[che_gl]

Hi all,
i have a very basic question due to rotating frame:
+ let simplify :i want to model the fluid flow a fan in a closed room

+ i would defind a fluid rotating frame in the room. Within the rotating frame the geometry of a fan is included using subtraction of boolsche operation and the exterior of fan will be defined as wall.

Now i have some questions:

+ Is there any rule of thumb of: how much bigger than the fan-geometry should the rotating frame be defined?. For my understanding: every elements within the rotating frame rotate with the defined rotational speed. If the frame is much bigger than the geometry of the fan (that don't have a frame or housing itself), there would be too much fluid that was "forced" to rotate?.

+ Since i would have a closed room, i would not really have an inlet and oulet. Could it work with the exterior of the room e.g. left hand side inlet (zero gradient pressure) and right hand side outlet (zero gradient pressure)


Many thanks in advance for any suggestion !
Best
Che

[ghorrocks]

Some general principles:

It is better to have the interface between domains in regions where the flow is simple. So it is better to have the interface away from separations and boundary layers.

Domain interfaces (GGIs) reduce accurately (slightly), use more memory and slow the simulation down (again, only slightly). So only use interfaces/GGIs when you really need them.

It works best when the flow has a low velocity relative to the domain. In other words, a straight flow will run best in a stationary domain and a swirling flow will run best in a rotating domain with the domain rotation equal to the swirling speed. This will reduce round off errors.

In your case you could put the entire region in a rotating domain. This eliminates any interfaces (desireable) but does mean lots of largely stationary flow is in a rotating domain (undesireable).

Alternately you could put the rotor in a small rotating domain which has enough clearance around the rotor that all separations and boundary layers are kept away from the interface, and then the rest of the region can be in a stationary domain.

Which of these options is best will depend on your exact circumstances.

[vuongcongdat]

Originally Posted by ghorrocks

Some general principles:

It works best when the flow has a low velocity relative to the domain. In other words, a straight flow will run best in a stationary domain and a swirling flow will run best in a rotating domain with the domain rotation equal to the swirling speed. This will reduce round off errors.

In your case you could put the entire region in a rotating domain. This eliminates any interfaces (desireable) but does mean lots of largely stationary flow is in a rotating domain (undesireable).

Alternately you could put the rotor in a small rotating domain which has enough clearance around the rotor that all separations and boundary layers are kept away from the interface, and then the rest of the region can be in a stationary domain.

Hi Ghorrocks,

- Could you explain more specific about the velocity round off errors?
- The two approaches you said are different methods? I think when you put the entire region in a rotating domain, you are using single rotating method. But when you put the rotor in small rotating domain and the rest is in stationary domain, the you are using sliding mesh. Is that true ? How come the simple problem (fan in a large space) have to resort to sliding mesh? I think the first approach is good and accurate for the nature of the problem, what do you think ?

[ghorrocks]

Round off errors - look in a CFD textbook for more details, or do some tests and work it out for yourself.

For your second point, why not just try them and find out for yourself?