SpaceX Starship (BFS) Engineering General Thread 2

[DaveJes1979]

Complexity of manufacturing rocket engines Vs electric cars is missing the point of the of Mueller's anecdote.  It was about driving down cost of that complexity.

Here are the actual questions posed, which I answered:

"And he said, “How much does that car weigh?” And I said, “About 5 thousand pounds.” And how much does a Merlin engine weigh? I go, “About a thousand pounds?” So, he’s like, “So why the heck does it cost, you know, some fraction of a million dollars to make a Merlin engine?”

And I mean, he has a good point. And the material you’re using isn’t aluminum, it’s not stamped, so I’ll give you a factor of five. So it’s equivalent to a five-thousand-pound rocket engine. So why’s it 20 times the cost?"

[Cheapchips]

Complexity of manufacturing rocket engines Vs electric cars is missing the point of the of Mueller's anecdote.  It was about driving down cost of that complexity.

Here are the actual questions posed, which I answered:

"And he said, “How much does that car weigh?” And I said, “About 5 thousand pounds.” And how much does a Merlin engine weigh? I go, “About a thousand pounds?” So, he’s like, “So why the heck does it cost, you know, some fraction of a million dollars to make a Merlin engine?”

And I mean, he has a good point. And the material you’re using isn’t aluminum, it’s not stamped, so I’ll give you a factor of five. So it’s equivalent to a five-thousand-pound rocket engine. So why’s it 20 times the cost?"

They weren't the point of the anecdote.  The point was the paragraph you omitted:

So that’s the way we look at it and the way we think at SpaceX trying to get the amortization cost of the rocket down. Once you start reusing it, the real big cost becomes the amortization cost of the rocket, the operational costs, and the fuel costs, which is basically the same model as the airliners.

I wish I'd just included the Merlin price line I'd bolded now! I just thought people might find the whole statement interesting if they'd not seen it. 

[john smith 19]

So the question is how big a tank do you need to cryo form a 9m tank or can you build a sheet metal stretching machine that can operate at LN2 temperatures?

I wonder if filling the tanks with liquid nitrogen and pressurizing will do it.

That might work. Second paper I found said they had to be stretched on the order of 10 to 20 percent. I don't know how you could do this without a jig to minimize distortion.

John

You need to keep in mind the scale of this thing. It's 9m in dia but a full tank is what, about 40m long?

I think the thermal gradients between the inside and the outside (unless it's immersed in a tank, or pit) will ruin the effect.

What has changed since these experiments is the massively improved ability to instrument a test rig and the ability of FAE software to cope with that data to predict what will happen, rather than build the thing as well as possible, then build a second one based on how far out your  engineering experience was.

[john smith 19]

The other paper I found got good notch sensitivity using low silicon low carbon 3XX SS. See graph and text a few pages back.

John

Carbon, Silicon and Manganese were the elements that were thought to make cryo notch sensitivity worse for cryo formed steel. So lowering 2 of 3 of them should definitely have improved things.

[john smith 19]

1. Not difficult for joints- whether fastened, welded, or bonded, to be far stronger than constituent parts. Challenge is to make them light.

And that's a pretty big challenge. Hence the plan to wind the CFRP tanks in one piece and the interest in either eliminating stiffeners (pressure stabilizing) or making them in one piece, first as "Waffle grid" then later as Isogrid.

2.  Fuselage structure will not be the high risk item. I wonder about the servicing requirements after reentry, as far as maintaining the finish, but otherwise this will not be what sinks Starship.

Although the Russians built a number of Supersonic aircraft in steel in the 50's and 60's nothing has been on this scale or operated over this speed range. It's that speed range (and it's associated environmental range) that make this application tough with any material system.

3. Active cooling system and the actuated fins and canards will be where the risk and complication comes in. The first time we see a Starship lost, put money down with your Las Vegas bookie on this being the cause.

Those would seem to be some of the highest risk parts of the design (along with engine restarts after the trip to Mars).

That said the control surface on the Shuttle, and their actuator systems, never failed in 135 actual missions, including on orbit thermal cycling for up to 14 days. That's relevant because this design is also going to be violently unstable without continuous computer control. What's different is (AFAIK) the actuator forces are much bigger. An interesting optimization from the RASV programme was the Shuttle APU's were sized by the TVC needs of the engines. The power needs for the control surfaces alone were much smaller. That means the APUs would be operating much closer to their full power, rather than barely idling (assuming the engines will be locked in place and only a few will be moving during terminal thrusting).

Which leaves the active TPS. The only things close to this would seem to have been efforts to mount radiators on the skins of a few fighters in WWII and the 50's, (the idea that the Howard Hughes steam powered car project used them seems to be a myth. Based on this item from CalTech http://calteches.library.caltech.edu/2907/1/howard.pdf )

In those designs AFAIK the radiators were in the upper surface of the wings to minimize damage and maximize heat transfer outward.

So actively skin is basically a complete unknown. It was researched by NASA in the late 70's and early 80's as part of the goal to build a High Speed Research Aircraft (I think the name changed over time as it's sometimes called the Hypersonic RA given it's design goal was M5+, ground launched).

NASA also looked at the idea of discontinuous systems using (effectively) a "bucked brigade" of heat pipes to transfer the heat to IIRC the LH2 tank (which was viewed as the most likely fuel for this vehicle).

Heat pipes would eliminate a pipe puncture releasing all the coolant in the system, but at a potential weight penalty. Interestingly active cooling looked to be a way to make a M5 from Aluminum.

The other obvious issues with Methane cooling is what do you do with the hot(ish) Methane? Driving thrusters or the APUs are the obvious uses. 

I also note that people are assuming Methane cannot polymerize in the way RP1 can. I'd check if any of the alloy constituents they make the cooling plumbing out of can catalyze Methane polymerization and what temperature they do so.

The petrochemical industry is built on cooking chemicals to either split them into shorter ones or merge them into longer ones. They know a lot about how to get polymerize Methane.
All of which can (should) be tested on the ground before  construction begins.

[rashomon]

The last US aerospace project that had as much stainless structure as the Starship and was actually built and tested was the Rockwell XB-70. That was a more than half a century ago; a lot of its structural analysis was probably by hand calcs, or a very fat deck of punch cards running some FORTRAN code on a mainframe that couldn’t come close to out-calculating an Apple Watch SOC. The tools today are what make the difference, even down to the ability to simulate the effects of alloy composition. Its going to be fun to see what SpaceX comes up with.

[livingjw]

So the question is how big a tank do you need to cryo form a 9m tank or can you build a sheet metal stretching machine that can operate at LN2 temperatures?

I wonder if filling the tanks with liquid nitrogen and pressurizing will do it.

That might work. Second paper I found said they had to be stretched on the order of 10 to 20 percent. I don't know how you could do this without a jig to minimize distortion.

John

.... I think the thermal gradients between the inside and the outside (unless it's immersed in a tank, or pit) will ruin the effect. ....

The wall thickness is 1-2 mm. A small amount of insulation on the outside will guarantee near equal temperatures across the wall with the tank filled with LN2.

John

[MikeAtkinson]

The wall thickness is 1-2 mm. A small amount of insulation on the outside will guarantee near equal temperatures across the wall with the tank filled with LN2.

John

Would they expand one tank at a time? Or both together? Common bulkhead would not be expanded if both together, missmatch between expansion at tank boundary if one at a time.

[RobLynn]

So actively skin is basically a complete unknown. It was researched by NASA in the late 70's and early 80's as part of the goal to build a High Speed Research Aircraft (I think the name changed over time as it's sometimes called the Hypersonic RA given it's design goal was M5+, ground launched).

The actively cooled skin is new to reusable launch vehicles but is a really quite mundane application of normal heat exchanger technologies, it is mostly only a challenge due to the large amount of detailing in the coolant manifolds.  Integrating heat transfer passages into metal skins has been done in vast numbers of applications for most of the 20th century with few surprises.  The fluids can be relatively low pressure, and the flow passages can have close-packed parallel redundancy to reduce sensitivity to single point failures.  And the cooling flows can be made very conservative to start with with to be only incremental optimised (reduced) as flight test data accumulates.  The loads on the tanks and structure are very low during re-entry - the tanks can be pressurised to just above the dynamic (stagnation) pressure of reentry (a few kPa).  And there is potential for a lot of performance enhancement through eventual use of LH2 instead of methane.

[RedLineTrain]

So actively skin is basically a complete unknown. It was researched by NASA in the late 70's and early 80's as part of the goal to build a High Speed Research Aircraft (I think the name changed over time as it's sometimes called the Hypersonic RA given it's design goal was M5+, ground launched).

The actively cooled skin is new to reusable launch vehicles but is a really quite mundane application of normal heat exchanger technologies

Indeed, including in all of SpaceX's flame trenches in McGregor.

https://www.teslarati.com/wp-content/uploads/2018/04/Merlin-stands-crop-Aero-Photo.jpg

[livingjw]

The wall thickness is 1-2 mm. A small amount of insulation on the outside will guarantee near equal temperatures across the wall with the tank filled with LN2.

John

Would they expand one tank at a time? Or both together? Common bulkhead would not be expanded if both together, missmatch between expansion at tank boundary if one at a time.

You would have to do something like that, if that is the approach at all. Also, what do you do about the nose section and aft section beyond the tanks? I don't know. To many options to nail anything down. We live in interesting times.

John

[RobLynn]

I think the Raptors will be cheaper to make than $4m unless their super alloys are extremely expensive. Merlin's are less than a million.

Tom Mueller talking about cost conversations with Musk:

Like, here’s a conversation I had maybe about five years ago on the Merlin 1D when we first developed it. He asked me; he said, “How much do you think it costs to make a Model S?” And I’m like “I don’t know; 50 thousand dollars?” He said “No, about 30 thousand dollars.” That’s the marginal cost for that car.

And he said, “How much does that car weigh?” And I said, “About 5 thousand pounds.” And how much does a Merlin engine weigh? I go, “About a thousand pounds?” So, he’s like, “So why the heck does it cost, you know, some fraction of a million dollars to make a Merlin engine?”

And I mean, he has a good point. And the material you’re using isn’t aluminum, it’s not stamped, so I’ll give you a factor of five. So it’s equivalent to a five-thousand-pound rocket engine. So why’s it 20 times the cost? So that’s the way we look at it and the way we think at SpaceX trying to get the amortization cost of the rocket down. Once you start reusing it, the real big cost becomes the amortization cost of the rocket, the operational costs, and the fuel costs, which is basically the same model as the airliners.

So can we take that as a Model S being 30k and a Merlin being 20 x 30k = 600k?  (Assuming that 20x number wasn't just picked out of the air?)

If that is the case then marginal cost of Raptors will certainly be a lot less than my $4 million guess.  They are likely designed to make extensive use of laser sintering (~$1-2000/kg) with minimum component count and minimum heavy flange joints.  In which case a 1000kg engine (assuming similar Trust to weight as Merlin) might be as little as $1-2million dollars.

If they can make the (likely) severely life limited high pressure thrust chamber + throat and first part of nozzle a replaceable component that costs only a few $100k to swap out then the rest of the engine might be OK for 100's or 1000's of engine firings.  That would drop the depreciation per firing a lot.  The depreciation per flight for Starship and Superheavy could then be greatly reduced compared to an engine with life set by a single component that cannot be made durable.

[RobLynn]

The wall thickness is 1-2 mm. A small amount of insulation on the outside will guarantee near equal temperatures across the wall with the tank filled with LN2.

John

Would they expand one tank at a time? Or both together? Common bulkhead would not be expanded if both together, missmatch between expansion at tank boundary if one at a time.

It's a bit tricky isn't it?

Another option would be to build in a second bulkhead dome right next to the 1st to create a secondary full diameter tank-between the LOX and CH4 tanks,  Pressurise the two end tanks to 3bar and the middle tank to 6 bar and then cut out the unwanted bulkhead dome afterwards.

I think it is most likely they will form the two tanks separately then weld them together and run another cycle of cryoforming on the whole tank assembly to strain the weld joint.

[DaveJes1979]

Although the Russians built a number of Supersonic aircraft in steel in the 50's and 60's nothing has been on this scale or operated over this speed range. It's that speed range (and it's associated environmental range) that make this application tough with any material system.

Thermal, inertial, and aerodynamic loads are well-quantified.  Once SpaceX completes coupon testing on their proprietary alloy there won't be any question marks.  Apparently, there will be some incremental improvements in mechanical properties vs. other stainless steel alloys.

That said the control surface on the Shuttle, and their actuator systems, never failed in 135 actual missions, including on orbit thermal cycling for up to 14 days.

I didn't even have power failure modes in mind.  Especially if they go with electric servos (a difficult but not crazy suggestion, and would increase reliability) and batteries.  Hydraulics and APU's should be eliminated if at all possible.

Moving parts have many failure modes.  See CRS-16 grid fins.  Not a power problem.

Also, the fins and canards will need TPS.  So some expense and complication with carbon-carbon, carbon-ceramic or whatever.

Which leaves the active TPS.

There won't be any real surprises with this.  It will work fine, when everything works nominally.

The problem will be when that is not the case, failures with pumps (top of the list), valves, and complex plumbing.

[DaveJes1979]

Two more comments, in response to some of the discussion:

1.  People keep assuming that additive manufacturing is a silver bullet.  While it can work fine for something small like a Super Draco, which is a very simple pressure-fed rocket, that doesn't mean it is going to scale up to a large, staged-combustion engine like Raptor, full of complex turbomachinery.  The size isn't the only problem, the mechanical properties of printed metals is eye-wateringly miserable.  That translates into a large weight penalty.

2.  There are no question marks, Starship will be structurally optimized by FEA software.  In our present year, 2019, software structural optimization tools are well-established.  Especially since we are now talking about isotropic materials (metals), it is a walk in the park.

[RobLynn]

Two more comments, in response to some of the discussion:

1.  People keep assuming that additive manufacturing is a silver bullet.  While it can work fine for something small like a Super Draco, which is a very simple pressure-fed rocket, that doesn't mean it is going to scale up to a large, staged-combustion engine like Raptor, full of complex turbomachinery.  The size isn't the only problem, the mechanical properties of printed metals is eye-wateringly miserable.  That translates into a large weight penalty.

I've working on biggish (0.2m scale) laser sintered metal parts.  The mechanical properties are generally pretty close to the ideal annealed metal strength, and mostly heat-treatable alloys are used to allow even higher strengths to be attained.
 You can use titanium alloys, maraging steels and wrought superalloys all with extremely high strengths, and you can light-weight designs using small webs etc in ways that are simply not manufacturable by any other means. 
EG check out properties of wrought superalloy Inconel 718 (which I have used a lot): EOS 718 properties: https://www.eos.info/eos-nickelalloy-in718-nickel-alloy-for-aerospace-and-industry-d64f5c43276f3673

I know people working on much bigger laser sintered parts (0.5m scale), no problem with strength and with careful tweaking of build process no problem with distortion compared to competing processes like casting or forging.

Sure it is not suitable for every component and not suitable for every material (particularly heat treatable aluminum alloys), but is enabling a big step forward in weight and cost reduction for low volume aerospace manufacture, particularly for hard to machine and hard to join materials.  If they can use it to make the Raptor turbomachinery and preburner casings and even injector manifolds with all their large numbers of intricately shaped internal passages in only a few pieces then they will likely be able to save cost, manpower and maybe weight compared to more conventional casting and forging fabrication.

[MizaruSpaceXNut]

Well, I saw a video where something fell into the rocket part with legs, perhaps a piece of lumber! I hope no

one was below or something breakable? They were moving a chunk of bulkhead with the crane when

something dislodged!

[Slarty1080]

Two more comments, in response to some of the discussion:

1.  People keep assuming that additive manufacturing is a silver bullet.  While it can work fine for something small like a Super Draco, which is a very simple pressure-fed rocket, that doesn't mean it is going to scale up to a large, staged-combustion engine like Raptor, full of complex turbomachinery.  The size isn't the only problem, the mechanical properties of printed metals is eye-wateringly miserable.  That translates into a large weight penalty.

2.  There are no question marks, Starship will be structurally optimized by FEA software.  In our present year, 2019, software structural optimization tools are well-established.  Especially since we are now talking about isotropic materials (metals), it is a walk in the park.

I agree FEA software will be used on Starship a lot. The only questions will be around how good the FEA software is when it comes to hypersonic or orbital entry speeds. I'm not sure that this area is fully characterised in sufficient detail. That said no doubt SpaceX will use their early tests to improve the FEA software and by the time the prototypes are done any such issues will be resolved.

[Rei]

Complexity of manufacturing rocket engines Vs electric cars is missing the point of the of Mueller's anecdote.  It was about driving down cost of that complexity.

Here are the actual questions posed, which I answered:

"And he said, “How much does that car weigh?” And I said, “About 5 thousand pounds.” And how much does a Merlin engine weigh? I go, “About a thousand pounds?” So, he’s like, “So why the heck does it cost, you know, some fraction of a million dollars to make a Merlin engine?”

And I mean, he has a good point. And the material you’re using isn’t aluminum, it’s not stamped, so I’ll give you a factor of five. So it’s equivalent to a five-thousand-pound rocket engine. So why’s it 20 times the cost?"

It reminds me of when I traveled to the US with my then-fiance, who was an former auto mechanic, and took him to JSC. It was his first chance to see rocket engines up close, and his reaction was.... "That's... it?  But I mean.... I could build most of that myself!" He expected some sort of magical-level of technological wizardry, and was shocked to see how "familiar" everything was.  I was worried I was going to have to dissuade him from trying to build a liquid-fueled rocket of his own when we got home 

Yes, liquid-fueled rocket engines use expensive and sometimes difficult to work with alloys.  Yes, some parts are made with unusual manufacturing techniques or have unusually high precision requirements. But ultimately when it comes down to it, they're manufactured products, and there's nothing stopping them from going into bulk production except demand.

[Rei]

I agree FEA software will be used on Starship a lot. The only questions will be around how good the FEA software is when it comes to hypersonic or orbital entry speeds. I'm not sure that this area is fully characterised in sufficient detail. That said no doubt SpaceX will use their early tests to improve the FEA software and by the time the prototypes are done any such issues will be resolved.

I think you're mixing up FEA and CFD. FEA deals with the transfer of forces imparted to solid objects by the world around them. CFD tells you what those forces will be.  FEA doesn't care whether you're sitting on the ground or entering the atmosphere at Mars, the physics is the same.  It's CFD that varies as you travel through different atmospheric / exospheric** environments.

** Technically an exosphere isn't a "fluid", but you'd still model it in CFD... it becomes a fluid when compressed at your bow shock