[KarenRei]

Hi everyone - I'm new to OpenFOAM (so apologies in advance!) and have relatively limited experience with CFD in general. I'm looking to model a virtual expansion nozzle for a rocket motor and have been reading through the documentation and tutorials with that in mind. I want to make sure I'm approaching this in a reasonable manner.

Since I primarily just want to evolve the nozzle shape (implementing it as an optimization problem for scipy.optimize), and because I'm just a beginner, I ruled out modeling the combustion chamber; hybrid rocket combustion in general is a pretty complex process, and the variant in question here has multiple fuels and significant chamber lining ablation which makes it even more complicated. So I figured I'd just use my generalized results from CEA2 and inject the gases leaving the throat into the simulation, and just use a relatively simple solver that handles compressible fluids.

Since then, though, everything I've read and the more I've thought about has made it more complicated.

One, the exhaust gas in question is not the same as the air around the rocket, it has a different average molecular weight and other physical properties. So the solver has to be able to handle varying mixtures of gases, not just temperature/pressure/velocity.

Two, the combustion isn't complete at the throat. Particularly the aluminum takes time to burn, and then condenses to solid particles, releasing a lot of energy in the process. So we have a continual release of heat along with changing average molecular weights. I figured that maybe there's a way in OpenFOAM that I could manually have it inject additional heat and change the properties of the exhaust downstream based on what CEA2 says happens with the exhaust as it expands, but doing it in an accurate and automated way (for differing nozzle configurations at that) would surely be tricky. Also, with a virtual nozzle there's some potential for combustion of fuel-rich mixtures with atmospheric air, which CEA2 would be of no help on.

So maybe the right solution is to use some form of chemFoam to model the continued combustion of the material leaving the nozzle? Even though that would obviously be very slow, I know that chemical reactions have to be simulated with very fine timesteps. Either way, I guess I have one advantage in that the output from one nozzle configuration's run can be fed in as the starting environment for the next nozzle configuration.

I'm a bit confused by all of the numerous solvers. In particular I found "chemFoam" online, but the "tutorial" is just a tarball with no explanations... then I found people refer to "dieselFoam", which has a good tutorial writeup with it and involves chemical reactions... but what's the difference between dieselFoam and chemFoam (and "reactingFoam", found that one mentioned too)? Can they handle condensation? What sort of solver do I actually need to be using?

Either way, from the look of it I'm going to be needing to learn CHEMKIN also**. And I'm going to need to figure out how to extract at least the net force on the rocket (very simple shape) in batch for the optimization function to use, although I'm sure I'll run into how to do that sooner or later in the docs or in my searching. Right now my top priority is to make sure that I head down the right path toward doing this right, so I figured I'd register and ask: what do you people who actually have experience with OpenFOAM think I should pursue as a path towards this goal?

**Update, re CHEMKIN: or maybe I don't need it? therm.dat appears to be already provided and generic to a wide range of reactants (looks like what CEA2 uses) and chem.inp looks like something that I could probably piece together on my own (hopefully!). Good, because CHEMKIN appears to be a commercial product :Þ

[KarenRei]

Update: I'm trying to see what I can do on my own here. I searched through studies trying to find out the reaction rate coefficients needed for aluminum (all of the ones having to do with C, N, O and H are all widely available). I found a couple research papers, one of which cited several other research papers. It gives the Arrhenius constant and activation energies for the oxidation of aluminum powder from different studies at:

7,41e10, 222
6,31e3, 119
1,84e12, 230

Gee, got to love it when your results vary over nine orders of magnitude They cite some other studies elsewhere and I think the study's values of 7,41e10 s^-1 and 222 kJ/mol is most realistic.

There's still a problem though in that this is just Al+3 (solid/granular) <=> Al2O3. But it's not pure aluminum dust going through the throat. CEA2 estimates it as:

Al2O3: 42,3%
AlOH: 25,2%
Al2O: 13,1%
*Al: 4,8%
*AlO. 3,9%
Al2O2 2,5%
AlH: 1,2%
Al(OH)2 0,6%
Al 6,4%

Really, if I want a realistic model here I need the half reactions. But I can't seem to find them anywhere. Clearly CEA2 is dealing with the half reactions, but how it's doing that and where it's getting the figures, I have no clue. I don't see them in the data files.

Still don't understand the difference between all of the different chemical reaction solvers... I'll take a break from trying to dig up reaction data since I'm not getting anywhere with that now and see if I can get a better understanding of the differences between the solvers.

Update - left a post on the Cantera page asking if their software can output the sort of data I need. AFAIK Cantera is free and seems to deal with these sorts of things. Also, I found a list of OpenFOAM solvers:

http://www.openfoam.org/features/standard-solvers.php

But the descriptions are horribly vague - for example PDRFoam and XiFoam have the exact same description. I mean, the description sounds good - turbulent mixing will surely play a role at the boundary layer during expansion - but what's the difference? Does either one support condensation, including the release of the latent heat and the effects of the condensate on the flow? I don't need individual particle tracking, just general bulk solutions on how the combustion steady-state will affect the overall pressure on the wall segments that define the rocket. Also, rates of flamefront propagation aren't important to me, only the steady-state combustion solution. A lot of the talk in what I've seen so far in the XiFoam tutorial (can't find a tutorial for PDRFoam) seems to be about the turbulence calculations being used for flamefront propagation calcs. And there's lots of model options related to flamefront propagation that are beyond me, and the goal seems to be solving for flame speed.

Hmm, maybe if whatever solver I need to use doesn't support condensation I could maybe fake it by adding in a fake reaction to generate an extremely high molecular mass "gas" at a given temperature ... aka 10000 Al2O3 => 1 Al20000O30000, with a huge negative temperature exponent so that it condenses quickly at the boiling point but not before then... That would work, right?

[KarenRei]

So, a followup, in case anyone's actually reading this. Thanks to some help from the Cantera people, I ran into this the other day:

"Report No. 2, 2011 Al/H2O Combustion"
http://www.tfd.chalmers.se/~valeri/A..._newAL_H2O.pdf

Apparently this guy's model for ALICE combustion (Aluminum-Ice) involves a chem.inp file containing 31 reactive aluminum-based species involved in 133 reactions. A couple of the more important reactions' arrhenius parameters are mentioned in the paper, but obviously getting the chem.inp file would be ideal! I've emailed him (hopefully at the right email address... I found a university address, but it says he's retired now)... I'll keep following up, hopefully it's possible to acquire.

As for OpenFOAM itself, since I haven't gotten any help explaining the difference between the different combustion solvers, I'm going to go with my gut and assume that rhoReactingFoam is best, since density is relevant and the two solvers focused on turbulence seem to be mainly about flamefront propagation, and since I'm not burning droplets like in dieselFoam, and chemFoam seems to be focused on single-block cases. Hopefully I'm interpreting all this correctly. Feedback would be quite welcome! That is, feedback on A) whether I'm headed down the right direction concerning solvers, and B) also it would be nice to know whether the solver will handle condensed species on its own or whether I have to fake it.