SpaceX Falcon 9 : CRS-16 (Dragon SpX-16) : December 5, 2018 - DISCUSSION

[AJW]

Through many decades of hard lessons, modern aircraft use both multiple actuators on each control surface as well as multiple independent hydraulic systems.  In commercial it is common to have at least three hydraulic systems using even different power sources.  As an example, Captain ‘Sully’s decision to start the APU early was considered critical in keeping his aircraft flying.   There are additional best practices for areas such as hydraulic line routing that need to be followed, so a single event is less likely to sever multiple controls. (UA232)

I think it has been an error to not have used redundant hydraulics, a standard that was used even on the Saturn V guidance and control system.   The CRS-16 control failure was just a matter of time and with landings moved out of the 'experimental' stage, I won't be surprised if RTLS is stopped until this single point of failure is properly addressed.

[LouScheffer]

I think it has been an error to not have used redundant hydraulics, a standard that was used even on the Saturn V guidance and control system.   The CRS-16 control failure was just a matter of time and with landings moved out of the 'experimental' stage, I won't be surprised if RTLS is stopped until this single point of failure is properly addressed.

It's clear they need to do something.  A 1 in 30 chance of losing a $10m booster costs them $300k each time on the average.  But it's not obvious to me the best thing to do is make it redundant.  If instead they can make it 100x more reliable, then the cost of $3K would likely be less than redundant pumps, lines, tests, etc.  And since it only runs for 8 minutes at a time, finding or creating a pump that fails less than once per 8 minutes x 3000 missions = 24,000 minutes = every 17 days, does not seem to be impossible.  Lots of other hydraulics are more reliable than that.

Obviously, I don't have the detailed data to say whether a redundant pump, or a better pump, is the right solution.  I'm just pointing out that low cost is the goal, not reliability per se, and the more reliable system may not be the cheapest overall.

[Lars-J]

I think it has been an error to not have used redundant hydraulics, a standard that was used even on the Saturn V guidance and control system.   The CRS-16 control failure was just a matter of time and with landings moved out of the 'experimental' stage, I won't be surprised if RTLS is stopped until this single point of failure is properly addressed.

The landings are clearly not out of the experimental stage. You seem to not acknowledge the difference between flight critical systems (that are redundant) to landing systems (some which are not).

Did I miss Saturn V landings that exhibited redundant landing systems?

As for RTLS being stopped, I wouldn’t bet on it. Odds are SpaceX might address it before it would become an issue anyways.

[niwax]

I think it has been an error to not have used redundant hydraulics, a standard that was used even on the Saturn V guidance and control system.   The CRS-16 control failure was just a matter of time and with landings moved out of the 'experimental' stage, I won't be surprised if RTLS is stopped until this single point of failure is properly addressed.

The landings are clearly not out of the experimental stage. You seem to not acknowledge the difference between flight critical systems (that are redundant) to landing systems (some which are not).

That's exactly the point. They have missions that are just light enough to be landed - FH center core and that heavy GTO record for example. Adding extra mass for non mission-critical items that pushes these launches to expendable boosters is not an effective solution. Maybe there are completely different solutions such as an improved hydraulic system combined with some sort of insurance - their landing record is now as good as some of the best launchers.

[Jcc]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

[vanoord]

The landings are clearly not out of the experimental stage. You seem to not acknowledge the difference between flight critical systems (that are redundant) to landing systems (some which are not).

Did I miss Saturn V landings that exhibited redundant landing systems?

As for RTLS being stopped, I wouldn’t bet on it. Odds are SpaceX might address it before it would become an issue anyways.

While landings are no longer described as 'experimental' on the SpaceX webcasts, as far as I can recall there was always an expectation that some cores would be lost on landing.

The current success rate for landings is around 82% overall and 89% for Block 5.

If they lose 1 in 10 over the next couple of years, that shouldn't be too surprising.

Over time that figure is likely to come down as new issues are discovered, dealt with and the landing process refined.

Landing the booster(s) is not the primary mission objective, but has been a useful bonus which has turned into an important part of the business - and there's no reason to halt attempts just because one booster has made a contingency landing at sea.

[Zed_Noir]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

Since SpaceX only have the OCISLY ASDS available on the East Coast now. The ASDS have to be available and not transporting another recovered core at the time or laid up at dock for maintenance.

Consideration also have to be given to how much priority the future cores being recovered might be. Giving redundant landing opportunity to a RTLS core might cost you the availability of the ASDS until you disembarked the landed core and refitted the ASDS. So you might want to spend more effect on recovering newer cores and less on older cores.

IMO SpaceX needs at least 2 ASDS if they want at least one available for down range recovery operations every 2 weeks or so. Without even accounting for unscheduled downtime for any of the ASDS.

[edzieba]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

[Rocket Science]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

If I had to perform a cross-controlled water landing in ocean swells I doubt I would have fared better...

[SimonFD]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

Yes, late on in the descent it seemed to be attempting a landing on a specific bit of water rather than the water directly below it.
I suspect some extra code along the lines of "if we're not landing on land, don't bother veering off 10 feet to the left, 50 feet off the water" will be added.
Of course, it could have been a gust of wind which triggered the "correct for wind" algorithm. The above still applies though.

[Jim]

the engine was able to recover a wildly spinning ship .

Not that incorrect statement again

[Rocket Science]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

Yes, late on in the descent it seemed to be attempting a landing on a specific bit of water rather than the water directly below it.
I suspect some extra code along the lines of "if we're not landing on land, don't bother veering off 10 feet to the left, 50 feet off the water" will be added.
Of course, it could have been a gust of wind which triggered the "correct for wind" algorithm. The above still applies though.

I'm sure they will look at all the data and contingency. Let's not forget that the stage would be coping with wind shear at different altitudes while rapidly descending through them...

[meekGee]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

Yes, late on in the descent it seemed to be attempting a landing on a specific bit of water rather than the water directly below it.
I suspect some extra code along the lines of "if we're not landing on land, don't bother veering off 10 feet to the left, 50 feet off the water" will be added.
Of course, it could have been a gust of wind which triggered the "correct for wind" algorithm. The above still applies though.

I'm sure they will look at all the data and contingency. Let's not forget that the stage would be coping with wind shear at different altitudes while rapidly descending through them...

Yup, but the tilt at about 5 seconds before touchdown looked like a very deliberate counter-wind maneuver.


IIRC it was windy during launch, and touchdown was with very little horizontal speed.

-----
ABCD: Always Be Counting Down

[Rocket Science]

What I would like to know is how precisely did the stage hit the water target. It seems pretty clear that if it had landed on a hard surface it would have landed upright, so could it have landed on the ASDS?

From the tracking video, the stage had a decent amount of angle and sideslip on 'land'ing. We've seen that combination in the past with CRS-6, and it did not end well for the stage.

Yes, late on in the descent it seemed to be attempting a landing on a specific bit of water rather than the water directly below it.
I suspect some extra code along the lines of "if we're not landing on land, don't bother veering off 10 feet to the left, 50 feet off the water" will be added.
Of course, it could have been a gust of wind which triggered the "correct for wind" algorithm. The above still applies though.

I'm sure they will look at all the data and contingency. Let's not forget that the stage would be coping with wind shear at different altitudes while rapidly descending through them...

Yup, but the tilt at about 5 seconds before touchdown looked like a very deliberate counter-wind maneuver.


IIRC it was windy during launch, and touchdown was with very little horizontal speed.

-----
ABCD: Always Be Counting Down

Like I said they will look at "all the data"...

[edzieba]

the engine was able to recover a wildly spinning ship .

Not that incorrect statement again

For the record (using 'rotation about the Y axis' as 'Roll' as 'rotation about the stage's long axis' convention):

- A centre mounted engine can do nothing about spin purely about the rocket's own Y axis.
- A centre mounted engine can arrest rotation about an axis perpendicular to the Earth's surface (i.e. Y in world coodrinates) provided the stage's own Y axis is not aligned to this axis
- The amount of world-coord-Y-rotation the engine can arrest is proportional to the angle between the stage-centric Y axis and the world-centric Y axis (i.e. Sin(offset)). 0% at 0° offset, 100% at 90° offset or the stage perpendicular to the ground.
- Or alternatively, the inverse of the offset (Cos(offset)) is the residual 'spin' the engine will be unable to arrest.
- Given sufficient time (which a stage hurtling towards unscheduled Lithobraking does not have) atmospheric coupling can be used to convert stage-centric-Y roll to world-centric-Y roll, which the engine would then be able to arrest.

tl;dr: how much 'spin' a centre-mounted engine can arrest depends on what your specific definition of spin is, and how much time you have between atmospheric entry and "Oh no, not again".

[meekGee]

the engine was able to recover a wildly spinning ship .

Not that incorrect statement again

For the record (using 'rotation about the Y axis' as 'Roll' as 'rotation about the stage's long axis' convention):

- A centre mounted engine can do nothing about spin purely about the rocket's own Y axis.
- A centre mounted engine can arrest rotation about an axis perpendicular to the Earth's surface (i.e. Y in world coodrinates) provided the stage's own Y axis is not aligned to this axis
- The amount of world-coord-Y-rotation the engine can arrest is proportional to the angle between the stage-centric Y axis and the world-centric Y axis. 0% at 0° offset, 100% at 90° offset (i.e. Sin(offset)), or the stage perpendicular to the ground.
- Or alternatively, the inverse of the offset (Cos(offset)) is the residual 'spin' the engine will be unable to arrest.
- Given sufficient time (which a stage hurtling towards unscheduled Lithobraking does not have) atmospheric coupling can be used to convert stage-centric-Y roll to world-centric-Y roll, which the engine would then be able to arrest.

tl;dr: how much 'spin' a centre-mounted engine can arrest depends on what your specific definition of spin is, and how much time you have between atmospheric entry and "Oh no, not again".

It's pretty simple...  The effect of leg extension is easy to calculate and it will simply reduce the rate of rotation by the ratio of moments of inertia, before and after deployment.  It cannot null the rotation.

But somebody clearly did, and it wasn't the fins.

So this leaves the engine.  The majority of the work done by the engine was to righten the stage (as would be the priority of the control system) but as it did so, the moment of inertia of the stage clearly decreased - so why didn't speed of rotation increase?  (Skater analogy again)

The reason was, again, because the control system was actively taking spin about the Z axis out of the system.



-----
ABCD: Always Be Counting Down

[Rocket Science]

For conventions sake: Would apply to the entire stack including booster return.

[edzieba]

I was only talking about the engine's contribution specifically. In addition, there will be drag from the fins (actuated or not), drag from the legs (extended or retracted, they protrude from the stage), moment of inertia change from leg extension, drag from the stage itself (skin drag), drag from the stage body being not perfectly aligned to the world-centric axis of spin, the actions of the cold-gas thrusters, and the reduction in induced spin from the offset fin as the stage slowed.

No one factor is "the reason" the stage was able to (just about) arrest the induced rotation. Only some of these factors would be ones the stage's GNC could potentially control, and it is likely that only the cold-gas thrusters were ones it was actively attempting to use to arrest the spin, with the aerodynamic effects and the engine being offset likely being coincidental help rather than deliberate action (though I wouldn't put it past SpaceX to make them deliberate actions eventually, as with the 'if the stage blows up, deploy the chutes' logic for Dragon after CRS-7).

[Kansan52]

Part of the investigation will be "How did we do that!?". That landing recovery seemed doomed when the control problems began.

[Rocket Science]

Part of the investigation will be "How did we do that!?". That landing recovery seemed doomed when the control problems began.

Just to play "devil's advocate" for a second... How do we know the did not try this failure out in a sim? A great question for Chris to ask...