Combustion Design

[CH3NO2]

PS - I can't model transpiration BLC in Solidworks. The program crashes when thousands of holes are introduced into its calculations.
I'll have to do all the basic flow modeling without BLC.

[finiteparts]

Hi Tony,

One thing that is common is to determine if a finite element mathematical solution (CFD, Stress analysis, etc) has any grid dependency. This is done by refining the mesh sizing and rerunning the case. If there is any change, you might not have captured all the relevant physics at the previous mesh sizing. Usually, this is done several times till you see a stable solution, irrespective of mesh sizing. You might want to try this.

Additionally, with the swirler, it will be very dependent on the type of turbulence model that you selected and also on how the solver is resolving the fluid to wall interaction. There are two primary ways this happens. One is to mesh the near wall region with highly refined cells so that you "capture" the boundary layer turbulent energy production. As can be anticipated, this is very computationally "expensive" since it requires a huge increase in the number of nodes and cells. The second method is using a wall function. This uses relations to calculate the effect of turbulent boundary layers without having to specifically model the physics. My guess is that is what Solid Works is using. For a wall function to predict correctly you need to make sure the mesh is set up so as to give the correct stand-off height from the wall, which is call y+ in the CFD world. I would recommend reading through the Help in SW to make sure that you are capturing that correctly. Just because you get an answer doesn't mean you have to believe it.

Yes, most of the time, meshing programs have rules in them to insert a minimum number of nodes in spaces so as to capture the physics, so if you have thousands of tiny holes, your number of nodes in the mesh with sky-rocket beyond the memory capability of your computer. You can see if SW offers a porous mesh as a "poor-mans" way of simulating the cooling flow through the wall, if you really want to do that.

As for the percentage flow for cooling, I have no idea. The real problem is you have no idea of the heat load on the liner, so how can you have any idea on what the cooling flow needed to counter that unknown?

My personal thought is that I have seen only limited need for adding liner cooling. Cooling the dome corners has been the one exception. There is the potential for burning in the near wall, cooling flow, which may make things worse (there was a paper done by G.J. Sturgess when he was at Pratt on this as I remember...but I just can't seem to find it now. I will post a link when I do). I would suggest making the liner uncooled, then let the metal oxide layers tell you if you need cooling.

Good luck,

Chris

[finiteparts]

Just went through this post and fixed all my images. I need to get back to this post soon...I have so much more to add when I get some free time.

Chris

[finiteparts]

HI Tony,



I thought that I would move this conversation over here, since it is more in line with this posts topic and I didn't want to hijack anyone else's posts.

So in reference to this,

Jul 2, 2018 18:22:03 GMT -4 CH3NO2 said:

Hi Chris,



Never mind the numbers shown in the rainbow colored legend. I dont know why it always shows the low numbers but I do know they are not correct. So I dont bother with the legend numbers anymore. I know the legend isn't correct because in simulation, I vented the flame can into a hard vacuum and the legend still shows impossibly low deltas. So the legend is useless. The CFD is however useful in illustrating the flow field directions, patterns and it gives a good breakdown of mass flow partitioning when needed. Everything in Solidworks CFD seems to work great... except for the legend.



Using the Lefebvre equations I sized the total orifice flow area to create a ~2.5 PSI delta P with an assumed 0.57 Cd at a 3.5 pressure ratio.



And yes, ideally, it's best to have near zero velocity just before the orifice, but in my experience the majority of dynamic pressure is conserved at 4X due to V^2/2g. 4X is my "safe" practical, minimum ratio. I try not to go less than 4X unless there is some other demanding compromise forcing it. Going up from 4X is always good until it comes at the cost of other compromises. The ratio is of course flexible depending on the circumstances.



When I start my own build thread I'll put in all the design and calculation details while eagerly looking forward to some good peer review. Rip it up all the way. The more the better!



Tony

If the legend is giving values that are incorrect, you probably have incorrect results.  CFD is wonderful at providing completely correct answers to very incorrectly posed questions.  If you changed your boundary conditions that drastically and there was no real change, you probably have a poorly-posed boundary condition.  Usually I use a mass flow on the inlet and a pressure condition on the exit plane.  What boundary conditions are you using?  If the pressures are incorrect, the mass flow partitioning and flow-field will definitely be incorrect.

Ok, we are in agreement on using the 4X as a target minimum.  But to be clear to all the readers, when you are using "V^2/2g", it is referring to the incompressible form of the energy equation and should not be applied to anything over Mach = 0.3 due to a loss in accuracy.  And also to be correct, you need to multiply that by density to actually capture the dynamic head, rho*V^2/2g.  Since it is the incompressible form, the density is assumed constant everywhere and as such that is the primary source of the error.  So, this is ok for use in the flowfields around the combustor since the local Mach numbers are very low (sub M = 0.1)...but it's use in other areas of the engine are not advised.

Generally, it is about the same amount of effort to use the compressible equations and eliminate that source of error.

- Chris

[CH3NO2]

Hi Chris,

Yeah, something is wrong with the boundary conditions but I still need to figure out what it is.
The compressor diffuser outlet conditions are set to 2.75lbm/sec air @ 400'F, 55psia, with the flame can outlet open to atmosphere and no heat addition. The flame can outlet was also flowed to a hard vacuum but it essentially produced the same results.

I'll try another simulation where the flame can hole area is uniformly reduced until it produces a ~2.5 psid. We will see if the flow area results look reasonable. ...this may give a clue on what the problem is.

Can you import Solidworks files? I can email the model to you for a sanity check.

Thanks for all your help.
Tony

[finiteparts]

I think I can convert a SW file. In the CFD systems that I have used, I usually specify a total pressure condition on one side and a flow condition on the other side. If you are specifying the mass flow and accidentally have total pressures on both ends, you may be over constraining the problem. You are not confusing "reference pressure" with one of the inlet or outlet static/total pressure? (I am not even sure if SW uses a reference pressure???) Unfortuantely, I just don't know much about SW flow solver.

You might try to change your boundary conditions around to see how the simulation changes. Can you try to specify a total pressure on the inlet of 55 psia and the mass flow condition on the outlet? Or try to use the inlet mass flow condition and then apply an averaged static pressure on the exit?

- Chris

[CH3NO2]

Hi Chris,

Yeah we already tried working the flow conditions both ways (inlet total pressure with an outlet mass flow, and inlet mass flow and outlet pressure) and got the same results.
Jetspecs prescribes 13.68"^{^2} of actual hole area so this is what we started with.  This hole area is what produced the ~0.5 psid across the flame can walls.

Yesterday we reworked all the Lefebvre equations (independently) and found we needed an actual hole area of 10.9"^{^2}.   This is with a C_d of 0.57.

Then we plugged it all into Solidworks and ran it overnight @ high mesh.   This morning we found the results were ~0.3 psid. We decreased the hole area and the delta p went down. 

Obviously something's not right.   I can't be sure if its something I'm doing or if we're asking Solidworks to do more than it can handle.  It is a rather complicated flow system.  Solidworks is great for making drawings but it isn't really a dedicated CFD program.  We have modeled small scale examples of  flow through an orifice and it works perfectly... but once we try to flow the entire combustion assembly it gives questionable results.
I'll just consider myself lucky it can do this much.  Even a crude CFD is a lot better than nothing.  At least it can show when a recirculation bubble forms or doesn't form in the primary zone.   

Another useful aspect is what it shows for relative numbers of pressure and velocity.   Even if the absolutes are off, at least it gives some relative proportionality.  As an example, I have a crude elbow/tee diffuser I'm trying to model.   CFD says I am dropping 0.3 psi across the tee... but it also says I'm dropping 0.3 psi across the flame can holes.... So now I can get a feel for how much pressure is being dropped across the diffuser elbow/tee.
This is definitely helpful to know.

We'll baseline with the Lefebvre numbers and start building it from there.   This is my first turbine build so I expect to go through a big learning curve.

Tony

[CH3NO2]

Good news on the Solidworks. We are remodeling in CFD the various parts in much smaller pieces and are getting much more believable results.
For example, the diffuser elbow/tee, we put all the flow through just that segment only and found a 2.3 psi drop. This part is believable.
I dont like losing 2.3 psi in the tee, but that's the cost of having a really short diffuser.
We'll be sure to experiment more in this area.


So now we will remodel everything (in smaller segments) so as not to overload Solidworks.   We should be able to get a much more realistic answer.

We've also been working on the Lefebver equations a little more.
 



Keeping fingers crossed.

Tony

[racket]

Hi Tony

Very interesting seeing this done properly .

Cheers
John

[gleenooks]

Feb 19, 2017 22:23:42 GMT -5 finiteparts said:

My design values are thus far:

   Liner outside diameter = 5.0 inch
   Combustor casing inside diameter = 6.0 inch
   Compressor discharge temperature, T3 = 340 F
   Compressor discharge pressure, P3 = 47 psia
   Pressure drop = 5%
   Liner flow through effective area = 1.838 in^2
   Casing to liner annular passage flow speed = 24.1 ft/s




HOW did u find 5 inch & 6 inch?

[finiteparts]

I just picked the 6 inch diameter, because this is the casing size that I wanted to use. Lefebvre has an equation in his book for finding the reference area (which you can convert to a diameter) that requires the inlet Mach number, pressure loss ratio and the pressure loss factor to solve. These can be estimated, but for this turbo, a 6 inch case is maybe over sized. So then I sized the liner based on the outer annulus flow velocity and the combustor cold flow speed. I plan to cover Lefebvre's method to size a can combustor in the future when I get time and discuss this further.

- Chris

[finiteparts]

Tony,

I am unable to import the SW files. My guess is that they are a newer format than the software that I used was written to accept.

- Chris

[CH3NO2]

Right on Chris. Thanks for trying.
I think SW can probably work it out in segments, with 1/2 space and 1/4 space boundaries, decreasing or increasing mesh size in various areas to optimize focus as needed. We're working on it now.

This has been a great learning process.

Tony

[ripp]

Hello Chris,

Can you figure out why Combustion Disign is different?
Allison 250 / GTD350
Striking is that both combustion chamber have plenty of room after the Tertiary holes
That seems very important.


youtu.be/V10uHxJA2V0

www.users.zetnet.co.uk/gas/combust.htm



Thanks, Ralph

translate.google.com

[smithy1]

I have elements of the RR250 C20B chamber in my GT6041 powered go-kart...seems to work fine.

Cheers,
Smithy.