[boreal]

Hello people

Nice board. I'm new here and probably looking for help from time to time, because I'm dealing with CFD simulations again recently after a long time out.
Currently I am trying my hand at a natural gas reformer. I am interested in the residence time distribution within the reformer as this can give an indication of the conversion of the natural gas.
The model looks simplified as follows. Inside the reformer, the gas is distributed by a structure of fins. The simulation is laminar.

reformer.jpg


I calculate the residence time with an additional equation analogous to the tutorial "Junction Box Example 2: Integrated Residence Time Distribution" in the CFX Modeling Guide. i.e. I use the one additional transport equation with the variable RTD and Source with unit [s/s].
Now I am struggling with the interpretation of the results. I put an x/y plane in the reformer and show me RTD there. What I would have expected is that the Residence Time increases constantly from the Inlet to the Reformer to the Outlet. Instead, there are higher residence times at the impact flow (inlet) and lower again towards the outlet. If I display RTD on a single flow line, then also in this case the RTD first increases, then decreases again and increases again towards the end. Am I correct in assuming that this is wrong and there is an implementation error somewhere? Or have I misunderstood RTD?


Thanks for shedding some light on this.
boreal

[Gert-Jan]

What you see are mixing effects. Liquid with high RTD mixes with liquid with low RTD giving you an average RTD at the outlet. This is the same as if you would have regions with low and high temepratures that mix towards the outlet.
You expect some kind of residence time distribution at the outlet, of liquid with short RTD and liquid with high accumulated RTD. For this you need discrete values. This can be achieved by studying discrete streamlines in Post or using massless particles as Lagrangian Particle Tracking.
Then you have discrete values where the time is accumulated giving the real Residence time per strealine or particle. Both methods provide the time by default. No additional variables required.

[boreal]

Hi Gert-Jan


Sorry for my delay. But that's for your reply, it was very helpful for me!


Merry christmas,

boreal

[boreal]

Hi guys


I just thought again about Gert-Jan's answer.

What i still don't understand: How can there be mixing effects when i don't have a diffusive term in my equation for the residence time?

With temperatures it makes sense for me, since there is a diffusive part in the energy equations.

Edit: It's probably important to state that i set the kinematic diffusivity to zero. In my understanding this should lead to now diffusion - hence no mixing. Am i mistaking here?

Thanks and best regards,

boreal

[ghorrocks]

How can there be mixing effects when i don't have a diffusive term in my equation for the residence time?

There will be mixing from:
* molecular diffusion (you say you turned this off)
* flow induced diffusion (eg turbulence, but laminar structures can do it as well)
* numerical diffusion (can be reduced by more accurate modelling - ie finer mesh)

[boreal]

A happy new year to you all!


@ghorrocks
Thanks for your reply.

It's your second point which I don't really understand. What exactly leads to flow induced diffusion in a laminar flow? Do you have some literature about it? I couldn't find a proper explanation in my books.

Numerical diffusion could be an issue. There are a lot of details in the geometry (e.g. narrow channels) and the mesh resolution is not always satisfying. But I don't see a huge impact if I refine the mesh.

Originally Posted by ghorrocks

There will be mixing from:
* molecular diffusion (you say you turned this off)
* flow induced diffusion (eg turbulence, but laminar structures can do it as well)
* numerical diffusion (can be reduced by more accurate modelling - ie finer mesh)

[Gert-Jan]

Given your geometry, you have a wide mid section and a narrow inlet and outlet.
Imagine you zoom into 2 numerical elements in the wide mid section, next to each other. Then, starting with t=0s on the inlet, imagine that in 1 element the time of the fluid has accumulated to a residence time of 5 seconds, in the other next to it, the time has accumulated to 6 seconds.

Now, imagine that on the way to the narrow outlet, the 2 elements are connected to a single element. Then the time in that element will be (5+6)/2 = 5.5 s. This is averaging, and can also be also considered as mixing!!!!!
This simple numerical exercise is valid, provided both elements have the same flow and that the additional residence time is small. Otherwise, it might be 5.6, or 5.6 s, depending on the local velocity and element size.

This averaging is there because there is only 1 scalar time variable. If you want to distinguish between the fluid of 5 and 6 seconds, so ignoring the averaging, then you need to follow streamlines, of massless particles. These are discrete numbers, without mixing of averaging. But I already explained that before. Apparently you did not understand, or I was unclear.

[ghorrocks]

On your question about what laminar flow could increase diffusion - flows like the Von Karman vortex street or Rayleigh Taylor instabilities are laminar flows which will increase the effective diffusion. On the micro scale the diffusion is still happening at the molecular diffusion rate, but on a longer length scale the diffusion is accelerated.

[boreal]

Hi ghorrocks & Gert-Jan


Thanks for clarification.

Steamlines of massless particles is no option for me. As an input parameter for a mathematical model i need a distribution with volumetric flow on the y axis and residence time on the x axis. Therefore, the discused approach seems to be appropriate.

I was confused because i didn't expect mixing effects and questioned my setup. But now i begin to trust the results.



best regards,

boreal