Modelling Adhesive Component In Bonded Joint PATRAN 1

[darkden321]

Hi

I was wondering if there was any suggestions to how the adhesive componenet of the bonded joint could be modelled in PATRAN?

 

[blakmax]

Good luck. I am not a FE person but I believe that the aspect ratio of the elements leads to errors. The fellows I used to work with developed special elements to handle adhesive bonds.

However, if you do get an analysis to work, please do NOT rely on an average shear stress approach, and do not rely on an elastic analysis. A very significant amount of the strain energy to failure for adhesive bonds is derived from plastic behaviour.

Regards

Blakmax

 

[RPstress]

For simple analysis the glue can be modelled with spring elements which can have the stiffness GA/t of the area of glue in shear they represent. This is adequate for elastic behavior. The shear stress in the glue can be recovered by a simple bit of post processing (divide the force in the spring element by the area of glue it represents). This can be compared with some "thin lap shear" test data for a very rough indicatiion of failure. Thin lap shear test specimens are usually 1/2" long so have a lot of peaking of shear plus they tend to bend adding in some peel. This makes them passably conservative for real structures. You will of course need to test the final configuration with at least some sort of subelement test.

If you want to get a bit more sophisticated then increase the mesh density a lot and model the glue with solids. With linear 8-node bricks You will probably need at least three elements through the glue thickness (glue is often about 0.01" or 0.25 mm thick).

It is more usual to do a simplified 2D model of a section through the glueline (perhaps towards the end of a glueline to estimate the elastic stress concentration). In that case the glue is just more 2D elements modeling its thickness. You still need a lot of elements and again about three elements through the thickness. A nonlinear analysis is advisable.

In the limit you can do what blakmax's friends did and essentially write your own code. I think that would only be a big advantage if you were worried about faliure of the glueline and wanted to model its behavior in detail non-linearly including plasticity. These days it might be more common to use a very nonlinear code (perhaps even an explicit one like Dyna or Abaqus Explicit) with appropriate elastic/plastic properties. You would need to get a measure of the glue's onset of yield (which is at quite a low strain) and the 'knee' where the glue gets more flexible in shear. See the Cytec datasheet on their FM 300 and FM 300-2 film adhesive for some representaive stress/strain curves. (E.g. page 8 of this datasheet

http://www.cemselectorguide.com/pdf/FM_300-2_092711.pdf.)

Like all analysis you must plan about what use you want to make of the results (or what you want the results to tell you) and allocate resources appropriately. For instance if you want to model the effects of defects in the glueline you will probably need to consider the glue's toughness.