[mlanahan]

Hello all,

This question is fairly general as I looking for guidance and resources on an appropriate modeling strategy for a class of flow. Specifically, the aim is to model the mass distribution of a heated Helium (~500-1100 C @ 10 MPa) coolant to impinging jet arrays for the cooling of a nuclear reactor. The Helium properties are modeled as temperature dependent. A sample image of the Jet Arrays is provided in the image labeled tillack_T_tube_layout. A sample image of a single impinging jet is shown in slot_jet.jpeg. The actual reactor will feed ~ 10 rows (in the horizontal direction) and ~ 200 columns deep (into the page) of the slot jets shown, which is probably not reasonable to simulate in it's entirety so simplifications will have to be made. The whole array will have to be fed using a single coaxial pipe, which will also function as the return. The mass flow rates are large and I expect that the flows will be ~ Re = 1e4-1e5 in the jets, and at least 1e4 in most of the distribution pipes, depending on design decisions. It is my task to manifold the access pipe to the jet array.

The challenge is presented by the non-uniform heat flux incident on the top of the jet tiles, an example of which is provided in the image entitled "kr_smoothed_hf.png" - the profile can be imagined to be 2D and running horizontal on the tile image. I believe that this non-uniformity and scale will create a hot-striping effect, and the helium exposed to significantly higher temperatures will expand, forcing less flow to go to these channels and exacerbating the temperature differences and thus forcing the flow distribution to behave in an unstable manner.

The coupled nature of this flow is complex, and I want to be sure to capture this effectively through simulation. I have found that modeling the jet using any of the two equations models with appropriate calibrations can capture at least the expected temperature distributions quite well when compared to experimental data. In this case however, I have limited experimental data to compare against- it has been difficult to find literature on the subject.

From a modeling perspective: Should I consider more exotic approaches than two equation models, such as LES or Reynolds Stress Models for modeling the distribution of the flow in the manifolding? I have read Pope and the Fluent Manual, but couldn't find information specific enough to convince me that more complicated models will do a better job of capturing the behavior.

Regarding computational resources: I have a 256 GB Memory 20 Xeon Gold 5th generation Processor machine for mesh generation, pre and post-processing, and access to a computing cluster that I've tested up to ~80 cores 512 GB of memory but I'm sure can handle more depending on how Fluent handles the additional resources.

I am using Fluent as my simulation software and ICEM as my meshing tool.

Thank you very much for any advice

[LuckyTran]

Going to LES or RSM models is changing the physics. You only really need to go there if you think the physics in those models affects what you are doing.

You should convince yourself that your problem (the helium heating up) is not being modeled adequately by the two-equation models. If the concern is just how the flowrate of helium changes... Is the flowrate of helium not being adequately captured by your two equations models? Is there some crazy integral scale turbulent structures strong enough to influence the flowrate? Can you already tell me right now what the specifics names of these structures are in your application? If no, then either they don't exist OR even if they did, you wouldn't recognize them.


It sounds to me like you are not yet convinced RSM or LES provides any benefit. So don't bother. You'll just be wasting your computational time and worse, your time.

[mlanahan]

I think you make a good point here.... I have not really spent adequate time thinking about WHY two-equation models would not capture the dynamics of concern from a fundamental point of view. Here are some initial thoughts: I do believe that there will be a wide range of length scales in the turbulence ranging from the width of the manifold to the entries of the coolant distribution lines, down to the slot jet widths. Two equation models consider both a length scale and a time scale, so any inaccuracies would have to result from incorrect assumptions in the production/dissipation of turbulence, or non-linearities in the turbulent viscosity i.e. secondary flows.

I would suspect some sort of system level instability due to the change in density of the gases created from large temperature differences, which would be of concern at the inlet and outlet plenum - but I have no reason to believe two equation models could not capture this behavior unless the instabilities introduce distortions in the flow . I have heard (but not verified) that two equation models fail to capture large scale vortex generation due to thermal mixing - of concern at the outlet manifold. I am also fairly convinced that the underpinning assumption of homogenous turbulence in any eddy viscosity model is dubious at best in thermal mixing with large scale vortex structures, and perhaps even in the transition between manifold and distribution lines, which indicates to me some modeling considerations need to be made - inhomogeneities can be handled by elliptic relaxation RSM or LES I believe. I ultimately have no idea how these dynamics will affect the system performance, or model prediction therein.

In short - I don't really have the expertise to say what is the correct approach. I think a more thorough survey of the literature is probably in order. It seems to me that a good strategy would be to create the simplest possible model to capture the dynamics of interest here - for both the inlet and the outlet manifold, and try both approaches to examine the differences, if any, in the flow predictions.