Impact of the intlet and outlet design on flow in a pipe

[frederic.gaillard.7]

Hello, 

i'm currently studying the effect of the geometry of a pipe. Ultimately i want to know what are the best dimension of the pipe to maximise the air flow toward the fan (araignee_vitesse.png) . Adding a bevel or not on edges are also a topics on this study.  

My question are more oriented to the geometry of the intlet and the outlet. 

1. Intlet : 

Is that correct if i put P=0 on every surface of my air bloc ?

or it's better to simply add a surface condition on the two littles entrance and suppress the rest of the air bloc ?

2. Outlet 

This part is more complex because the outlet is a really environnement where there has physical wall, pressure and opening. 
For the moment, i'm only interrested about what's happening in the pipe so knowing that, what should be the best outlet design to freely evacuate the air, regardless the pipe design. 

And finally,  from your perspective,  do you think the flow who's evolving in the outlet can i have an effect on the flow in the pipe ? 
Before i go further i want to put some light on this aspect. 
Thx You 
Fred 

PS     ''araignee_vitesse'' is the velocity profile in the region of interest (L pipe) 

[Jon.Wilde · Accepted solution]

Hi,

Nice to see another rotating region model 🙂

Be careful that you only have flow going in one direction over the larger P=0. Otherwise your pressure drop might be out - if this is the case, consider making the length longer.

Do not put P=0 on all regions, you really want a clearly defined inlet and outlet.

I would say that yes, the outlet flow can have an impact as it will all affect the overall pressure drop, right?

Thanks,

Jon

[frederic.gaillard.7]

Understood ! 

Thx for the advice. 

I change a little bit my set up : 
1.  The  ''square outlet'' is now short and circular, where I define P=0 on the circumference surface .  

2.  The intlet is more longer (achieve a fully developped flow) and with only one pressure boundary condition at the top 

I think this will allow a unidirectionnal flow and a clear defined intlet and outlet. 

Furthermore, what should be the best turbulence model for modelizing the pipe ? 

I'm considering to use SST-Komega with 7-10 wall layer instead of the k-epsilon  

Merci beaucoup have a great day 
Fred

[frederic.gaillard.7]

i just notice that in my last design, air comes and goes throught the same boundary condition. I have to change it 

[Jon.Wilde]

My apologies, this might be worse... I would try to keep the outlet really simple and be sure that flow is uni directional over it.

What actually happens in reality, how is it set up and tested?

The inlet looks better 🙂

[frederic.gaillard.7]

Hi John, 

the set up, is taking part in huge oven volume. If you check the CFZ in the first post, you will see the dimension. The Z and Y coordinate are the real length. The x coordinate are only the 1/8 of the total length. As you previously said, this set up allow flow in and flow out throught the same surface. 

To avoid this, i first design a short and circular  outlet around the RR

   - model                (short_outlet.png)

   - velocity vector    (velocityvector_short_outlet.png) 

one of the problem that i notice was a suction on the upper part of the outlet caused by a high pressure gradient (see picture) 

Then i design a longer outlet to have a lower pressure gradient  and supress this effect (straight_edge_spider.cfz) 
what i realize is for both scenario i have a transitional regime where the flow is not stabilize. So i guess my two design are correct i just need to run them a little longer. 

A question remain : is it preferable to have a longer outlet or a short one for this specific application 
Thx John 

[frederic.gaillard.7 · Accepted solution]

Hello Jon, 

Quick follow up on this topic, 

This geometry allow an unidirectionnal flow through the whole model with P=0 on both end. 

The huge air box allow to create convection pattern which is directly affecting the flow who enter in the pipe. 

The simulation was perform with k-epsilon - IWF 
the result seems to convergence after 500 iteration although the Y+ was under 35.
Furthermore my NAR values were over 100 close to the surface of my model. 
But the main problem was solved 

Thx to you and david . 
Regards 
Fred