Falcon Heavy Cross-Feed

[go4mars]

Elon seems to imply that the fuel comes from the "adjacent engine". 

Would it be practical to have each octopus mouth feed 2 pipes on the side core? 

Multi-splitter like the octopus, but with less outputs:

Each pipe coming off this "Hexapus" is big enough to feed 2 engines.  So on the side boosters, 6 lines come off from each Hexapus.  Each of these lines has a splitter (like a Y-connector for your garden hose).  The 6 outermost engines on a booster are fed from 3 split lines.  The three engines on the side facing the middle core are fed in the same way, but (like the garden hose Y) there is a valve on the pipe going to a quick-disconnect system on the outer edge from the other split inthe Y's on this side.  These hook up to counterparts on the middle core to feed the "adjacent" engines. 

In the middle core, there is the traditional F9 style "octopus", but with valved branches facing the side cores on the feedlines for the 6 relevant engines. 

SpaceX had mentioned at one point that they hope to be able to completely fuel-up for a launch in less than 15 minutes IIRC.  Having these big fat valves available might help speed that up. 

Criticism welcome.     Thanks.

[RDoc]

Wouldn't it be simpler to just run a single line each RP1 and LOX from the outboard tanks to the core, then distribute the fuel/lox from there? That would only require two couplings rather than six on each side. Perhaps with valving, the current fill/empty lines could be expanded a bit and used?

Looking at the octopus (decapus?) jpg, and guessing the LOX tubes are about 8" in diameter, and a burn rate of about 4 cu ft/sec, I come up with a flow velocity of around 12 ft/sec. I would have guessed faster, but to keep the resistance low I suppose it's possible. Is that kind of velocity believable?

If so, the instabilities in switching from one feed to another wouldn't be that extreme as there would only be a few hundred lbs moving relatively slowly in the lines. Perhaps more importantly, I wouldn't think it very hard to model and test.

[jimvela]

If so, the instabilities in switching from one feed to another wouldn't be that extreme as there would only be a few hundred lbs moving relatively slowly in the lines. Perhaps more importantly, I wouldn't think it very hard to model and test.

Answer this: what happens to a cryogen in a feed line while the valve is closed and it sits there warm soaking?  When that valve is opened, what does the inlet side see from the valve?

[RDoc]

Answer this: what happens to a cryogen in a feed line while the valve is closed and it sits there warm soaking?  When that valve is opened, what does the inlet side see from the valve?

Well if I were designing it, I'd look into allowing a small flow through the valve to keep everything cold.

[Robotbeat]

Just a reminder of what the octopus looks (or looked) like without any legs. (from here)

Thank you for posting that.

[RDoc]

Answer this: what happens to a cryogen in a feed line while the valve is closed and it sits there warm soaking?  When that valve is opened, what does the inlet side see from the valve?

Actually, thinking about this a bit more, if the valve was a single Y valve (open before close) located right at or just astern of the octopus, I'm not sure it would be subjected to significant warm soaking.

First, the flow from the running tank would chill the entire valve, and secondly, convection back up into the octopus body would tend to carry away warmed LOX. The acceleration would help the convection effect and it wouldn't be far to a large reservoir of cold LOX.

Again, this would need modeling and testing, but it still doesn't sound very hard to engineer.

[sdsds]

I wouldn't think it very hard to model and test.

Is there any indication SpaceX has ever (on a test stand) switched propellant tanks feeding an engine during a burn?

[RDoc]

Is there any indication SpaceX has ever (on a test stand) switched propellant tanks feeding an engine during a burn?

I'm unaware of any cross/side feeding live fire tests at all. Have there been any?

In any case, wouldn't they try it on a test bench without an engine, etc. first? If my estimates of flow rates are anywhere near close, it doesn't sound very hard to do some modeling and then do a bench test to see how good the model is.

[MP99]

One option is to just pay a mass penalty, lengthen both feeds slightly and install large Ts and valves as Jim suggests.  I still don't think the center core feeds can be "off" while in flight because of problems with flow when we want to slam those valves all the way open at booster separation.  This is probably the easiest and best option.

I'm no expert by any means, but are these sort of valves the only ones that might be used?

ISTM that, if possible, it would make much more sense to gently shut down the side-feeds while (or after) opening up the feeds from the core. That would avoid shocks to the engine being fed.

Taken to an extreme, the RP-1 & O₂ feeds to each of the side-fed engines could be switched individually - twelve small (but slow) transients instead of a couple of big (but slow) ones.

cheers, Martin

[MP99]

If so, the instabilities in switching from one feed to another wouldn't be that extreme as there would only be a few hundred lbs moving relatively slowly in the lines. Perhaps more importantly, I wouldn't think it very hard to model and test.

Answer this: what happens to a cryogen in a feed line while the valve is closed and it sits there warm soaking?  When that valve is opened, what does the inlet side see from the valve?

Would a cryo valve have to be closed on the ground? Once closed in-flight, I don't think it would need to be opened again?

cheers, Martin

[RDoc]

Would a cryo valve have to be closed on the ground? Once closed in-flight, I don't think it would need to be opened again?

cheers, Martin

IFF some of the engines were to be switched from the side tanks to the core tanks (as opposed to being shut down) the LOX main tank feeds to them would start out shut on the ground, then opened in flight so there's the question of keeping them chilled.
 

[cuddihy]

Wouldn't it be simpler to just run a single line each RP1 and LOX from the outboard tanks to the core, then distribute the fuel/lox from there? That would only require two couplings rather than six on each side. Perhaps with valving, the current fill/empty lines could be expanded a bit and used?

Looking at the octopus (decapus?) jpg, and guessing the LOX tubes are about 8" in diameter, and a burn rate of about 4 cu ft/sec, I come up with a flow velocity of around 12 ft/sec. I would have guessed faster, but to keep the resistance low I suppose it's possible. Is that kind of velocity believable?

If so, the instabilities in switching from one feed to another wouldn't be that extreme as there would only be a few hundred lbs moving relatively slowly in the lines. Perhaps more importantly, I wouldn't think it very hard to model and test.

Simpler conceptually, but much more expensive. And this discussion has pretty much answered my question for me. From looking at the pictures above, it's evident that although it's not that big a change to go to the 12/3/12 configuration, because it looks like it can be done by splitting the individual smaller feed lines, and without major mods to the tank feedthrough, no added Kerosene plenum, and minor mods to the thrust structure. Also it allows you to keep a lot more commonality between the core and side feed systems and thrust structures.

On the other hand if you go to main feed line cross connects you now have to extend the feedlines, create a kerosene plenum to allow for smooth flow through the cross connects, extend the thrust structure to accomodate the big feed line cross connects, and you have major differences between the side booster feed areas (with two feed cross connects each) versus the four in the core. Add instability questions from big feed valves shutting and it's easy to see why they wouldn't do it for just another 5 or 8% boost in payload when it's already higher than the top of the market.

[simonbp]

I might have missed it in all this discussion, but if the center tank is 1/4 full at separation assuming everything throttles the same, couldn't you throttle down the "center three" before sep to lower their flow rate, allowing the center to be more than 3/4 full after sep?

[Robotbeat]

I might have missed it in all this discussion, but if the center tank is 1/4 full at separation assuming everything throttles the same, couldn't you throttle down the "center three" before sep to lower their flow rate, allowing the center to be more than 3/4 full after sep?

The other way...

It's normally supposed to be ~3/4 full after sep because the outer tanks effectively each feed three of the core's engines. Throttling would allow you to increase that to, say, ~7/8s full.

[simonbp]

Yes, that's what I meant. Which would conform better to Shotwell's statements at Space Access last year that the center would "nearly full" at staging.

[baldusi]

What if the booster's tanks were held at higher pressure than the core tank? And in the feeds to the outboard engines with a Y connection. On the core side you put a poppet valve from the core tank and on the Booster side you put a one-way valve. Thus, as long as you have the booster feeding at higher pressure, it would keep the poppet shut by pressure differential and keep the one way open, thus feeding from the side. Before staging you simply lower the pressure on the boosters side, thus, the popper valve would open and the one way would shut.
If you keep the poppets on the octopus, for example, there would be no sublimation problems. But the one way might generate some cavitation.
May be they can design sort of a 5 way valve, with the actuation activated by pressure differential. In fact, you could handle that with the pressurizations system directly.

[kevin-rf]

What if the booster's tanks were held at higher pressure than the core tank? And in the feeds to the outboard engines with a Y connection. On the core side you put a poppet valve from the core tank and on the Booster side you put a one-way valve. Thus, as long as you have the booster feeding at higher pressure, it would keep the poppet shut by pressure differential and keep the one way open, thus feeding from the side. Before staging you simply lower the pressure on the boosters side, thus, the popper valve would open and the one way would shut.
If you keep the poppets on the octopus, for example, there would be no sublimation problems. But the one way might generate some cavitation.
May be they can design sort of a 5 way valve, with the actuation activated by pressure differential. In fact, you could handle that with the pressurizations system directly.

What is the head pressure on 180 feet of LOX at 4g? That is what you would need to increase the pressure in the outriggers by. Meaning beefed up heaver less mass efficient structures.

And how will the turbo pump react when the pressure suddenly drops and the center tanks begins feeding?

[baldusi]

What is the head pressure on 180 feet of LOX at 4g? That is what you would need to increase the pressure in the outriggers by. Meaning beefed up heaver less mass efficient structures.

I understand that the old tanks where pressurized at 50psi (3.4kg/cm²) for flight stress support. And LOX is 1.141g/cm³. At 54.8m of column that's 7.73kg/cm² at Earth's gravity. Plus the 4g, that's 30.92 kg/cm²+7.73kg/cm²+3.4kg/cm²=42.03kg/cm=617psi at the head, right?
You'll have to help me here, but isn't that the same pressure the Boosters have to support early in the flight?
Another issue is that poppet valves can be balanced easily. You don't need higher pressure on one side. You need higher than a threshold pressure on such side. This is a basic exercise of surface differentials. In fact, that's where it sort of becomes a 3/4/5 way valve.

And how will the turbo pump react when the pressure suddenly drops and the center tanks begins feeding?

The pressure drop of the boosters tanks wouldn't be affected. It would be falling either way. What you get when you open the poppet is a pressure spike. Beside the problem of the initial hammer punch (that's how you call it?), which can be solved with the poppet's opening speed, you'd get back to the situation when the boosters where at whatever the core is at staging (I guess around 80%?)

[RDoc]

Wouldn't it be simpler to just run a single line each RP1 and LOX from the outboard tanks to the core, then distribute the fuel/lox from there? That would only require two couplings rather than six on each side. Perhaps with valving, the current fill/empty lines could be expanded a bit and used?

Simpler conceptually, but much more expensive.

That certainly doesn't seem obvious.

Assuming the fill/empty lines can be enlarged and are used for the cross feed to 3 engines in the core, then single cross feed for fuel and LOX requires for each booster:

Probably enlarging the 2 fill/empty lines
2 T's
2 line separation system
2 1x3 manifolds in the core
Space and support for 2 lines

Running separate lines using the booster feed lines requires

Probably enlarging 6 feed lines
6 1x2 manifolds
6 line separation system
Space and support for 6 lines

[oldAtlas_Eguy]

The technical problems that crossfeed design must overcome as outlined so far in this thread: (If I have missed one or more please indicate what they are?)

1) Difference in pressure between the booster and center cores on the operation of the crossfeed engine during changeover. Can be tested on the ground with a single engine, although a problem, solutions can be thoroughly tested on the ground with modifications to the M1D single engine test chamber. A secondary feed with a pressure reducer can easily simulate this condition.

2) Booster separation while center core engines still firing. Atlas heritage propellant quick disconnect would be a starting point. But having six propellant lines coming from the booster would be a higher risk than a larger two prop lines setup during booster separation. Some testing can be done on the ground but the real test cannot be simulated on the ground, it takes an actual flight to prove out the design, too many small variables involved with separation. Test jigs can reduce the risk but not resolve all risks.

3) Testing cross effects of propellant flow changes on the 9 engines of the center core. This can mostly be tested on the ground but at only 1 g and not the g level seen in flight.

4) The boosters will have a different “octopus” propellant feed structure than the center core (the center core would have a normal F9 “octupus”).  Simulation of propellant feed effects for the booster can be done only at 1g on the ground and not at the flight g levels.

A summary list of the technical challenges and whether ground testing can thoroughly test or can only mostly be tested except for flight conditions such as g levels, should be a task of this thread so that SpaceX test facilities can be evaluated for whether what technical challenges posed by the crossfeed as proposed can possibly be addressed by these existing or planned facilities like the new tri core hot fire facility at McGregor.