Stiffness mismatch

[khoppe83]

Hello, I am trying to analyze a filler material, i.e. with a CTE of around 1E-4 1/F and modulus of around 500 psi. It is bonded in between two materials both with moduli greater than 1E5. The delta T of all three materials is around 150 F.

My question is, is there a general technique that allows this to be modeled accurately? When I equivalence the nodes, the large stiffness mismatch/CTE mismatch creates pretty unrealistic stresses. I then ran it with a glued contact solution (Patran sol 600) between the different materials which creates more realistic stresses than the equivalencing the nodes. Any thoughts on how to improve the fidelity of these analyses?

 

[gwolf2]

Are you using first or second order elements?

 

[khoppe83]

Hello, I am using first order elements (Hex 8).

 

[gwolf2]

Sorry, been busy.

Patran Sol 600 actually means NASTRAN nonlinear of some sort. Nastran is the solver which often hides behind Patran. Patran will have created some sort of 'unusualness' w.r.t. contact which you will have to investigate by looking at the NASTRAN input file and bulk data deck manual.

You are using the best elements for this type of thing in NASTRAN. It's difficult to know how you can improve it. NASTRAN just isn't good at contact - it's a Patran bolt-on probably using gaps or MPC's, not core NASTRAN functionality.

You could tidy it up by using identical meshes on either side of the gap, 1D none-node gap elements, a fine mesh near the contact edges.

Otherwise, use Hertzian hand calcs or get a better contact code like ABAQUS or MARC.

Regards,

Gwolf.

 

[rb1957]

your "glued contact" solution appears to be closer to reality. when you equivalence the nodes the same node is used for both elements ... which doesn't sound to be real. it would seem better to have a finite stiffness between the two elements, since you'd expect the two materials would displace differently.

 

[khoppe83]

Thanks for the help. A bit of clarification on my part will probably help the follow on question a bit more, sorry for the confusion.... Nastran sol 600 actually simply exports it as a .bdf file, however it is not running Nastran, it is actually running the Marc solver (counterintuitive, but it's currently like Mentat without all of the nifty gadgets).

As a follow on question, do you think nonlinear material properties of a filler material such as RTV would make a significant difference if the gap is only 15 to 30 mils?

Thanks again!

 

[gwolf2]

Regarding the PATRAN-NASTRAN-MARC issue: Well, you have MARC in there so that's good but what happens in the deck translation process inbetween is anybody's guess!!!

What exactly do you mean by mils? Are you modelling a very thin glue line?

gwolf

 

[SAITAETGrad]

Any chance this could be modeled as a laminate plate element?