Cylindrical Model Meshing Help

[william.smith-keegin]

Hello,

I'm rather new to the world of AutoDesk, and particularly to AutoDesk Simulation. I'm carrying out some heat transfer FEA for a research project, and I'm having a bit of trouble making a decent mesh out of the model.

The model consists of a simple cylinder (50mm wide, 100mm long), with a 1.1mm hole drilled from one end to the midpoint. Into this hole is inserted a 1mm thermocouple (currently modelled as a pair or concentric cylinders). The air in the hole surrounding the thermocouple is modelled as a separate body. The geometry is imported as a 3D model created in AutoDesk Inventor. 

I'm meshing using pretty standard parameters (I can't quite remember the specifics) – a pretty low element size, as well as a low aspect ratio and warp angle limit. I think I had around 20 refinement points, all situated along the thermocouple. The mesh generated looks fairly good on the surface but really messy on the interior.

In some analyses, the air gap was at a higher temperature than both the thermocouple and cylinder – which seems impossible to me. This is presumably not a meshing issue though.

Here are some pictures. I guess what I am looking for is some advice as to how to make the mesh a bit better. It seems like a mesh composed of subdivided concentric circles would be a lot better. If anyone has any advice as to how to improve anything, that would be much appreciated. Various models could be supplied if required.

Thanks,

Will

A 3D view of the entire model.

An interior view (via slice plane) of the model.

[Joey.X]

Could you provide more specific info for this problem?

- What kind of BCs (boundary condition) and IC(Initial conditions)? 

- Does the composed by multiple parts which has least a part for the air (you mentioned separate body)? if yes, are the parts bonded?

- Do you use steady ot transient analysis? 

[william.smith-keegin]

Of course, sorry. I should have supplied more details originally. 

The cylinder, air gap, and inner- and outer-layers of the thermocouple are all made up of separate parts with different thermal properties. I don't know what you mean by bonded, so I shall assume that they are not. 

The analysis is transient. Everything starts at 20ºC, and the bottom is heated by an applied temperature of 1900ºC for 15-20 minutes. There are no other conditions applied to the other surfaces (which I assume means they're adiabatic).

Will

[Joey.X]

 Hi, Will, 

Thanks for providing more info, it clarifies that the model is transient conduction only analysis.

For model with multiple parts, ASIM has the options for the part contact on the interface (from Tree View/Bottom), the default is bonded. For your case, you have a part for air already, so it will be appropriate to have bonded (no temperature jump between the interfaces of the parts). Please double check it. 

It is acceptable to model the air as a part, however, it brings the simulation accuracy issue due to the sudden (and large) change of the thermal conductivity, in this case, is there a large thermal conductivity change between cylinder to air or thermocouple  to air?  To avoid or reduce the temperature result inaccuracy as mentioned in your first post, the boundary layer mesh is highly recommended especially in the part of air which has very low thermal conductivity and high temperature gradient is expected. For a easy shot without changing mesh, changing (artificial) thermal conductivity in air part to make it close the property of cylinder to see the result changes.

PS, for transient analysis, small time step also helps for better accurancy. 

The other modeling thought is deactivating the air part, which is assuming the complete adiabatic in air part, this assumption usually acceptable if the other parts has larger difference in thermal conductivity. 

Regards,