[ThunderstruckGER]

Hi everyone,

I'm new to OpenFOAM and CFD in general. I'm sorry if I'm asking some rather simple questions... Right now I'm struggling to gain an overall picture of how to approach my case and its step-by-step implementation in OpenFoam. I would appreciate any input whether I'm heading into the right direction!

What I'm trying to do is simulating total outward heat losses of a sub-assembly. I'm having trouble understanding how exactly the environment of my system should be included into the OpenFoam simulation.

Thus far, I have simplified my geometry into a cylinder inside a cylindrical shell. I uploaded a quick thumbsketch of my geometry here. There is a gas flow entering through the inner cylinder, which consists of an electrically heated active coal textile and leaving the system through the bottom circular ring. The inner cylinder is the heat source of the system, held at a constant temperature and the heat transfer is directed outwards (= external heat losses of the system). In order to work with this multiphase case, I plan on using chtMultiRegionFoam.

Since the geometry is radially symmetric, I have quickly managed to mesh a quarter of this geometry. But I'm not sure how I should continue with my case.

Because I need to do a transient simulation of the heat transfer, the outer shell of my system is not isothermal, but will heat up over time. Therefore, I won't be able to define the boundary between my system and its surrounding environment as isothermal. Also, as a consequence, radiative and convective heat transfer rates to the surrounding environment are not constant in the simulation.

Now I have a couple of questions about how one usually sums up such a case in a simulation:

Will I have to mesh some kind of a "sample room", set standard conditions for it, set the walls, the bottom and ceiling of the room as isothermal and put my sub-assembly into the middle this room?

Or will I be able to mesh only the geometry of my sub-assembly and then later set boundary conditions at my outer lateral surface area with pre-calculated/estimated/approximated radiative and convective heat transfer rates?

In case a mesh of the sub-assembly itself is sufficient: Will I be able to use changing heat transfer rates instead of fixed ones? Are there existing approximate equations for the convective and radiative heat transfer rates, or will I be required to add them somewhere in openfoam? I guess it could be justifiable to set the convective heat transfer coefficient to a fixed value and to somehow feed openfoam the approximate equation for radiation heat transfer between a small and a very large surface.

These are my thoughts so far.

I hope my questions make sense and thanks in advance for your attention and any help you can offer!

[pbrady2013]

Hi,


First up I like your approach of starting with relatively simple solutions and building complexity from there. Jumping in to the full case straight away is usually a very bad idea.


To further reduce you could start to axisymmetric but the returns may not be that great.


In answer to your specific questions, with my work with natural and forced convection I model the whole space. I find that it makes a significant difference with more realistic results as I push the solution domain out. I have found that the cells in the outer domain can be quite large though, particularly the further away from the assembly you get.


For example we had a small "chimney" but heated from the sides - think the double skin of a building with brick on one wall and glass solar panels on the other with a small air gap in between - set up in a lab. When simulating to validate the experiments we got the best results when simulating the entire laboratory room. Looking at the results the answer was obvious in that the equipment within the laboratory as well as the pressure and temperature stratification in the room lead to highly non-uniform flows entering the base of the chimney. These flows were then drawn into the chimney and drove the unsteady observations. The effects increased both as the chimney heated and the room heated through the course of the experiments.


There should be a paper coming out on this shortly as we are just tidying it up.



Hope that helps,
-pete

[ThunderstruckGER]

Hey,

Thanks a lot for the nice words and your comparison! It was very interesting and also helpful in clearing my thoughts about what I actually want to achieve with my simulation and how I could get there.

Now, before I explain my thoughts, I think I need to explain the wider context of my simulation first: I want to simulate the impact of a radiant barrier between the active coal textile and the outer cylindrical shell (e.g. by thermal spraying the inside of the cylindrical shell). I want to build a simulation for the heating process, modify material properties and compare end results using different materials/constructions. Therefore my simulation set-up a) needs to be comparable by all means and b) should preferably produce roughly realistic results. I'm not sure yet to what extent b) can be achieved simultanously. After reading your answer right now I tend towards focussing on a) and maybe checking on b) later on.

Your experiences stated above made me realize that my approach can consider the impact of the surrounding environment on convective heat transfer in no way accurately. The sub-assembly I want to model is a rescaled version of the original one and part of a larger prototype device. Originally there are fittings on and next to this cylindrical module, as well as flanges for sensors on the module itself and so on. Also the closed top and bottom outlet of the module are not to be modelled (yet) in order to simplify the geometry. What I'm trying to say is that the simplification of turning it into a radially symmetric, cylindrical geometry have probably altered the heat losses (radiative+convective) a great deal already... Assumed my approach is somewhat feasible, because of those (hot) fittings, etc. and the reflective materials used (steel, alu), the real heat losses would maybe be less than the final result in OpenFoam as I will try to explain below.

Since I opened this thread I reviewed the steady-state equations for radiative and convective heat transfer rates of my geometry. That helped to reflect on a possible fitting approach.

Regarding heat radiation, basically, a way to make my simulation comparable is simulating only the heat radiation of my sub-assembly and excluding the reflection of any heat radiation by the surrounding room (and the objects in it). That's pretty much what the approximate equation for radiative heat exchange between a small surface enclosed by a very large surface does. For example I was always told to think of a household radiator in a room.

This would be like simulating my sub-assembly in an empty, "infinitely" large room filled with air. The temperature of the air and the walls of this "infinitely" large room would then be set at standard conditions. Since any reflections of heat radiation are excluded, I would obtain the maximum radiative heat transfer for my model from my simulation. Therefore it's probably the most "conservative" way I could calculate radiative heat losses. I could approximate this model by making a concentric, cylindrical mesh with an outer diameter of ~100 times the diameter (and consequently ~100 times the lateral surface area) of my sub-assembly.

Another, much easier way of dealing with this could be setting the emissivity of the surrounding room walls as 1. This would eliminate reflection of the surrounding room back to the sub-assembly as well.

Heat transfer by natural convection in an empty, closed room should be mostly independent of its size. (Provided the boundary layer is not interrupted by placing the room wall in close contact with the heat source.) So there should not be much of a change by scaling the surrounding room up/down.

In case the general idea looks good to you, I'm guessing I could mesh a small room (~two times the diameter of my module as surrounding space?) which depicts convective heat transfer correctly (?) and then set the emissivity of the surrounding room walls to 1. In that case I would be able to keep the mesh as a quarter of my radially symmetric geometry for now, which would be very nice. Then, the next step would be reading up and working with chtMultiRegionFoam.

What are your thoughts on this? Does this sound reasonable, or am I missing something? Am I wrong about the influence the size of an empty room has on natural convection?

[pbrady2013]

Hi,


Great to be of some help.


I think your general approach seems sensible.

As you have a prototype I'd always try to get some real data at least for one configuration. That way you can validate your model in one method/configuration - then take a leap that the process holds, change the materials and have some faith in the answers.

With respect to the room at 2x I think that answer depends on how long you are going to run the simulation rather than boundary layer interactions with the walls per se. My problem was that with long experimental runs the internal air became startified due to the heat I was injecting through the experiments. This created two problems: (1) the chimney outlet would choke as the thermal gradient dropped and (2) as the stratification descended through the room we started so suck hot air in through the bottom.

As for the radiation, I think you should be fine there. The question to always ask: is my assumption reasonable? In this case, you should have a rough idea of the energy you are trying to dump through the radiation, right? Could be as simple as assuming an ideal black body at a given temperature. Then you can do a very rough estimate on what will be reflected for different objects and distances. That can give you an idea of sanity for your assumptions.

This would be like simulating my sub-assembly in an empty, "infinitely" large room filled with air. The temperature of the air and the walls of this "infinitely" large room would then be set at standard conditions. Since any reflections of heat radiation are excluded, I would obtain the maximum radiative heat transfer for my model from my simulation. Therefore it's probably the most "conservative" way I could calculate radiative heat losses. I could approximate this model by making a concentric, cylindrical mesh with an outer diameter of ~100 times the diameter (and consequently ~100 times the lateral surface area) of my sub-assembly.

I'm not sure about the logic in this quote. Wouldn't this configuration be non-conservative? As you say, this is the maximum cooling rate, which would be unlikely to be achieved in real life?


Cheers,
-pete

[ThunderstruckGER]

Hi,

As you have a prototype I'd always try to get some real data at least for one configuration. That way you can validate your model in one method/configuration - then take a leap that the process holds, change the materials and have some faith in the answers.

I know the rough steady state temperature of the prototype module (~100°C at the outside). But depending on how much my rescaled geometry changes the heat transfer and how much a radiant barrier improves the heat insulation I think this might be of a limited value.

With respect to the room at 2x I think that answer depends on how long you are going to run the simulation rather than boundary layer interactions with the walls per se. My problem was that with long experimental runs the internal air became startified due to the heat I was injecting through the experiments. This created two problems: (1) the chimney outlet would choke as the thermal gradient dropped and (2) as the stratification descended through the room we started so suck hot air in through the bottom.

The run time of the device is set at 10 minutes for now. To be honest I have no hunch if this will lead to similiar effects you stated above. But I guess you were talking about hours-long runs or something like that?

As for the radiation, I think you should be fine there. The question to always ask: is my assumption reasonable? In this case, you should have a rough idea of the energy you are trying to dump through the radiation, right? Could be as simple as assuming an ideal black body at a given temperature. Then you can do a very rough estimate on what will be reflected for different objects and distances. That can give you an idea of sanity for your assumptions.

I have thought about that, too. I have already estimated my steady state heat transfer rate with the outer temperature, an estimate of the steel emissivity and the approximate equation I talked about earlier. (Taking the equation of a black or even gray body does not take the temperature of the receiving area into account.) With the aproximate equation, heat transfer rate should be about 500 W if no reflection is going on at all. That should be a bit more accurate than the equation for a gray body at given temperature, which gives about 830 W. So I'm having a very rough idea what I can expect of the simulation.

'm not sure about the logic in this quote. Wouldn't this configuration be non-conservative? As you say, this is the maximum cooling rate, which would be unlikely to be achieved in real life?

The heat loss is to be reduced. So, the highest cooling rate would be "worst case". The higher I'm estimating heat loss by setting up my model in a specific way, the "better" (=lower) the heat loss will automatically be in practice. That's why I called it conservative. (I don't know, maybe there is kind of a language barrier going on here for my part?)

What I want to try now is the following: I will update my mesh to twice the diameter to be able to start working with chtMultiRegionFoam (there are still many things unclear to me with the solver and the simulation, that's why I want to start working on that asap). I need to see if I get everything to work as planned. Then I could compare the outer temperature of my simulation with the outer temperature of the prototype. This should give me a clue about the reliability of the simulation and about what possibly needs to be changed.

[pbrady2013]

Hey,


Sounds like you have a pretty good handle on the physics and what you are expecting. I like the idea to push the boundary out so looks good.


Run the models, build the complexity and test with sanity checks as you go.


Please let us know how you progress - if you can commercially.



Cheers,
-pete