ABSTRACT
The strength of an adhesive joint is not a property of the adhesive alone but is a system property depending on adherends, the adhesive, the joint geometry, and service (or test) conditions. Most standard test methods specify giving the results as an average stress at failure. Use of this average stress is not very useful in the design and prediction of the strength of other adhesive joints that differ in even slight detail from the geometry of the test specimen. On the other hand, fracture mechanics methods have shown real promise in quantitatively explaining how differences in geometry can affect joint strength. The effects of modifications of lap specimens, similar to those described in ASTM D3165 ‘Strength Properties of Adhesives in Shear by Tension Loading of Laminated Assemblies’ have been investigated. An earlier paper discussed changes in strength resulting from tapering of the adherends over the overlap region. Here this type of analysis is extended to lap joints in which large debond regions were introduced in the adhesive joint. In both cases, the finite element/fracture mechanics predictions were in good agreement with experimental observations on steel-epoxy lap specimens. Next, a study of failure in cleavage specimens is discussed. ASTM D3433 ‘Fracture Strength in Cleavage of Adhesives in Bonded Joints’ is an ASTM standard whose basis lies in fracture mechanics. This standard makes use of specimens composed of ‘double cantilever beams’ (DCB) bonded together with the adhesive to be evaluated. The equation provided in ASTM D3433, to calculate the adhesive fracture toughness G Ic, is based on the equations from mechanics of materials for flexure stress and vertical beam shear stress and assuming ideal cantilever end conditions. This equation, therefore, neglects any energy stored in the adhesive or through rotation of the cantilever end and/or stored in the beam beyond the crack tip (assumed cantilever point). This equation was carefully analyzed using both classical (from the literature) and numerical methods. One would, of course, anticipate that the original assumptions would be most valid for long slender beams and very thin adhesives. Careful finite element/numerical analysis indicates, however, that when using dimensions well within those suggested in ASTM D3433, the energy release rate or G Ic determined including these factors can differ by more than 50% from that obtained using the recommended equation. A quasi-elastic adhesive was used with aluminum sheets to manufacture DCB specimens. A group of these specimens were tested to failure and the results used to determine a reference G Ic. This G Ic was then used with finite element methods to predict performance for test specimens with different geometries, i.e. the adhesive thickness and/or adherend thicknesses were altered. These test results confirmed the validity of the fracture mechanics/finite element approach. The calculated failure loads for the geometries typically differed by no more than 8% from the experimentally determined loads. The results for one sample lot differed by slightly more than 12% from the reference value but, as explained in the text, this difference might be attributed to mode dependence. On the other hand, use of the standard equation with the test results to calculate G Ic yielded values that differed by as much as 61650% between the different sample geometries. The finite element methods were also used to calculate the energy release rate for various assumed crack paths through the adhesive. It was hypothesized that the fracture locus should follow paths of maximum energy release rate. Experimental observations of crack paths for a number of different geometries of DCBs were consistent with these predictions in every case.
