Spew fillet

Real bonded joints are unavoidably formed with a fillet of adhesive spew at the overlap ends (Fig. 4.10(a)). Even if this is removed, some slight amount of adhesive still remains and truly sharp corners are not encountered. It is in these regions of maximum stress, where failure is initiated, that most analyses are remote from reality, and here Adams and co-workers at Bristol University have been amongst the main investigators, using finite-element techniques(5). Indeed the adherend corner is likely to be rounded rather than truly square, which actually has the effect of reducing stress concentrations in the adhesive layer. Joint failure begins in these spew fillets of adhesive... 

Adhesive Joints tend to form a fillet of adhesive spew, owing to the squeezing of the adherents during bending. The existence of a spew fillet can reduce the stresses at the free end appreciably, because of a stress relief mechanism, so unless otherwise agreed it is desirable to maintain a form of spew fillet. 

The joint geometry of our samples surely implied the presence of spew fillets, being the shaft press-fitted into the hub. The external spew fillet was visible after the coupling and it was removed before the adhesive cure. The internal one was observed after the breaking of the joint, but the geometry of the sample made impossible to remove it. Indeed, according to the pin and collar test method, a cord shaped adhesive spew must remain in one of the collar ends (Abenojar et al., 2013). 

Tsai M Y and Morton J (1995), The effect of a spew fillet on adhesive stress distributions in laminated composite single-lap joints . Compos Struct, 32(1 ), 123-131. 

Adams et al. [6] detailed axisymmetric analysis of a butt-tension joint and considered the influence of the detailed geometry at the edges of the joint on the stress distribution. The single lap joint was further analysed by Crocombe and Adams [7] using the finite element method and the effect of the spew fillet was included which was seen to significantly redistribute, and decrease, the stresses at the ends of the adhesive layer. In complementary work, Crocombe and Adams [8,9] analysed the mechanics of behaviour of the peel test and included the effects of non-linear deformations and also plasticity in their work. Harris and Adams [10] extended this work and accounted for the non-linear behaviour of the single lap joint. Crocombe et al. [11] quantified the influence of this non-linearity, both material and geometric. 

The geometry used in these analyses consisted of the single lap Joint configuration with a relatively short overlap of 12.5 mm. In a parallel experimental programme, any spew fillet that had been formed on manufacture of the Joints was machined off subsequent to the adhesive curing, so this was also removed in the model. 

In similar joints analysed and tested by other authors [17] the lack of a spew fillet for single lap joints with composite adherends had been seen to lead to... 

Finally, the excess adhesive squeezed from the joint, commonly termed the spew fillet , may contribute to its overall strength by reducing the stress concentration at the edge, so the fillets should only be removed for overriding practical or aesthetic reasons. 

Figure 6.15 Diagram of lap joints showing different types of adhesive edge, (a) Square (90°) edge, (b) 45° spew fillet [61].

It was shown above that the existing closed-form solutions predict that the highest stresses should be near the ends of the joint. However, they do not take into account the influence on these stresses of the spew fillet which is formed at the ends, and so it is in just these regions of maximum stress and where failure is bound to occur that the assumed boundary conditions of the previous theories are the least representative of reality. 

In their early work, Adams and Peppiatt (1974) used constant strain, two-dimensional, triangular elements which give the stress at the centroid of the element. Since joints tend to be wide compared with the thickness, the problem was considered to be one of plane strain. This assumption should be satisfactory for the adhesive layer, but less so for the adherend. The spew fillet was approximated to a 45° triangular fillet of varying size. The restraints used in their models are shown in Fig. 18 and a typical mesh in Fig. 19. 

Figure 24 shows a typical crack in the spew fillet of a double-lap... 

Fig. 24. Typical crack on loading double-lap joint made with spew fillets (joint has been cut and polished).

Comparison was made with the analysis by Allman (1977) by averaging the finite-element stresses across the adhesive thickness. Results for adhesive peel and shear stresses are shown in Fig. 30. Several points should be made here. First, when a square-ended adhesive layer was modelled, the results agreed closely with Allman s. However, when a full-depth spew fillet was used, both stresses were... 

However, when a finite-element analysis is made of a square-ended adhesive layer (as analysed by Allman), the comparison worsens markedly. Figure 31 shows the magnitude of the principal stresses predicted for a square-ended adhesive layer and one with a full-depth spew fillet. It can be seen that a high, tensile principal stress is predicted at the loaded adherend surface in Fig. 31(a) and small compressive stress at the corner. Averaging these stresses gives us 59 units. Any failure criterion based on average stresses with square-ended layers would therefore be in error by about 100% It is also in the... 

Fig. 31. Finite-element principal adhesive stresses around the end of the overlap for (a) square-edged joint (b) joint with full depth spew fillet, carrying the same load (arbitrary units).

In the earlier part of this chapter, it was shown that increasingly complicated mathematics is required if the stress situation in a single adhesive lap joint is to be determined. Even so, the law of diminishing returns applies, together with the irony that end effects, particularly where a spew fillet is involved, are the most difficult to model accurately while this is the most critical region since failure almost always occurs here. 

All the above analyses, with the exception of those by Kuenzi and Stevens (1963) and by Alwar and Nagaraja (1976a) make the assumption that the adherends are infinitely stiff compared with the adhesive. Although the adherends will normally have elastic moduli at least an order of magnitude greater than those of the adhesive, there will always be some Poisson s ratio strain in the adherends when the joint is loaded. Also, the interface does not remain plane under load since the stresses in the adhesive are not uniform. None of the above analyses makes any reference to the radial and circumferential stresses in the adhesive, except for that by Kuenzi and Stevens (1963), which makes the unrealistic assumption that the axial stress distribution in the adhesive is uniform. As in the case of torsion specimens, the presence of a spew fillet may affect the stress distribution, and none of the above analyses takes this into account. 

---