Showing results for 
Search instead for 
Do you mean 

Diverging Model in Thermal Analysis



I'm doing a thermal analysis with NX Thermal (not Nastran). As I'm searching the convergence temperature of a steady-state-problem I'm playing with the number of finite elements in a component of the system.


The problem is that if I go with the element length under 0.05 mm (for example 0.04mm) the simulation seems to have problems and does not converge. Over the iterations the model solid maximum temperature is jumping around in the range of  nearly 0 to 6000 °C; also the minimum temperature is even below 0 (impossible... the model is cooled with fluid at 22°C). I know that the max temperature should be around 160 - 170 °C.


What can I do to get a converging model again?


Re: Diverging Model in Thermal Analysis


I guess you need to use 2D elements in stead of 3D elements of such small elements.I had the same issue when I wanted to mesh a fluid with small gaps of 30 µm. I solved that by removing the small gaps in the fluid. After doing that I could mesh the fluid again with 3D elements.Another way could be to mesh the part with a 2D mesh and connect the surrounding parts with a thermal coupling.




Re: Diverging Model in Thermal Analysis

Dear Fisher,

The element size is important regarding accuracy results, but if for reducing element size it caused that the mesh gets higly distorted then solution could diverge. Also, you do not mention what type of Finite Elements you are using: 3-D TET/HEX, or 2-D elements, this is important to understand the problem.

Mesh quality is critical, then take a look to mesh distorsion, inspect your model, check for Jacobian, Aspect Ratio & TET collapse ratio, and see if mesh mass the minimum threshold quality requirements. Activate the flag "Auto Fix Falied Elements" during Meshing process, the software NX AdvSim uses different strategies to repair element quality issues depending on the quality criteria that the element failed:

  • For elements that exceed the specified Max Jacobian value, the software straightens the elements' midnodes to eliminate the issue. The software does not remesh the associated surface or body.
  • For elements that exceed the specified Aspect Ratio tolerance, the software remeshes any surfaces associated with those elements. The software never removes elements that exceed the Aspect Ratio.
  • For elements that exceed the specified Tet Collapse tolerance, whether the software removes the failing elements or remeshes the associated surfaces depends on the element's topology and the degree to which the element exceeds the tolerance.

If the software removes an element, it maintains the connectivity of the nodes on the surrounding elements such that the mesh remains contiguous. The software removes elements that fail Tet Collapse when:

  • The element is completely flat and all its nodes lie on the outer surfaces of the volume. For all other element configurations (for example, when an element is not completely flat or at least one node located in the volume's interior), the software remeshes all surfaces associated with that element.
  • The element's Tet Collapse ratio exceeds 200.
  • The element's Tet Collapse ratio is between 100 and 200 and all four of its corner nodes lie on the outer surfaces of the volume.

The mesh quality is the trick!. Avoit to use at all TET elements and mesh with HEXAEDRAL elements, you will see a great improvement in both mesh quality & accuracy results!!.
Best regards,

Blas Molero Hidalgo, Ingeniero Industrial, Director
Blog Femap-NX Nastran:

Re: Diverging Model in Thermal Analysis

I'm moving this to a new post since it's a slightly different topic