Thermal stress analysis

[rishipillaifsi]

Hi,

I am new to the forum and I have a question already.I am doing a thermal stress analysis in Ansys. I transfer the thermal loads from Ansys CFX to Ansys Multiphysics and use the Multifield tool.

To reduce structural stresses I have constrained my model which by the way is huge(800000 elements) only on the bottom surface in the Y(vertical) direction. But due to high temperatures as you can imagine the material is yielding as is evident form the static linear analysis performed. This generates a rigid body motion due to less constraints.

I see the problem in not allowing a large deformation analysis or a plastic material property or not accounting for elasticity modulus change with temperature. Any of the three methods provide almost the same results though the large deformation method may predict less stresses.The problem is that I see my part moving in x direction only i.e. the nodal displacements in x-direction is unidirectional which cant be in a thermal expansion.What correction would you recommend?

Could you please suggest something and a discussion would be an icing on the cake?

Thanks a lot.Keep up the good work.

Rishi Pillai

 

[corus]

I'm not sure why you expect rigid body motion just because the material has yielded. It doesn't make sense. In your case you have only one restraint (in the y) so I don't see how you managed to get any results at all. If you're looking to see what stresses you get from differential thermal expansion within the body due to some temperature distribution then you need to pin the body in the X and Z directions at single positions to prevent not only translational movement but also rotational. This can be tricky where temperatures are concerned as you don't want to induce any restraint to the body that will cause false stresses. If you restrain the body properly you should see no noticeable stresses at the points you restrained. If possible, use symmetry constraints.

corus

 

[rishipillaifsi]

Thanks corus. what I meant by rigid body motion is that i know I have an insufficently unconstrained model.I forgot to mention that I am using Solid45 which eliminates ROT dof. Secondly in order to simulate free expansion and as I mentioned to eliminate strutural stresses generated by fixtures I constrained the model as explained(the way it is in reality;just placed on a flat surface);its like a roller joint.You already mentioned false stresses which is exactly what i didnt want.I cannot use symmetry constraints as I dont have loading symmetry even though i have geometric symmetry.

Suggestions for constraints required. I am waiting for your reply. I am trying a few things simultaneously with my model.

Rishi Pillai

 

[cbrn]

Hi,
first of all, using an element which doesn't have rotational DOFs doesn't imply that you won't have rotations in the model: with a solid element, the element itself can pivot around one node (one node has all DOFs restrained while the others have all translational DOFs free...).
This said, you'd better use 3-2-1 constraining method (already described in a former thread), bearing in mind that you have necessarily to eliminate 6 independent DOFs from your model. You won't have stresses auto-induced by boundary conditions if you keep them rigourously isostatic.
However, if you allow for FREE thermal expansion, there will be no thermal stress in the structure, only deformation.

Regards

 

[cbrn]

... I forgot: the thread I wanted to point you to is:
"What is 'Soft Spring'"

 

[rishipillaifsi]

Thanks cbrn.The pivoting is valid for only the the element where u apply nodal constraints.Otherwise rotation is not possible.I said "simulate" free expansion i.e. keep the constraints minimum. Lastly but never the least FE methods treat thermal deformations as initial strains and the stress is a function of this strain and the total strain.This is what is done in metallurgy in releiving stresses and so on and mind you with no external forces or unwanted fixtures.

Thermoelastic analysis is different from thermal-structural analysis in a special way.

Thanks a lot again.I will try the 3-2-1 method.I received excellent results complying with experiments with my previous run with lower temperatures.Only when the temperatures went too high was the solution fallible.Will run this one with 3-2-1 constraints.

Rishi

 

[corus]

riship,
What cbrn refers to is the rotation of the whole structure. Obviously a 3D solid element doesn't have rotational degrees of freedom, but the whole body can rotate translationally. For instance restraining one node in x, y, and z will restrain the whole structure translationally, but the whole body can rotate about that point. You just have to visualise it.
I'm not sure what you mean by a thermo-elastic model as opposed to a thermal-structural model. Stresses are caused by differential thermal expansion. If the body is held between two zero restraints then the stresses are caused by differential thermal expansion from points at a zero temperature, effectively. If the body has no restraints then the stresses will be caused by differential thermal expansion from within the body. It's the same whether it's thermal-structural or whatever you want to call it. Non-linear and/or temperature dependent properties complicate matters a little, but otherwise the principle is the same.
In reality your body is restrained by the ground physically, and by friction transversely. If you restrain the whole of the bottom of the structure (say in the y) then you may prevent any bending upwards to cause the bottom to seperate off the ground. If you don't have contact modelled then you may have to 'guess' where the body will contact and restrain it in the Y at those points. You'd need at least 3 points (not in the same plane)restrained in the Y to ensure contact of course. These 3 restrained points will automatically give the body rotational restraint about the x and z . Restrain one of these 3 in the x and z to give you full translational restraints to the whole body. A soft spring attached to the body in either the x ot y will give you the rotational restraint about the Y axis to give younumerical stability for a solution. The choice of spring stiffness will determine the effect on the overall solution. Too high and you'll see a hot spot of stress. It's best to attach the spring in a region where you're not too concerned about the stresses there. Of course you could assume that the body won't slide and that friction will prevent any movement of the base at all. Then you could apply a restraint to the base to prevent expansion and stop rotation of the whole. That would be your judgement but generally friction is ignored.

corus

 

[torpen]

Hello,

A 3D structure has 6 rigid body modes whatever the finite element type. You can perform a free-free modal analysis and check the first eigenmodes to verify what cbrn said.

Regards,

Torpen