Learning Forum

[SR786]

I am attempting to create a geometry with a magnetic rod stirrer (cylindrical) lying on the base of the reactor. I have used a boolean operation to subtract the rod from the reactor (at typical approach when modelling a stirred tank reactor) but I notice the cell skewness is very poor (0.99) meaning that I cannot run this mesh in ANSYS FLUENT as the simulation will blow up (even if I use PISO with skewness correction it doesn't work).  

Does anyone have any suggestions on how to put the rod stirrer on the base of the reactor without the simulation blowing up (i.e., how to lower skewness of mesh)

[Federico Alzamora Previtali]

Hello, 

what did you use to create your Mesh? Where are those highly skewed cells? Is there a contact between the cylinder and the tank?

[SR786]

Hi, I used ANSYS workbench to create my geometry and mesh. The skewed cells are located at the bottom of the tank, where the cylindrical stirrer meets the tank bottom

[Rob]

Move the mixer off the bottom of the tank by a small amount. In reality there will be a gap too, it's just rather small. 

[SR786]

 

Hi, I’ve tried doing this before (adjusting mixer by a small amount off the tank bottom) but the skewness is still poor (cells exceed skewness value of 1). The skewness only goes down if the rod is moved by a more noticeable gap (see image below) so that I can run simulation but as you can see at bottom (underneath stirrer) there is some flow when there shouldn’t be which is why I'm trying to get the rod touching the bottom

 

 

[Rob]

You also need to have enough cells between the mixer and tank to let the mesher work. If you add inflation into the mix you're going to struggle without a very small minimum cell size. 

[SR786]

Okay I see, so it's not possible to put the stirrer exactly at the base of the reactor due to overlapping cells?

[Rob]

Overlapping cells? 

It's possible if you remove the sharp angle at the contact (look at how the automotive models do tyres) but the question is, how close do you need/want to be, and how will that effect your model? Or rather, can you afford the cell count relative to how much it's going to alter your solution.