 7. A. chsner, L. F. M. da Silva and R. D. Adams, in: Modeling of Adhesively Bonded Joints, L. F. M. da Silva and A. chsner (Eds), pp. 131154. Springer, Heidelberg (2008). 8. R. D. Adams and N. A. Peppiatt, J. Strain Anal. 9, 185 (1974). 9. A. D. Crocombe and R. D. Adams, J. Adhesion 13, 141 (1981). 10. R. D. Adams, R. W. Atkins, J. A. Harris and A. J. Kinloch, J. Adhesion 20, 29 (1986). 11. L. Dorn and W. Liu, Int. J. Adhesion Adhesives 13, 21 (1993). 12. M. Y. Tsai and J. Morton, Composite Struct. 32, 123 (1995). 13. T. P. Lang and P. K. Mallick, Int. J. Adhesion Adhesives 18, 167 (1998). 14. G. Belingardi, L. Goglio and A. Tarditi, Int. J. Adhesion Adhesives 22, 273 (2002). 15. M. K. Apalak and A. Engin, J. Adhesion Sci. Technol. 18, 529 (2004).

836
16. 17. 18. 19. 20. 21. 22. 23. 24. 25.

X. Zhao et al. / J. Adhesion Sci. Technol. 25 (2011) 819836

L. Andreassi, R. Baudille and M. E. Biancolini, Int. J. Adhesion Adhesives 27, 458 (2007). L. F. M. da Silva and R. D. Adams, Int. J. Adhesion Adhesives 27, 227 (2007). J. Deng and M. M. K. Lee, Composites Part B 39, 731 (2008). L. F. M. da Silva, T. N. S. S. Rodrigues, M. A. V. Figueiredo, M. F. S. F. de Moura and J. A. G. Chousal, J. Adhesion 82, 1091 (2006). H. L. Groth, Int. J. Adhesion Adhesives 5, 19 (1985). R. D. Adams and P. Cawley, in: Research Techniques in Non-destructive Testing, R. S. Sharp (Ed.), pp. 303360. Academic Press, London (1985). L. F. M. da Silva and R. D. Adams, J. Adhesion Sci. Technol. 19, 109 (2005). J. M. Gere, Mechanics of Materials, 5th edn. Brookes/Cole, Pacic Grove, CA (2001). R. S. Raghava, R. Cadell and G. S. Y. Yeh, J. Mater. Sci. 8, 225 (1973). V. L. Hein and F. Erdogan, Int. J. Fracture Mech. 7, 317 (1971).