SpaceX vs ULA : What's the deal with engine performance?

Factual Questions
1

Ok, I read thisarticle. Among other claims was that the RD-180 engine is needed to get “high energy” payloads into the high orbits needed by certain types of satellites.

The author also makes fun of SpaceX’s engine sizes, implying that a rocket with 27 engines is automatically less reliable than one with just 3-6.

I took a look at the 2 engines in question using wikipedia :

There’s the “superior” russian engine, here.
Relevant stats : Sea level ISP 311, Vacuum ISP 338, 78.4 TWR.
Then there’s SpaceX’s cruder engine, here.
Sea level ISP 282, Vacuum ISP 311, 150 TWR.

You also need 5.5 Merlins to equal one RD-180. This would possibly be an issue if the merlins weren’t drastically cheaper. SpaceX is quoting total launch prices at 1/3 of so of what ULA is quoting.

Now, as I understand the basic mechanics of rocket flight, you only need full engine thrust right at liftoff. Shortly after liftoff, you have to throttle back in order to reduce aerodynamic drag, and later in flight, you have enough speed that you’ll make orbit even with less thrust (albeit this will cost you some delta-V that the mission may have needed).

This makes a design with more engines more reliable, because if one engine fails you can still make orbit, so long as the failure does not occur in that brief period of time right at liftoff.

Anyways, what’s this guy talking about? A quick bit of figuring with a delta-V calculator show’s a difference of ~458 m/s assuming 80% of the lower stage is fuel. At 70%, it’s 342 m/s. However, the better TWR of the Merlins negates some of this advantage, though I’d need more precise numbers of the dry mass of the different stages. There’s also a benefit from launching from Brownsville rather than Cape Canaveral.

Is the problem actually with the upper stages, where ULA can use liquid hydrogen?

2

SpaceX has so far put two satellites into geosynchronous orbit (SES-8 and Thaicom 6) so there obviously isn’t a problem with them reaching high orbit.

3

That’s basically it. “High-energy” means LH2 fuel in this context.

Making some simplifying assumptions, if you have two rockets that have the same payload to LEO, the one with LH2 engines on the upper stage will have a higher payload to GTO. The reason is basically that as your payload fraction goes down, the energy density of the stage becomes more relevant.

This is easy to see in the limiting case: if your payload weighs 1000 kg and your stage 10 kg, then halving your stage mass to 5 kg makes almost no difference. If instead the stage weighs 1000 kg and the payload 10 kg, then halving the stage mass to 500 kg makes an enormous difference. Obviously things aren’t so extreme in reality, but the same basic principle applies.

At any rate, I found the article rather silly. It focuses on irrelevant design details. Someone looking to launch a satellite only cares about payload, cost, and reliability (well, and bureaucracy). SpaceX has already said that they can’t compete on every single launch, in part to the upper stage and in part from the lack of Falcon Heavy; they just want the ability to compete on launches they can handle.

4

There are several misapprehensions in the o.p which lead to inappropriate apples to oranges comparison of performance characteristics. The first is that thrust-to-weight ratio (TWR) is a critical parameter in assessing vehicle-level performance. A TWR is a good metric for comparison of thrust performance at the engine level, but the engine is actually small part of the dry mass of the vehicle, and even less than the total gross lift off weight (GLOW) of the loaded vehicle, so the weight of the engine(s) by themselves is not strictly germane. As the o.p. notes, maximum thrust is typically only used at liftoff, and once the vehicle is moving at an appriciable speed the engines are throttled down so as to reduce aerodynamic loading and flight vibration environments at the point of highest dynamic pressure (max-Q alpha). Then engines are then typically throttled back up once the atmosphere is thinned out to reduce gravity drag losses but there may be additional restrictions on the acceleration thrust load that the payload can tolerate. So, high thrust is nice to get moving (hence, why solid propellant boosters are used as “Stage 0” strap-ons desipte their relatively poor mass-specific performance), but isn’t a critical measure provided that some threshold level of thrust over unity can be delivered. Since most of that thrust is used to lift propellant that will be used later in flight, a lower TWR with a higher specific impulse (the normalized measure of propellant mass per unit thrust) may give better overall performance in terms of payload to a given orbit, which is exactly what the RD-180 with its 3800 psi chamber pressure and sea level I[SUB]sp[/SUB] of 311 s offers versus the Merlin-1D with a chamber pressure of ~1400 psi and sea level I[SUB]sp[/SUB] of 282 s.

Second, the Atlas V uses one NPO Energomash RD-180 engine, which is a dual combustion chamber, dual nozzle engine (perhaps giving the impression that it is multiple engines). The RD-180 is derived from the four chamber RD-170 engine used on the Energia launch vehicle and the Zenit rocket (in modified form). Given the incredibly high performance from this engine and its extended and problematic development it has performed with remarkable reliability with only one propulsive failure (debris in turbopump). In static fire testing the engine has endured up to twenty full duration static fire tests without measureable degradation. It is basically considered the crown jewel of the Russian rocket propulsion system development and a version of it (RD-190 and -191) is intended to be used on the Angara system, the Russian replacement for the existing Proton vehicle expected for operation through 2065. BTW, Pratt & Whitney attempted to build this engine under license from Energomash but was unable to build it to the required quality standards and repeatedly failed qualification testing. Unfortunately, that also leaves EELV dependant upon a foreign-sourced engine to power its RP-1/LOX launch vehicle. (The Delta IV is all cryogenic with the attendant problems that come along with that.)

Third, when we talk of reliability, it has to be looked at from a system-level context. Yes, if one or even possibly two of the Merlin-1D engines on the Falcon 9v1.1 experienced a non-catastrophic propulsive failure which allows controlled shutdown, the vehicle may still be able to delivery the primary payloads to orbit (depending on payload mass and orbital parameters), but despite efforts to isolate and protect the engines from one another, there are still failure modes which could result in multiple engine failures or loss of control of vehicle, and of course all of the additional plumbing, controls, mounting hardware, filters, turbopumps, et cetera associated with each engine. For instance, a water hammer event that feeds back into the propellant manifolds, or loss of thrust vector control system, or any of a number of failures which are not simple shutdowns could result in a loss of vehicle criticality for which there is no recovery. Propulsive failure or even a significant thrust imbalance in a side core of a triple core system (Falcon Heavy or Delta IV-H) would probably result in adverse bending modes or structural imbalance resulting in loss of guidance and control. All things being equal, nine engines increases the potential for failure exponentially to order of the number of engines, e.g. a Falcon 9 with n=9 engines will have a composite probability of failure of P[SUB]f[/SUB]=1-R[SUP]n[/SUP], where R is the per-engine reliability. So, for a R=0.997 (the nominal “three sigma” level of reliability), a vehicle with a single engine (n=1) such as the Atlas V or Delta IV will have a P[SUB]f[/SUB]=0.3% for propulsive failure, while the Falcon 9 with n=9 will have a P=2.7% chance of propulsive failure; nearly an order of magnitude greater. On the same basis, a triple core single engine vehicle like the Delta IV-H with n=3 will have a P[SUB]f[/SUB]=0.9%, while the Falcon Heavy with n=27 is P[SUB]f[/SUB]~8%. (All of these numbers are for illustration only, and these are not the predicted reliability of the Merlin-1D engine and Falcon 9v1.1/Heavy propellant feed systems.)

Of course, we can’t just assume that reliability of all engines and propellant feed systems is the same; each has to be considered in the context of the system in terms of design robustness, redundancy, complexity, and ultimately demonstrated reliability of the overall system which cannot be assessed except by a significant body of flight history. It may be that the Falcon 9v1.1 and Falcon Heavy are as or more more reliable than the Atlas V (which has a nearly spotless record to date) but we do not have enough data to make that assessment, and handwaving that having multiple engines automagically offers additional redundancy when that is actually only true in some portion of some trajectories for some types of failure is glossing over the additional complexity having so many engines requires.

Fourth, there is a dramatic performance benefit to using all cryogenic propellants on an upper stage vehicle as the propellant mass carried in the second stage is close to a 1:1 trade for payload mass for a two stage vehicle, hence why Atlas V and Delta IV cryogenic upper stages. However, this comes with all the attendant problems of LH2, e.g. propellant expansion, thermal stresses, leakage, embrittlement, high dynamic slosh, low propellant density, et cetera. SpaceX decided to trade the performance of LH2/LOX for the greater simplicity and higher volumetric energy density of RP-1/LOX propellants. Of course, this also adds some other issues that aren’t typically experienced with LH2/LOX engines, such as fuel gelling, manual purging of lines, payload contamination, et cetera, so there are tradeoffs beyond performance to be considered.

As for launch costs, I don’t know that we can really make a head-to-head comparison. On one hand, it is pretty clear that SpaceX is still eating a lot of costs as part of their development efforts, and what they are currently charging customers for launches does not reflect their actual costs including unplanned labor, design and process modifications, et cetera. On the other hand, the costs of the ULA vehicles are ridiculously higher than the original cost targets for the EELV and are effectively subsidized by the Air Force bearing the cost to maintain launch facilities and other sundries to support the program, whereas SpaceX currently maintains their own launch facilities (albeit with sweetheart lease deals). If SpaceX can launch for even half the cost per payload kilogram to orbit that ULA can, it will still be a significant improvement over current costs, and would make SpaceX a viable and badly needed competitor to ULA.

The real question is whether SpaceX can build up enough of a commercial customer base to justify the F9v1.1 core vehicle. Right now it has way more capability than nearly any commercial satellite operator needs for a single payload; in order to keep costs per payload mass reasonable, SpaceX will have to demonstrate reliable multi-payload deployment capability, which is a highly challenging endeavor that ULA has mostly avoided with their vehicles. If they can do that, and hold launch costs to a reasonable level, and demonstrate good (~98%) reliability, and get commercial providers to buy into building satellites that can be integrated horizontally into a multiple payload stack, then there is a good chance that they can foster nascent elements of the commercial spacecraft industry which have been waiting for significant reductions in launch costs. This would be a good thing for the entire aerospace industry, which is largely dependent upon government (primarily Department of Defense) expendatures to sustain space access, and would potentially open up multi-billion dollar sectors of new space applications in Earth surveillance, satellite communications, space resource utilization, et cetera. But despite the marketing glossies from the SpaceX PR department, this is still far from certain. Space launch and access to space is still a very challenging problem–as SpaceX themselves have discovered–and a rocket launch vehicle is always a fraction of a second from literally going sideways and burning up hundreds of millions of dollars of payload and hundres of thousands of hours of person-effort from liftoff all the way up to orbit.

Stranger

5

Your caveats are well-noted, but I wanted to go through the math a little more here.

First, although the reliability decrease is indeed exponential, that may give the wrong impression due to the nature of exponentials for numbers close to 1. To a good approximation, the decrease is in fact linear. 0.997^9 =~ 0.97332, while (1 - 0.003*9) = 0.973. The difference is in the noise.

SpaceX has said they can handle a single-engine failure from seconds after launch (this ignores non-isolated failures, of course). The odds of 8 or 9 engines succeeding, given a 0.003 failure rate, is:
(1 - 0.003)^9 + 9*(1 - 0.003)^80.003 = 0.97332 + 90.00293 = 0.99968

At 0.03%, the overall failure rate is better by an order of magnitude than the single-engine rate.

Again, clearly I’m ignoring non-isolated failures and the other issues you mentioned; I just wanted to show how much redundancy helps them in one class of failures.

6

They didn’t make a rocket with 27 Merlins.

SpaceX are working on the larger engines for the heavy lifter rocket, the Raptor project. The Raptor methane/LOX engine will use the more efficient staged combustion cycle … so there’s two reasons (fuel and process) you don’t simply scale up Falcon stats to the Raptor.

7

The Falcon 9 can only recover from a single engine shutdown early in flight in a significantly underloaded condition. The loss of performance at low speed is a significant hit. Closer to the end of Stage 1 operation, they can shut down one or even possibly two engines and still get nearly the same delivered impulse by burning slightly longer. And as you note, this is only for non-catastrophic failures for which the engine can be shut down in a controlled fashion. The greater complexity of the propellant feed system, avionics and controls, structures, et cetera which goes to provisions for all of those engines magnifies the potential for problems from component failures to errors in assembly. A long standing lesson from the annals of launch vehicle history is that excessive complexity is always asking for trouble, and the value of redundancy over robustness and simplicity is overrated in terms of reliability. The most reliable rocket launch vehicles in history–Soyuz, Delta II, Minuteman, Peacekeeper–with sufficient history to have a predicted reliability of 97% or greater have relied largely on simplicity and robustness over redundancy. The one exception to this was the US Space Transportation System (Space Shuttle), which was the most expensive launch vehicle in history and required a massive army of quality control and support personnel to keep it operating in what was, in retrospect, obviously marginal condition.

Falcon Heavy is the world’s most powerful rocket, a launch vehicle of scale and capability unequaled by any other currently flying…Its first stage **is composed of three Falcon 9 nine engine cores whose 27 Merlin engines **together generate nearly 4 million pounds of thrust at liftoff.

The Raptor engine is designed for use on the so-called “Mars Colonial Transporter”, basically an upgraded version of the Falcon X and Falcon XX that was presented back in 2009. The Raptor is intended to be a full flow staged combustion design using LCH4/LOX. As no one has ever operated a full flow staged combustion engine of this size, nor a methane/LOX engine larger than a small RCS thruster, it is presumptive to argue that the development of this vehicle will smoothly follow plan. In general, accepting Elon Musk’s pronouncements about what SpaceX will do at a detail level has a poor history of matching reality. This is not to detract from the accomplishments SpaceX has demonstrated–which are still quite impressive–but rather to take such claims with a grain of salt until the evidence shows otherwise.

Stranger

8

Yes, I second Dr. Strangelove. I did this same calculation in another post. What Stranger is correct about is that suppose out of all possible engine failure modes, x% of them are such that they will kill the rocket anyway, even if the avionics attempt to shut down that failed engine.

For instance, a failure that causes an engine to literally fall off the rocket, letting fuel drain freely out of the rocket would be a mission killer.

For those failure modes, more engines means more risk. Do you have any possible data on what “X” might actually be? If x is small, SpaceX’s approach is superior, if large, Delta’s approach is better.

9

SpaceX claims that the Merlin 1D is only running at 85% of its capacity in the F9 1.1 configuration. It’s not clear precisely what this means–I suppose only they know for sure. But it may hint that they have enough thrust reserve to fully compensate for an engine out. It may be that running the engine at a higher thrust level increases the failure probability past where they would prefer under normal conditions, but clearly that’s better than losing the payload completely.

I don’t disagree that there’s great value in robustness and simplicity, but there’s still a time and place for complication. As you note, the RD-180 is a remarkable engine, in no small part due to its efficient staged combustion cycle. It’s more complicated than a single Merlin 1D with its gas generator cycle. Are 9 simpler engines better or worse than a single complicated engine? It’s design tradeoff and there’s no single right answer.

Yup.

I was thinking about the Shuttle recently and realized how much I underestimated the technology of the external tank. It seems to get short shrift mindshare-wise, since it isn’t reusable, and it neither looks as cool as the orbiter nor puts out the incredible thrust of the boosters. But it’s a reliable LH2-LOX tank, with an impressive mass ratio (nearly 30:1), and furthermore handles incredible side loads coming from the boosters and orbiter. Not an ideal situation for a cylindrical tank.

Rip out the reinforcement you need for side loads, put some engines on the bottom, and you’d have a pretty decent SSTO…

10

It is easier to test and characterize a single engine versus all of the controls, plumbing, mounting structure, and co-induced environments from a gaggle of engines. One of the lessons from the N-1 rocket and the Saturn V (both superheavy lift vehicles) is that the complex interactions between multiple engines. The advantage of being able to tolerate a single engine shutdown–which, again, is only advantageous in a fairly limited set of circumstances–is massively overweighed by having sufficient confidence that engine shutdown is a highly infrequent event. The reason SpaceX has chosen to gang together so many engines isn’t because of the purported reliability, but because they already started with an engine of a certain size and scaling up a Merlin-class engine to an order of magnitude higher thrust is a very expensive and challenging propsition, as SpaceX propulsion VP Tom Mueller would know having previously worked on the TR-107 development at TRW.

I assume this is coming from the SSTO “thought experiment” proposed by Gary Hudson. However, Hudson makes a number of non-conservative assumptions about the mass of the ET, essentially assuming that a thrust structure to accommodate the six SSMEs would be 30,000 lbm, with a GLOW of 830 metric tons and a vehicle TWR of 1.34 providing almost 27 tons of payload to an orbit velocity of 9.1 km/s. However, this neglects to consider the completely different structural load path and also skin reinforcement necessary to withstand both the axial loads and protect against “oil can” modes. The External Tank (ET) as currently designed is really a minimally loaded stucture with it and its propellant largely dragged by the Solid Rocket Boosters; as a primary thrust structure it would require significant reinforcement. We don’t have to just theorize about this, though; we can look at the Space Launch System (SLS) core stage, which is essentially an ET structure modified in a fashion similar to Hudson’s SSTO (albeit to accept four RS-25s rather than the six that Hudson specifies, and uses two SRBs to provide the bulk of thrust (>80%) from liftoff through T+270 seconds). The SLS core stage has an dry mass of greater than 85 tons versus the ET at 26.5 tons, a difference of almost 60 tons; we could only expect an SSTO with a purely axial thrust structure to be even heavier.

Note, too, that the advertised 60 ton “payload” mass includes the payload adatper, fairing, and other necessary elements. While I haven’t actually run a POST or TAOS simulation for a vehicle with these parameters, given the poor TWR of the proposed SSTO combined with this highly optimistic mass ratio I think such a vehicle is likely to be marginal at best. And this is the general problem with SSTOs that doesn’t scale; the dry mass to orbit versus GLOW ends up being so high that the additional capability left over for payload is almost negligble, and if you want a reusable SSTO all of the necessary protection systems and added robustness further detract from available payload mass. Until propellant mass efficiency improves by at least ~30%, a heavy lift SSTO is probably not viable, though a case can be made for the integrated aerospike-propelled Chrysler SERV concept if sufficient propulsive performance could be achieved. Realistically, a working continuous wave detonation rocket engine is probably necessary to achieve useful payload (beyond personnel) to orbit in a single stage vehicle.

Stranger

11

Scary to think that I’ve been reading the two of you discussing enough stuff that I didn’t even have to stop and think about what SSTO meant…:slight_smile: