CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Modeling a Fan by the Multiple reference frame (MRF) method in CFX. (http://www.cfd-online.com/Forums/cfx/84921-modeling-fan-multiple-reference-frame-mrf-method-cfx.html)

saisanthoshm88 February 12, 2011 03:45

Modeling a Fan by the Multiple reference frame (MRF) method in CFX.
 
1 Attachment(s)
I'm a newbie to CFD. I'm simulating the hot air flow inside a heating oven. My system consists of a rotating fan a bit adjacent to a heater coil. So the fan is meant to cause forced convection in the oven. The information available to me is the fan r.p.m and the temperature of the heater coil and the heat transfer from that in Watts.

My domain can be thought of a recirculating system. I mean it doesn't have a inlet and outlet.

I'd like to Model the fan ( the image is attached to this post.) by the Multiple reference frame (MRF) method.

To do this I've chosen a small fluid zone that surrounds the Fan surfaces to be a rotary domain.

But now the problem is that I can't set up the fan to be stationary with respect to the rotating domain in CFX.

I mean in Fluent there is an option to explicitly specify the velocity of an object relative to it's adjacent cell zone ( which is to be set to zero in my case) and Star-CD supports the approach of specifying an equivalent negative velocity ( so that the relative velocity is zero).

But it seems that CFX doesn't support either of these approaches. I tried to define the Fan surfaces as a counter rotating wall and even to assign them explicitly a negative angular velocity equal in magnitude to that of the adjacent fluid domain . But in either case I get a runtime error with the CFX- solver. However, i'm sure enough that I've chosen the axis of rotation correctly.


1.So I'm not indeed clear in modeling the Fan in CFX using the MRF method. Can some one please help me out.


2.It is a steady state simulation and I wanted to choose the physical time scale to help the problem converge.But I didn't know how to calculate the physical time step.

Can some one please help me in this.


3.For the sort of problem I defined here, Can some one suggest a few quantities which can be chosen as monitor points.

Thanks in advance.

ghorrocks February 12, 2011 06:25

The CFX tutorial Flow in an axial rotor/stator shows how to set up MFR simulations. Follow the approach used there.

If this is a steady state model, make sure you have physically set it up to be steady state. If the heating element is providing heat, that heat needs to leave the domain somehow for a steady state to exist. In an oven I presume this is due to heat losses to the outside. You will need to include this for a steady state to exist. Have you done this?

For setting the time step, read the documentation or this link http://www.cfd-online.com/Wiki/Ansys...gence_criteria

Set monitor points to look at parameters of interest. Only you know what you are interested in.

saisanthoshm88 February 12, 2011 07:03

Thanks for the reply Glenn, I've gone through the Axial stator/rotor Tutorial in CFX. But it is a case where the Stator and rotor are simply modeled in different
frames of reference but as I found it from different sources this isn't the approach usually followed for modeling fans and blowers.

I mean the general approach is not to model the Fan in a rotary reference frame ( which indeed involves meshing the interior of the Fan increasing the computational effort and such approach doesn't stand out to be effective in replicating the physics). The approach I'm talking about is discussed much in relevance to Fluent than in CFX.

Please refer to this Fluent tutorial ( The content in the 12 th page can throw a better elucidation.) , available on the link: http://my.fit.edu/itresources/manual...f/tg/tut09.pdf

ghorrocks February 13, 2011 07:03

? I think you do not understand how multiple frames of reference work. The approach in CFX and Fluent is similar, and very simple in concept. You put the rotating bits in a rotating frame of reference and the stationary bits in a stationary frame of reference.

I have no idea what you are talking about with comments like:

"But it is a case where the Stator and rotor are simply modeled in different
frames of reference but as I found it from different sources this isn't the approach usually followed for modeling fans and blowers"- The CFX multiple frames of reference has been used by many rotating machine modelling.

"which indeed involves meshing the interior of the Fan increasing the computational effort and such approach doesn't stand out to be effective in replicating the physics" - I have no idea what you are talking about. You do not mesh the interior of the fan, you mesh the fluid domain around it. This approach is used in CFX and Fluent.

The tutorial you quote seems to use the same concepts as CFX.

saisanthoshm88 February 13, 2011 09:19

1 Attachment(s)
Well, I mean that by default any CFD solver treats a wall boundary as being in the same frame of reference as the cell zone to which it is attached.

So it shall be required to make the relative velocity of the fan surfaces zero with reference to the adjacent rotating fluid zone in the MRF method.

In the 12th page of the fluent tutorial , this is realized by defining the blades as a wall and then setting the velocity relative to the adjacent cell zone to zero.

I've even attached an image that shows the settings I was referring to. Please look into the image which shows that the velocity of the blades relative to the adjacent cell zone is explicitly set to zero in Fluent

My problem is that such options are not available with CFX.

ghorrocks February 13, 2011 20:48

CFX sets the wall velocity to zero relative to its local frame of reference by default, just like Fluent. What makes you think CFX does not do this?

saisanthoshm88 February 14, 2011 01:54

1 Attachment(s)
oh, Thanks for the reply Glenn, indeed I was confused after reading some stuff related to the implementation of MRF method in Fluent and Star-CD softwares. Now it's clear to me that indeed in CFX If the domain containing the boundary condition is rotating and the Frame Type for the boundary condition is set to be rotating, then the velocity components are relative to the local domain rotating frame of reference instead of the stationary frame of reference but in Fluent (or) Star-CD this doesn't seem to be the case.

This was clearly elucidated in the CFX- solver modeling guide:
Quote:

Originally Posted by CFX solver modeling guide
2.7.1.6. Rotating Wall
This option applies to both stationary and rotating domains and enables the wall to rotate with a specified angular velocity. The angular velocity is always in relation to the local (relative) frame of reference (that is, relative to the rotating frame in a rotating domain.).

I'm attaching an image of the settings used in CFX to implement the MRF.

But can you please tell me if the relative velocity of a rotating wall in a rotating domain is by default taken to be zero in CFX even if it is not explicitly specified by the way I did it?

ghorrocks February 14, 2011 05:52

The best way to understand these things is just to try it and see what happens. Run your simulation (just the first iteration is enough, or even zero iterations if you want) and see if the velocity on the wall is how you expect.


All times are GMT -4. The time now is 19:28.