Comparison between mesh types

One of the most important processes when performing a numerical simulation is meshing, currently there are several commercial and free software, in another post we will talk about it, however the most used (I think) is the Ansys Meshing, which allows to develop and customize the mesh to avoid errors of non-convergence in the calculation. Also in ANSYS MESHING, there are several types (shapes) of mesh elements that we can use and combine to better adapt the geometry and reduce the calculation time.

Therefore we decided to do a test with some forms of mesh elements and thus observe the variation of results and the calculation time that takes in each case. We use mesh elements in three dimensions with ANSYS CFX and ANSYS FLUENT (later we will do more tests with identical meshes using other CFD software).
We observe that the polyhedral mesh is the one that contains a smaller number of elements but a larger number of nodes, unlike the tetrahedral mesh, this as we will see in the following tables of results directly affects the calculation time.


For the created model, the elements in the polyhedral mesh are reduced to less than 50% compared to the tetrahedral mesh. The combination of a hexahedral and prism mesh is in an intermediate-term keeping their values of nodes and elements close.

Case analyzed:

Simulation of a phase (water) in a pipeline, with activation of energy using the k-epsilon turbulence model considering an inflation in the walls.

Types of mesh: Tetrahedral, Hex / Prism and Polyhedral

To observe the differences between the meshes you can see the following video:

Results:

We observe that the polyhedral mesh is the one that contains a smaller number of elements but a larger number of nodes, unlike the tetrahedral mesh, this as we will see in the following tables of results directly affects the calculation time.

For the created model, the elements in the polyhedral mesh are reduced to less than 50% compared to the tetrahedral mesh. The combination of the hexahedral and prismatic mesh is in an intermediate-term keeping their values of nodes and elements close.

The results of the physics of the problem (temperature and velocity) at the outlet of the pipe are very similar for all cases, however, the distribution of temperatures and speeds is slightly different in the polyhedral meshing, you can see the video again (shown above ) and see this point in more detail.

Regarding the difference between the software used (CFX and FLUENT) we note that CFX obtains results in smaller iterations (we did not modify the residues of the convergence, these default values were maintained for both software), however, the calculation time in CFX is much smaller with the tetrahedral mesh and the combination of hexahedral and prismatic; up to this point CFX for a great alternative to reduce computational time but the polyhedral mesh (which can only be generated in ANSYS FLUENT) decreases the calculation time radically.

Therefore from this small test, we can give some scopes as seen.

  1. The elements of the mesh are those that define the dependence of the calculation time.
  2. The polyhedral meshing gives a shorter calculation time compared to other mesh types.
  3. The physical results of the polyhedral mesh are similar to the other mesh types analyzed in this test.

Recommendations

  • To know the error of the results, it is necessary to validate the simulation with an experimental test that allows us to know what type of mesh is the most appropriate.
  • We can not conclude with this test that one software is faster than another, this has only been a case in a single phase, it is likely that by adding more variables or changing the turbulence model, the results obtained will vary.

Hoping that these results can be helpful, we invite you to subscribe to our YouTube channel: CFD.NINJA where you will find varied information and CFD tutorials.

If you want to know how to generate a polyhedral mesh you can see this tutorial:

 

1 Comment

Leave a Reply

Your email address will not be published.


*