Pushing the limits of the hoverslam landing

[matthewkantar]

Indeed. There is some value to knowing those things to some extent, but a control loop can be pretty dumb and still work just fine getting you to where you need to go.

I think the question wasn't so much about staying on track, but in cases where fuel reserves are at a bare minimum, knowing whether or not there will be enough fuel remaining to complete the landing.

If some set of data,such as remaining fuel, is moot, or would not in any way change the control inputs, I don't think it would be included in the landing program. Finding out the stage is running out of propellants is a bummer, but unless there is something that can be done with that information it is useless. No filling stations on the way down.

There is one action the stage could take based on such data, but to me it seems unlikely it would switch to progressively more aggressive burns to save on propellant.

Matthew

[llanitedave]

Indeed. There is some value to knowing those things to some extent, but a control loop can be pretty dumb and still work just fine getting you to where you need to go.

I think the question wasn't so much about staying on track, but in cases where fuel reserves are at a bare minimum, knowing whether or not there will be enough fuel remaining to complete the landing.

Why does the stage guidance system need to know that? Either there is, and the landing presumably succeeds, or there isn't, and it fails one way or another. Telemetry lets the ground later figure out why it failed (e.g., due to lack of fuel) and they adjust things to hopefully improve odds of success the next time. Or decide landing isn't feasible with that payload and they don't even try further for a payload/mission of that type in the future.

You could save yourself the expense of patching holes in the barge.

[Herb Schaltegger]

Indeed. There is some value to knowing those things to some extent, but a control loop can be pretty dumb and still work just fine getting you to where you need to go.

I think the question wasn't so much about staying on track, but in cases where fuel reserves are at a bare minimum, knowing whether or not there will be enough fuel remaining to complete the landing.

Why does the stage guidance system need to know that? Either there is, and the landing presumably succeeds, or there isn't, and it fails one way or another. Telemetry lets the ground later figure out why it failed (e.g., due to lack of fuel) and they adjust things to hopefully improve odds of success the next time. Or decide landing isn't feasible with that payload and they don't even try further for a payload/mission of that type in the future.

You could save yourself the expense of patching holes in the barge.

They've hit the barge a few times, and holing the deck has happened once. Pretty obviously, at some point once they've reached the point where they think they've learned all they can, SpaceX won't continue to rocket-punch their barge if they don't think recovery is going to work. They'll go into some semblance of an "operational mode" - recovery will either be routine and expected to succeed most of the time, or it won't and they'll stop bothering for at least some heavy/high-energy payloads.

[malu5531]

And the computer doesn't need to know how much propellant remains. It's probably pre-programmed to do an entry burn of specific time duration, and the landing burn is probably pre-programmed to start at a certain altitude/velocity, based on pre-launch Monte Carlo simulations that give them a good idea of how much propellant will be used during the burns.

My impression is that they're not pre-programming control timings such as when to start engine burns, rather they tweak constraint parameters for an onboard algorithm for real-time optimisation of landing trajectory. I.e., the rocket decides in flight when to start / stop or throttle engines.

Attached is a nice paper by Lars Blackmore, the person responsible for F9 EDL at SpaceX. It describes an onboard "Powered Descent Guidance" algorithm, which optimise landing trajectory for minimal landing error and fuel use, with given limits set on throttle, speed, position, etc.

Continuous onboard optimization during EDL is needed since initial conditions at staging are not known beforehand AND conditions change during flight such as wind gusts, high altitude jet-streams, engine performance, control accuracy, etc.

I believe SpaceX is currently in a phase of iteratively adjusting constraints. If SpaceX is very aggressive in finding the constraint envelope, it may be that there's quite a bit more fuel left in the rocket after landing. I.e. the warning that they might crash stage may not be due to "landing on fumes", but rather due to constraint experiments in order to expand the envelope and evaluate the overall performance.

[Jcc]

You would think propellant load could be calculated based on knowing the thrust of the engines and the deceleration it produces. More deceleration for the same thrust means less mass of the stage, subtract the dry mass and you get the fuel. Maybe that can't be measured with enough precision, if so, upgrade the sensors.

[Bob Shaw]

You would think propellant load could be calculated based on knowing the thrust of the engines and the deceleration it produces. More deceleration for the same thrust means less mass of the stage, subtract the dry mass and you get the fuel. Maybe that can't be measured with enough precision, if so, upgrade the sensors.

No. Rocket fuel is notoriously difficult to measure in-flight, especially when in zero-G. It also sloshes, gurgles and bounces around when under thrust, changes density, and will generally seek the 'lowest' point at the end of the flight which may well be different from the 'lowest' point during launch, as the stage may be canted into the wind. You have to have reserves; the saving grace for SpaceX is that the thrust/weight ratio on the F9 is absurd when the tanks are almost empty and there's no second stage/payload on top, so they can get away with murder, eg the infamous hoverslam.

[philw1776]

And the computer doesn't need to know how much propellant remains. It's probably pre-programmed to do an entry burn of specific time duration, and the landing burn is probably pre-programmed to start at a certain altitude/velocity, based on pre-launch Monte Carlo simulations that give them a good idea of how much propellant will be used during the burns.

My impression is that they're not pre-programming control timings such as when to start engine burns, rather they tweak constraint parameters for an onboard algorithm for real-time optimisation of landing trajectory. I.e., the rocket decides in flight when to start / stop or throttle engines.

Attached is a nice paper by Lars Blackmore, the person responsible for F9 EDL at SpaceX. It describes an onboard "Powered Descent Guidance" algorithm, which optimise landing trajectory for minimal landing error and fuel use, with given limits set on throttle, speed, position, etc.

Continuous onboard optimization during EDL is needed since initial conditions at staging are not known beforehand AND conditions change during flight such as wind gusts, high altitude jet-streams, engine performance, control accuracy, etc.

I believe SpaceX is currently in a phase of iteratively adjusting constraints. If SpaceX is very aggressive in finding the constraint envelope, it may be that there's quite a bit more fuel left in the rocket after landing. I.e. the warning that they might crash stage may not be due to "landing on fumes", but rather due to constraint experiments in order to expand the envelope and evaluate the overall performance.

Looking at Dr Blackmore's impressive list of publications I notice a couple on balloons in Titan's atmosphere.
A future FH launched Dragon 2 destination in the 20s?

[Zed_Noir]

You would think propellant load could be calculated based on knowing the thrust of the engines and the deceleration it produces. More deceleration for the same thrust means less mass of the stage, subtract the dry mass and you get the fuel. Maybe that can't be measured with enough precision, if so, upgrade the sensors.

No. Rocket fuel is notoriously difficult to measure in-flight, especially when in zero-G. It also sloshes, gurgles and bounces around when under thrust, changes density, and will generally seek the 'lowest' point at the end of the flight which may well be different from the 'lowest' point during launch, as the stage may be canted into the wind. You have to have reserves; the saving grace for SpaceX is that the thrust/weight ratio on the F9 is absurd when the tanks are almost empty and there's no second stage/payload on top, so they can get away with murder, eg the infamous hoverslam.

Recent videos from propellant tank interiors show that visual markings inside the tank can be use for gauging the volume of remaining propellants. At least while the vehicle is under acceleration.

[Bob Shaw]

Oh, certainly - things will get smarter. But not all at once. The big thing SpaceX has going for it is, simply, that empty tankage is light. Everything else pales into invisibility compared to that.

[hkultala]

I just want to point out that the Falcon 9 first stage carries approximately 286,400 kg in LOX and 123,100 in RP-1, approximate total of 409,500 kg.

Source: http://spaceflight101.com/spacerockets/falcon-9-ft/

A savings of 2,000 kg of propellants is about 0.5% of the Falcon 9's total fuel capacity.

So more aggressive landings would only help in the most marginal of return scenarios, where the Falcon 9 has already depleted almost all of its available fuel and oxidizer.

2000 kg of fuel saved in the landing phase does not mean only either
1) Having 2000 kg more fuel to spent in the ascent and boostback phase (0.5 % here is not much)
2) Having 2000 kg less weight to first accelerate to staging velocity(~1.5%)
3) Having 2000 kg less weight to boost back. (>5% for this)

It means the combination of  ALL of these.

And because rocket equation is a highly nonlinear equation, the benefits of these together is much higher than the sum of either done separately.

[speedevil]

And because rocket equation is a highly nonlinear equation, the benefits of these together is much higher than the sum of either done separately.

It's only highly nonlinear if you are at the portion where the mass is changing rapidly.
At the state where the tank is nearly empty, and the fuel is no longer overwhelmingly dominant, it's gets close to linear.

[AncientU]

There seem to be a few factors that optimize fuel use (and thus delivered payload) by conducting a three engine landing burn:
1.  Extra fuel saved for landing is fuel that could have been used for boost during the most productive final few seconds of the burn -- 5g burn with almost empty tankage.  This is the fuel you most want to conserve -- it provides much more than 0.5% of acceleration (I think).
2.  Landing with minimal fuel also improves the ballistic coefficient, allowing the atmosphere to slow the booster more instead of the landing burn doing that deceleration, and
3.  Waiting till last seconds allows more deceleration in the thickest portions of the atmosphere. 
These last two each reduce the amount of fuel needed for landing.

[John Alan]

I have sometimes wondered the trades... of adding some sort of speed brake to be deployed when it goes subsonic...
A)drogue chutes (then cut all loose when speed <~20 knots)
B)form fitted 'speed brakes' that deploy in between the grid fins (likely NFG on a mass trade study)

Obviously these all need thick air to work in...
...and they have to sort out the 3 engine super slam first, to maximize that free resource to work with it... 

[MikeAtkinson]

There seem to be a few factors that optimize fuel use (and thus delivered payload) by conducting a three engine landing burn:
1.  Extra fuel saved for landing is fuel that could have been used for boost during the most productive final few seconds of the burn -- 5g burn with almost empty tankage.  This is the fuel you most want to conserve -- it provides much more than 0.5% of acceleration (I think).
2.  Landing with minimal fuel also improves the ballistic coefficient, allowing the atmosphere to slow the booster more instead of the landing burn doing that deceleration, and
3.  Waiting till last seconds allows more deceleration in the thickest portions of the atmosphere. 
These last two each reduce the amount of fuel needed for landing.

Agree, but there is likely to be a larger reserve of un-burned fuel on landing to ensure that none of the three engines run out of fuel.

[deruch]

There seem to be a few factors that optimize fuel use (and thus delivered payload) by conducting a three engine landing burn:
1.  Extra fuel saved for landing is fuel that could have been used for boost during the most productive final few seconds of the burn -- 5g burn with almost empty tankage.  This is the fuel you most want to conserve -- it provides much more than 0.5% of acceleration (I think).
2.  Landing with minimal fuel also improves the ballistic coefficient, allowing the atmosphere to slow the booster more instead of the landing burn doing that deceleration, and
3.  Waiting till last seconds allows more deceleration in the thickest portions of the atmosphere. 
These last two each reduce the amount of fuel needed for landing.

Though, it may not be a 1:1 transfer from savings on the landing to extra boost, as increasing the boost may also require that they add some to the entry burn in order to remain survivable/maintain reusable condition, etc.  But, any addition can be worth it.  The point is that optimizing the landing burn as much as possible either expands the envelope for what is considered a recoverable mission, allows them to improve the condition of recovered boosters by protecting them more on reentry, or improves the performance on what orbit a given payload can be delivered to.