Adhesively bonded joint analysis

[Mat_Aero]

Hello.
I have question regarding modeling and analysis of the bonded joint for thermoelastic load case. The bonded joint consists of CFRP tube from T700 prepreg and Titanium tube which are bonded together by epoxy adhesive, see attached figure.
The CFRP tube is modeled with shell elements while titanium tube and adhesive is modeled with solid elements. The titanium tube is fixed at the end by MPC while CFRP is unconstrained. All components were modeled node to node and merged together. Assuming ambient temperature of 0K and nodal thermal load of -70K, the thermal change corresponds to ΔT=70K.
However, due to high CTE mismatch between adhesive(CTE=80∙10-6∙K-1) and Titanium tube(CTE=8.6∙10-6∙K-1) or CFRP tube, large stresses in adhesive occurs which exceeds the strength of the adhesive. By setting CTE of adherends to 0, it is visible that large portion of these stresses are due to expansion/contraction of adhesive itself, see attached figure. My question is, if the modelling approach is correct and if it is correct to assign CTE to adhesive as well, considering its small thickness of 0.2mm and relatively small thermal change of ΔT=70K. I’ve tried modelling the CFRP tube with solid elements as well, however it had negligible effect on stress results. I would be very glad if any of you would share some insight on FEA of bonded joints. I attach the analysis model(bdf file),SI units were used, and figures of stress distribution in adhesive for these 3 cases:
1. CTE set for all components
2. CTE of adhesive set to 0
3. CTE of adherends set to 0
Thank you!
Kind regards,
Mat

 

[SWComposites]

Are you using linear elastic or nonlinear stress-srain properties?
Where are you getting your adhesive material properties from?
Do you have any test data to correlate your FEA to?
How many elements thru adhesive thickness?
Attachment wont open on ipad.

 

[Mat_Aero]

I am using the linear elastic properties. I managed to find stress-strain curve for the adhesive I assuming to use, and the curve has linear behavior for large portion of stress values and then remains practically flat. I’ve taken the adhesive properties from manufacturer’s datasheet. I haven’t performed any tests by myself, however I’ve found experimental results for double lap joint in design handbook so I’ve also tried to model the double lap joint Titanium-QI CFRP as well according to ASTM test method and try to correlate my FEM to those experimental results however with no success. My results were higher approximately by order of 10.
In the model I’ve attached, there is only 1 element through thickness, which may look insufficient, but I’ve also performed refinement with refined mesh, several elements through thickness and even with parabolic elements which improved accuracy, but not marginally. My main concern is, whether such thermoelastic analysis methodology by modeling adhesive with solid elements is correct, or if there is any other approach how to evaluate bonded joints due to temperature load. I’ve seen similar designs of CFRP tubes bonded to titanium or aluminium tube which were used for even higher ΔT with no problems.

 

[SWComposites]

Using linear properties will give unrealisticly high peak stresses. You need to use nonlinear properties. And multiiple elements thru the adhesive thickness. The vendor data is likely worth just about what you paid for it.