Learning Forum

[Alex Chiella]

Hi! I'm simulating some sort of y-elbow piping, as of now I have a prevalently hydrogen stream outflowing in ambient air.
The flow is going upwards, in the same direction of my Y  axis. If I set gravity at -9.81m/s^2 (with force of gravity pushing down), I get this:

The way the low density H2 flow is stopped and pushed back by the denser ambient air really leaves me puzzled, it really looks like the buoyancy effects are reversed here.

I tried changing the sign of the gravity value, and I see something which embodies the expectation I originally had:

The contour inside the pipe fades to red because the hydrogen stream is partly premixed with air.
And finally if I try to set it to 0 m/s^2, I get the obvious confirmation that gravity is playing quite a role in the first contour I uploaded.

I'm using the ideal gas model with operating density set to 0.

Am I beautifully mixing things up, or is there actually something strange going on?

 

[Rob]

What boundary conditions have you set? Is the model converged? What is the flow doing? Ie don't just focus on the species contours. 

[Alex Chiella]

Hi Rob! Always excited to get replies from you!

The square at the top is the symmetry cross-section of an air cylinder, that I set as an atmospheric (0 Pa gauge) pressure outlet. I am using the symmetry conditions on a halved domain.
I have a mass-flow inlet with a very low mass flowrate, which results in ~10m/s at the 50mm pipe inlet.
The solution seems to have converged alright, I mostly checked mass balance during iterations and the static pressure contour. Very mild static pressure differences through the whole domain by the way.
The hydrogen flow is exiting through that outlet at about 5m/s at normal conditions, a pretty ordinary scenario. 
As I said, this is a sort of y-elbow pipe, with an inlet, an outlet (pictured) and a vacuum-entrainment opening. For this given flowrate there's not much going on at the entrainment opening, so the outflow might be compared to a gaseous stream going through an elbow at low velocity.

This is what I get from Y velocity coloured symmetry plane pathlines, so the flow is indeed being pushed back to the ground.

I do know that light gas plumes tend to have a boost in Y velocity while dense ones are bent down, but I struggled finding more stuff about that online.

[Rob]

What material have you got for the backflow? And what operating density? If you initialise the domain from the outflow & then use the exact reported density for the operating conditions what does the result look like?

[Alex Chiella]

The backflow is pure air. The operating density for the domain is 0, were you asking about some pressure outlet specific density that I'm not aware of?

If I got what you mean about initialising the domain from the outflow, the issue is that the relevant value is the velocity at pipe outlet, which in my CFD domain is just an internal surface. The cylindrical exit region is not suitable for initialisation, because "pull" behaviour (via a radial velocity outlet) is inherently different from "push" (flow with a given inertia through a pipe opening into the atmosphere). But I might have entirely misunderstood your suggestion here!

[Rob]

Ah, read the 7.3.1.5 https://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v242/en/flu_ug/flu_ug_bcs_sec_operating.html section - you want to make the (rho - rho_op) term equal to zero. 

[Alex Chiella]

Dear Rob, hi again!
Took some time to try to get the Operating Density straight, thank you a lot for pointing me there!!

I tried running a slightly different geometry, with everything made from scratch including the SpaceClaim CAD (just to make sure I did not screw up axes choices and whatnot). I set the Operating Density to 0.142 kg/m^3, which is the value I got at the pipe outlet on precedent runs. That's pretty much the hydrogen density at normal condition, slightly increased because of some light air dilution. While in this new case I get some slightly different results (which is understandable), the issue at hand is exactly the same.

This is what I get using the Operating Density above both for ideal gas and incompressible ideal gas:

 

The lighter gas gets push downwards, again.

The pathlines are mushroom shaped:


I'm going to rerun the same exact simulation I showed above using non-zero Operating Density too, just in case.

P.S. For some reason now I have Fluent running the hybrid initialisation as soon as I try to Edit... the project (I'm working via Workbench because I'm hoping to perform a parametrical analysis one I get the DP0 down correctly). Have no idea how it happened nor how to stop it, in this way I can't checkout the saved solution because it gets wiped away. What did I do wrong this time?

[Rob]

I said the operating density needs to be that of the surrounding fluid, ie whatever the backflow condition is. Not what's leaving the domain. So, air at ambient conditions is around 1.182345kg/m3. I suspect you're seeing the effect of the hydrostatic pressure if you set the outlet as a uniform pressure. 

If you want to check the results you want the Solve part in Workbench. You've just found one of the many reasons we tend not to use WB! Note, there are parametric tools in Fluent standalone too, that may make life a little easier. 

[Alex Chiella]

Hey Rob!

Tried to do as you suggested and it seemed to work! I also went on to do some more research on how Operating Density works in Fluent and it really does sound like, in my case, that was just the only appropriate way to set things up.
I happened to notice it just now because it's with low velocities and high density gradients that the hydrostatic pressure takes the lion's share, so to say. I was a bit stuck with the school mindset of just forgetting about the gravitational head whenever you're dealing with gases.

So, thanks a lot for your help, big learning opportunity for me!!

[Rob]

You're welcome. I only know because we found the same effect with a domain that was around 500m high - the air flow was unusual to say the least. That was about 20 years ago!