Launch vehicles are not Tinkertoys. The system is tightly integrated. Yes, you could make a lower stage booster that feeds from external tankage above and can separate, but now you’ve created a whole new set of interfaces which also have to be non-destructively severed in flight, and then have to be reintegrated. This required a very complex propellant feed system on the STS, and while the External Tank (ET) was expendable there was serious consideration given to trying to recover the ET given the cost (>$150m per tank). The additional complexity of a separable propulsion system (which would still have to carry propellant to slow and return to the surface) would add more cost and inert (non-propellant mass). If the ultimate desire is to bring the stage down and reuse it with minimal refurbishment, bringing it down whole make sense and while challenging is feasible (as demonstrated).
Stranger, if I have understood you correctly, SpaceX documented costs are competitive vis a via European and US launch systems, but not so much against Russian or Chinese ones. Being a US company, it can bid for DoD business more easily.
From your numbers it seems that heavy lift vehicles have a lower cost per kg. Why then super heavy lifters of like of Saturn V or the. Energia not been produced. It seems that they would be more economical for talks like building the ISS.
Spin stabilization is used on smaller vehicles, especially solid propellant sounding rocket; the SPARK smallsat launch vehicle (also known as SuperStrypi) uses spin stabilization for the first stage.
Spin stab would not be desirable for a large liquid propellant vehicle because of the slosh it would induce during changes in spin rate or nutation due to changes in vehicle thrust axis versus the spin axis. It would also not be desirable for return because the vehicle would have to despin before landing. And you certainly could not spin stabilize a crewed vehicle for obvious reasons.
Stabilizing the lightened vehicle on descent is actually easier than the ascent, especially if fixed or deployable fins are used, hence why the New Shepard has the reverse position fins at the top of the vehicle, and I suspect SpaceX will add deployable fins to their Stage 1. The challenge is damping any residual propellant slosh, especially through the transonic regime.
I was under the impression that only the gyro rotors inside the vehicle need to be spinning, and not the whole spacecraft as if it were a gigantic rifle bullet. If one could somehow install enough tiny gyro rotors inside a broomstick, wouldn’t it be able to stay erect all on its own?
The cost per kilogram is lower in superheavy vehicles, but the overall cost is still very large. A vehicle like the Saturn V, which was purpose built to send the Apollo CSM to Lunar orbit, is just too large to use for any practical purpose, e.g. orbiting even large telecom satellites. Sure, it could carry multiple large satellites, but they’d each need to go to their own orbit. As a bulk goods hauler, its still too expensive, hence the proposals for “big dumb booster” type vehicles such as the Boeing Double Bubble or a low performance mass hauler like the Sea Dragon which are designed for simplicity at the expensive of performance and reliability by redundancy (although I personally think these vehicles would probably have made back the reliability via robustness and simplicity).
Again, the costs are the processing, and it doesn’t scale linearly with payload. Although you realize “economies of scale” with larger vehicles, the real overall cost reduction opportunity is streamlining and simplifying the integration process to minimize labor.
Control moment gyroscopes (CMG) are used for attitude control in some space vehicles but there is no way you could build a gyroscope with a high enough moment of inertia to stabilize a large rocket. Even if you could, you wouldn’t want to; it would render the rocket incapable of changing direction when it needs to turn or rotate. Gyroscopes are used in the inertial measurement unit (IMU) to detect rotation, but they’re free to rotate and don’t exert any net torque on the vehicle.
Of course, all this implies that building a big rocket and sacrificing payload to achieve other goals is a reasonable strategy. I recall SpaceX estimating a 70% payload hit for full reusability (all stages). That’s significant, but if ground costs are sublinear (and fuel costs are in the noise), then you may as well just build a bigger rocket. I expect that (in several years, when the tech has matured some more) some payloads which could be launched on an expendable Falcon 9 will instead be launched on a fully-reusable Falcon Heavy.
Stranger covers it pretty well, but there’s another significant factor. The stage, tank and all, acts as a lifting body. That is to say, it can use aerodynamics for a degree of maneuvering. The same couldn’t be said if the octoweb structure decoupled, since it is a dense structure that would descend ballistically.
In their Falcon 9 Reusable program (F9R, or “Falcon Niner”), they have experimented with using grid fins for extra control. Grid fins are quite effective at both subsonic and super/hypersonic speeds (though not so well in the transonic regime), but they still require attachment to an aerodynamic body. It’s not clear if or when they’ll add these fins to their real launch vehicles, but its clear that they think it’s a possibility (right now, they use cold gas thrusters, which are effective but not too efficient).
The parenthetical part seems odd to me. The Air Force definitely seems resistant to using what they’d consider unproven vehicles. But that seems as much in favor of landing legs as against. If SpaceX can meet the qualification requirements of the F9 with reusability kit, and especially if SpaceX maintains a high launch rate for that variant, then the Air Force should be, if anything, opposed to the removal of the equipment.
Of course, I’m sure SpaceX will be happy to launch whatever the AF asks. The Dragon capsule is also reusable and yet SpaceX builds a new one for each ISS resupply mission.
At any rate, it does bring up an interesting and more general point. The military appears to still view rockets as artillery. It may take quite an institutional attitude change to see things differently. In most other areas, we see a “fresh” machine as being less reliable than a gently used one. Generally, the failure rate for any machine follows what’s called a bathtub curve, with a high failure rate in the beginning (“infant mortality”, where outright manufacturing defects show up), a low mid-life failure rate, and then again a high rate as the device reaches the end of its design lifetime.
So once SpaceX’s reusability program is going full steam, the smart customer will want a rocket that’s flown once or twice already. There may even come a point–as with commercial aircraft–where customers simply will not accept a vehicle that hasn’t flown already. This is probably some time off and I’m certain you will remind me that rockets are not commercial aircraft, but in this particular case I think the same basic principles apply.
Not that this applies to a separable octoweb, but SpaceX will face a similar challenge with the crossfeed version of the Falcon Heavy. As you may be aware, SpaceX plans on feeding some of the center core engines from the booster core tanks. Any advanced player of Kerbal Space Program knows that this strategy increases payload by dumping dead weight earlier in the flight. It is a simple enough idea but it’s clear that the implementation will not be trivial.
Rocket launch vehicles are fundamentally unlike any conventional transportation systems. They operate under environments not experienced by any terrestrial system. In particular, the vibration, turbopump speeds, and the temperature extremes experienced by cryogenic valves and seals in the propellant feed system induce vibratory and thermal fatigue conditions in minutes that are equivalent to hours of service life in what even the most aggressive terrestrial applications such as fighter aircraft or mining excavation equipment experience. These conditions challenge the fundamental limits of material capability. The only way to increase service life is to either build systems which can help isolate against vibration and temperature loads, or reduce the pressures and conditions experienced by the components. With the limitations of current propulsion systems, minimizing inert weight and maximizing propulsive performance is crucial, but it also means that the systems can only be built to survive a limited lifetime. This lifetime for an expendable system typically includes a three of acceptance test cycles and the flight duration, which is established by the qualification durations and environments. In order to hold the same standard for a reusable system the qualification duration has to be much, much longer. (Similar standards apply to turbofan engines used for commercial aircraft, and yes, they do operate qualification units for thousands of hours and under extreme operating conditions before they are certified for use.)
The “bathtub curve” model of early latent failures and a long span of reliable operational life just don’t apply with the narrow margins allowable in rocket propulsion systems, especially with some many potential single failure points. And yes, I’m aware of the claimed redundancy of the Falcon 9, but if you rupture a propellant line or have FOD that feeds into all engines, having “single engine out capability” and triple redundant avionics won’t help. Under such abbreviated lifetime flight history does not provide guarantee of future success; it means that the vehicle is closer to an inevitable failure condition.
None of the EELV New Entrant Certification Flights carried landing legs. The landing legs detract from flight margins and should they deploy prematurely would almost certainly cause a loss of vehicle failure. The Air Force is not resistant to reusable vehicles per se, and in fact the EELV replacement program (the 2025 Reusable Booster System) planned to develop a fleet of two stage vehicles with a first stage with horizontal recovery (presumably using a glide body or horizontal wings). A 2012 report by the National Research Council reviewing the concepts and the ultimate development costs resulted in the system being cancelled for much the same reason that Orbital elected not to develop Antares into a reusable system, e.g. the rate of flights just doesn’t justify the investment in reusability on a cost basis. Air Force missions on the EELV vehicles are typically Class A or Class B payloads where the need for reliability trumps any hypothetical savings from reusability.
The many single failure points is exactly the reason why the bathtub curve is applicable. Consider the case of incorrectly installed inertial sensors (gyros, accelerometers). A surprising number of rockets fail for this reason–I am the farthest thing from a launch historian and yet I can name two examples off the top of my head. Now, a rocket with incorrectly installed sensors will fail, catastrophically and spectacularly. Therefore, you can be sure that if a rocket has a chance to fly a second time, it won’t fail for that particular reason.
Ground testing helps to eat some of the initial peak of the curve. But as you note, rockets are extremely complicated systems, and there are some things you can only test by actually flying.
Fair enough, but that wasn’t really my point. The military is traditionally conservative–once they’ve qualified a vehicle, they don’t seem to like changes. I’ve heard that SpaceX’s small fixes from one vehicle to next pose a problem in this respect. To me, this implies that the AF cares less about the legs themselves than they do about having a vehicle that will fly in the exact configuration as it was certified with.
That said, I see you are correct that the flights the AF has certified so far were sans reusability kit. Perhaps they will additionally certify the kit at some future point.
First of all, misassembly of components such as the IMU that was installed backwards in the Proton-M that crashed in June 2013 is not an example of a latent failure; this is a process and inspection failure (and possibly a design failure as well insofar as the part could be installed improperly and still function). Latent failures are failures of a component or system which are not detectable or inspectable by any feasible method. For instance, the occasional random failure of a bearing due to improper surface hardening or underperformance of a bolt cutter cartridge (even though the sample population in the lot inspection test showed good function) are examples of latent failures, because there just isn’t a practical way to check out each unit. And as IMUs with sufficient precision to be used in launch vehicles require extensive bench calibration prior to use this is likely a component that would have to be removed and recalibrated between flights, so having flown it once doesn’t mean it will be correctly installed in subsequent launches. The proper resolution to this kind of failure is to design the installation such that it cannot possibly be installed incorrectly (e.g. using a locating stud or asymmetric bolt pattern) or improper installation will result in an obvious and inspectable problem (e.g. cables are routed so that it will only mate up if oriented correctly). This, plus rigorous quality assurance provisions assure that such mistakes will catch such process failures.
The “bathtub curve” model is applicable to components and systems which have a very long lifetime compared to the infancy period; that is, components for which a latent failure will occur in a few tens or hundreds of seconds with an operating lifetime of many millions of seconds (or operating cycles for discrete function components) under relatively steady loads before wear- and age-related failures start to occur. This in no way describes the propulsion systems of launch vehicles and the components within them, which operated at very high loads and conditions for a few hundreds of seconds, experiencing considerable wear during every operational cycle. The “infancy” portion of a bathtub curve for such components usually occurs in the first few seconds of operation and is intended to be revealed by the acceptance test cycle (hopefully at the component level before a component is even installed) and certainly before the vehicle is flown with several hundred millions of dollars worth of payload. Such vehicles will experience continual wear and degradation of both performance and operating margins throughout their life between refurbishments. And while it may be the intention of SpaceX to develop a system that is sufficient robust to allow rapid turnaround with minimal refurbishment analogous to a commercial airliner, the current F9v1.1 in no way meets criteria. Just as with the Shuttle Orbiter Vehicle, current plans call for the engines to be removed and re-ATP’d and other components to be refurbished. A truly reusable vehicle that can be flown with no refurbishment will require either substantial advances in the state of the art of materials science or a tradeoff of performance for lower loads and longer reliable performance life.
It is not true that military systems are not upgraded or improved (i.e. “once they’ve qualified a vehicle, they don’t seem to like changes”); in fact, the bulk of sustainment costs on most military systems are continual upgrade programs that typically make more profit for the contractor than the original system development. Even a cursory look at systems like the B-52, F/A-18, or the Delta/Delta II launch vehicle (which saw over four decades of incremental upgrades) will evidence this to be true. However, when systems are upgraded they are requalified or delta qualified to assure that the changes don’t have some detrimental performance or reliability impact upon other parts of the system. What isn’t acceptable is to arbitrarily make incremental changes without some level of requalification, or claim that a system is qualified even with substantial changes. (SpaceX tried to pull this one by initially arguing that the F9v1.0 flights should be considered in certification of the F9v1.1 vehicle even though every major system except for avionics is completely different, and that certification of F9v1.1 should also feed toward certifying the three core Falcon Heavy.) When applied correctly, this methodology has resulted in highly reliable systems despite their complexity; Delta II is the reliability champion of US launch systems, and for all of their cost the EELV vehicles (Atlas V and Delta IV) have had remarkably successful records for new design vehicles (47 successes in 48 attempts and 26 of 27, respectively).
Its all well and fine to talk about the benefits from reusability in abstract terms but the reality is that attaining true, turn-it-around-in-a-day reusability in conventional, multi-stage rocket launch systems requires a lot of development effort and cost at the detail level and some advances in basic materials technology if the desire is to operate at high operating levels with the slender margins to compare to the performance of conventional vehicles. This isn’t going to be done with commercial off the shelf (COTS) components and the “soonest-is-best” programmatic approach; like the modern commercial aircraft, it requires a large amount of development testing and, most importantly, failures that are fully understood so that the boundaries of the capability “box” can be thoroughly characterized and understood. The other alternative is to go to lower performing systems with large operating margins (high robustness) and less complex design (simplicity), a la the Truax Sea Dragon, the Boeing Double Bubble, or other minimum cost designs (MCD) which trade propulsive efficiency for lower required tolerances and higher margins so that you don’t need to so precisely demarcate the edges of the box or worry about the complex interplay of various components creating a problem you didn’t anticipate in performance and reliability analysis.
This is a longer term goal: immediately, the falcon heavy will be very similar to the delta IV heavy, in that the center stage will throttle down for some time to conserve fuel, but without any crossfeed.
The advantages of crossfeed in real life are less than they are in Kerbal space program. In KSP, the engines have much lower thrust to mass ratios and the tanks have proportionally higher masses compared to the amount of fuel they containe, compared to real life. This makes it a huge advantage in KSP to dump engines and empty tanks as soon as you don’t need them, but the advantage is muted for real life rockets.
For real life numbers, SpaceX claims a high-end payload of about 53 tonnes to LEO with a falcon heavy with crossfeed, compared to 45 tonnes without. Keep in mind that the falcon heavy hasn’t flown once yet, in any configuration. I’ll believe 53 t, or even 43 t, when I see it. And if you push up to the upper end of your mass capabilities, you lose any chance at recovery, as well as your engine-out margin.
Does anyone know what the mass of the payload of the first falcon heavy flight is going to be? SpaceX apparently doesn’t really like to talk about the details of their missions.
I don’t see how classifying the type of failure excludes it from contributing to the bathtub curve. Significant design failures tend to get weeded out early on for mass produced devices, though even that’s not always the case.
This seems really unlikely to me. Modern IMUs don’t shake apart with a little vibration. And while they probably aren’t using sensors pulled from a smartphone, they also aren’t using a dead-reckoning unit from a nuclear sub.
No doubt, Murphy-resistant design is an important element of engineering. Amusingly, the original Murphy’s law was formulated in response to some backwards-wired sensors that were effectively to act as accelerometers.
Nothing’s perfect. It’s certainly a laudable goal, and launch providers do a commendable job given their low flight rate, but process failures will still happen. A single actual flight will tell you if there were any critical ones that slipped through QA.
That is by no means always the case. Some failures happen immediately. Others may take much more time, though still low compared to the design lifetime. In my personal experience, thermal problems of various sorts tend to have this type of behavior (thermal cycling, overheating, etc.), but there are other possibilities.
That would be the hope, yes. SpaceX does well in this regard with their clamp-down engine firings. It’s still not a flight environment, with the acceleration and vibration and other things going on.
Perhaps you will note my original qualification: once SpaceX’s reusability program is going full steam. SpaceX has already said they will have failed if each flight requires major refurbishment. Obviously they are not there yet. Whether they get there eventually remains to be seen, but that is their focus and what they (and I) consider to be an actual resuability program. Perhaps it will require a v1.2.
SpaceX has already said the Merlin 1D is running at 85% capacity. It’s not clear what this means precisely, but the most obvious interpretation is that they have given themselves some margin for the sake of reliability.
They also claim that the Falcon 9 v1.1 has a 40% structural margin, while other rockets are typically 25%. I’m not sure where these numbers come from or how realistic they are, but they certainly believe they have overbuilt their rocket.
I was admittedly unclear, but that’s not what I meant. There is a vast difference between significant upgrade programs vs. small per-vehicle tweaks (the way SpaceX has been operating so far). No one wants every vehicle to be slightly different, but SpaceX is using their customer’s payloads to prototype new hardware.
Obviously this speeds up development time (and reduces development expense) at the theoretical expense of reliability. An actual impact on mission success has not yet been borne out. It’s not surprising that the military would be suspicious of this approach, but that doesn’t mean they are entirely in the right.
It’s also not obvious that a reliability-at-any-expense is the right attitude, even if reliability is your prime metric.
I think it’s clear that SpaceX is using both approaches and hope to meet in the middle. Their engines use a relatively simple gas generator cycle and are lower efficiency than similar Russian engines. They run on RP-1/LOX. They appear to run well below their design limits.
On the other hand, they are progressively refining their design and flight envelope with each new flight. And at the same time, their Grasshopper/F9R effort appears to be modeled after the DC-X program, where the goals were not simply learning how to fly the rocket, but also how to achieve quick turnaround with a minimum of ground crew. They’ll hammer on each “long pole” for turnaround time as they reach them, and they can iterate much faster as compared to real launches.
I’m not going to attempt to address the line-by-line response other than to say that your arguments are based on a naive understanding of reliability analysis, acceptance and qualification testing, and vehicle integration processes. Rocket launch vehicles are fundamentally not like other consumer or commercial products. When they fail, you don’t pull off on the curb or send them back to the factory for a replacement; they fall out of the sky taking multi-million dollar payloads with them. This risk of loss may be acceptable for lower cost commercial smallsats looking for the cheapest ride they can get, but it isn’t for large many hundred million dollar telecom birds or highly critical national security payloads, hence why the EELV certification only applies to unflown vehicles.
The degree to which the Falcon lower stages are reusable remains to be seen, but the lesson of previous attempts to develop recovery and reuse capability is that it is full of unexpected difficulties. There is no reasonable expectation that SpaceX will be able to turn around a stage without significant refurbishment as currently designed, and no, they can’t just punch the reset switch on an IMU and reinitialize it as can be done with an aircraft navigation system; the degree of precision required for space applications requires bench calibration. It is clear from the recent delays in meeting manifests that SpaceX is not realizing the planned launch rates because they’ve underestimated the difficulties in integration and processing, and the same can be expected of attempts at refurbishment and reflight.
Perhaps not. But significant refurbishment is not a reusability program. If SpaceX finds that there’s no way around frequent rebuilds, off-vehicle recalibration of their instruments, and so on, they will discontinue their program. In the long run, it’s all or nothing. Only politically-motivated programs like STS could ever exist in this indeterminate state indefinitely.
I don’t think the evidence bears this out. They have indeed experienced long delays, but they have also shown short turnarounds (33 days from SES-8 to Thaicom 6, and 22 days from OG2 to AsiaSat 8). Obviously, the average is much higher than this.
This does not appear to be evidence that they simply underestimated integration and processing time. You would see a consistently long delay in that case. Instead, it suggests that there are several random factors that they do not have sufficient control over.
For instance, the OG2 mission was plagued by numerous delays related to helium leaks. Why was that? Internal QA failure? Poor process control? Variation from external supplier? Who knows–probably a combination of things. The point is, sometimes they “get lucky” and fly without issue, and in those cases a <1 month turnaround is doable. Once they nail down these issues, there’s no reason to believe they can’t sustain that rate.
I looked it up, and at least some Falcon 9 missions have used the Northrop Grumman LN-200 IMU on both the Dragon capsule and Falcon rocket. It’s certainly a high quality IMU, but there’s nothing extraordinary about it in terms of specifications. It uses MEMS accelerometers and fiber optic gyros. It’s qualified for deep space missions of up to 6 years and has been actually used for missions of over 10 years. It’s rated for shock/vibration/etc. loads well beyond what a rocket would experience.
So I highly doubt your claim that the IMU requires bench calibration after each flight. In fact, I doubt it requires any kind of recalibration at all–it is entirely a solid-state unit and there aren’t too many progressive degradation mechanisms. The brochure makes a specific point of the long lifetime. As you note, replacement would only offer the opportunity for misinstallation.
