Proton-M Failure Reaction and Discussion Thread - July 2, 2013

[baldusi]

Angara, as I understand it, will use multiple boosters, which means multiple drop (and cleanup) sites compared to Proton, so the comparison is not straightforward.  Even Vostochny, as I understand it, will need drop zones on land.

 - Ed Kyle

No I think. First stage of Angara (3 or 5 version) will use multiple boosters (URMs) but in one block. So no multiple drop sites.

I've seen on this site a few graph on the cross feeding of the A5. Even showing the check valve. Is it optional, a future development or simply a proposal.

[baldusi]

Another way to look at it is that there has been one Proton-M failure per year for the past four years...

The historical rate of failure of Proton has been 88.40%, or one in 8.77 launches. And the average launch rate from 2000 to 2012 has been 8.62. I don't think this is a coincidence. As I said before, you know the reliability of a Proton. Consistency is a good characteristic.

Not when it's a consistent failure rate of 11-12%. No other launch vehicle in the world is anywhere near as accident prone.

Tell me what's the reliability of the Zenith, of the GSLV and the LM-3? You want more reliable launchers with a great track record: Atlas V, Delta IV, H-IIA/B and Ariane 5. Save the las case, none are getting any commercial orders. And Ariane 5 would love to launch 10 times per year, the fully have the capacity for that. They just can't get enough clients willing to pay for that reliability.
You see, commercial launchers are a cost. You can't have good launchers or bad launchers. You have cheap or expensive launchers for their reliability and service levels. Given that a GTO launch cost between 20% and 45% of your capital investment, and given that you can buy insurance, taking a 12% of failura chance that's insurable to lower your costs by 10% or more might be financially responsible.
When one states that the whole market is wrong, then one must think that may be there might be some sound economic reason for the market to be "wrong". If not, I'm sure you could do some arbitrage and make a fortune.

[spectre9]

Elon has tweeted.

https://twitter.com/elonmusk/status/352221585010212864

[newpylong]

That may change, as Falcon Heavy (dual- or triple-manifested) could offer Proton-class prices at much, much higher reliability (and cheaper than Angara). But until that time Proton will continue to launch and crash...

As Jim likes to point out, paper rockets are always cheaper and more reliable than their real counterparts...

Could not have said this better. This is pure conjecture at best.

[Prober]

easy JimO this is the space age or google maps age?
 
should be easy to get high res pictures by some private sat company.

Well, I want them NOW. And they already exist, from aircraft. I'll be watching the on-line Kazakh press, tomorrow's editions are now being posted. You want to wait for a future fortuitous commercial overflight, do it your way.

Let me go check ISS daylight passes. But a 200 meter crater might not be an easy target for handheld cameras.

how are you making out on this, any progress?
 
would that shock wave break glass in the space center?
 

[edkyle99]

Not when it's a consistent failure rate of 11-12%. No other launch vehicle in the world is anywhere near as accident prone.

The current standard model, Proton M/Briz M, currently has a 90% expected success rate.  Not as good as its designer's wanted, I'm sure, but still better than the following launch vehicles.

Vehicle     Successes/Tries Realzd Pred  Consc. Last     Dates    
                             Rate  Rate* Succes Fail    
================================================================ 
Proton-M/Briz-M  60    66    .91  .90      4    12/08/12 2001-
Dnepr            16    17    .94  .89     10    7/26/06  1999-
CZ-3/3A          33    36    .92  .89     23    8/18/96  1984-
PSLV             22    24    .92  .88     20    9/29/97  1993-
Soyuz 2-1a/Fregt 12    13#   .92  .87     10    5/21/09  2006- 
Zenit 3SL/DMSL   31    35    .89  .86      0    2/1/13   1999-
Pegasus (H/XL)   37    42    .88  .86     28    11/4/96  1991-
Rokot/Briz/K(M)  16    18    .89  .85      2    02/01/11 1994- 
Ariane 5ES        4     4   1.00  .80      4    None     2008-
Soyuz-U/Fregat    4     4   1.00  .83      4    None     2000-
Soyuz 2-1b/Fregat 9    10    .90  .83      3    12/23/11 2006-
Soyuz 2-1b        3     3   1.00  .80      3    None     2008- 
Minotaur 4(+)     3     3++ 1.00  .80      3    None     2010-
H-2B              3     3   1.00  .80      3    None     2009-
Zenit 2(M/SB)    30    38    .78  .77      7    9/9/98   1985-
Zenit 3F/FregatSB 2     2   1.00  .75      2    None     2011-
Shtil'            2     2   1.00  .75      2    None     1998-
Vega              2     2   1.00  .75      2    None     2012-
Strela            2     2   1.00  .75      2    None     2003-
Delta IV-H        5     6    .83  .75      5    12/21/04 2004-
Zenit 3SLB/DMSLB  4     5    .80  .71      4(B) 4/28/08  2008-
Falcon 9          4     5    .80  .71      1    10/8/12  2010-
Soyuz 2-1a        1     1#  1.00  .67      1    None     2004-
Antares           1     1   1.00  .67      1    None     2013-
Shavit(-1,-2)     6     9    .67  .64      2    9/6/04   1988-
Taurus (XL)       6     9    .67  .64      0    3/4/11   1994- 
Safir             4     7    .57  .56(C)   0    9/2/12   2008-
KSLV-1 (Angara)   1     3    .33  .40      1    6/10/10  2009-
GSLV              2     7    .29  .33      0    12/25/10 2001-
Volna             0     1    .00  .33      0    6/21/05  2005-
Unha (TD-2)       1     4%   .25  .33      1    4/12/12  2006-
Proton-M/DM-03    0     2    .00  .25      0    7/2/13   2010-

It has a good chance to improve its rank, since most of its failures involved its relatively recent upper stage.  Proton K/DM-2M had a 93% expected success at the end of its service, better than its contemporaries Ariane 5G and Titan 4.

 - Ed Kyle

[mdo]

Anyone claiming a vehicle cannot reach the city of Baikonur within the 45 second FTS delay is forgetting that the vehicle continues flying after the engine thrust is terminated.

Let's throw in some numbers based on the Proton Launch System Mission Planner’s Guide. It states that Qmax occurs at T+66 s with v = 465 m/s. Therefore, an engine shut-down at the given T+42 s mark means that the speed is <=300 m/s. At an angle of 45 deg (v_hori = v_verti = 210 m/s) that gives only about 60 s of ballistic flight time before hitting the ground and the respective impact radius is <=17 km. That is the distance covered during powered + ballistic flight. The city of Baikonur is clearly outside this radius.

[simonbp]

The point is that Proton-M's failure rate over the past few years has been closer to 86%, which puts it worse than every rocket with more than 10 launches other than Zenit 2.

It's not a problem of design, but quality control. Same reason the failure rate for all other Russian/Ukrainian rockets has jumped in the past few years. But higher quality control would mean higher prices, and low-cost is only thing the Russian really have going for themselves right now.

[edkyle99]

The point is that Proton-M's failure rate over the past few years has been closer to 86%, which puts it worse than every rocket with more than 10 launches other than Zenit 2.

It's not a problem of design, but quality control. Same reason the failure rate for all other Russian/Ukrainian rockets has jumped in the past few years. But higher quality control would mean higher prices, and low-cost is only thing the Russian really have going for themselves right now.

Here's a chart showing the long view.  There have one or two more failures than might be expected during the past five years, but it is too soon to say that Proton's all time failure rate is shifting. 

 - Ed Kyle

[Danderman]

Anyone claiming a vehicle cannot reach the city of Baikonur within the 45 second FTS delay is forgetting that the vehicle continues flying after the engine thrust is terminated.

Let's throw in some numbers based on the Proton Launch System Mission Planner’s Guide. It states that Qmax occurs at T+66 s with v = 465 m/s. Therefore, an engine shut-down at the given T+42 s mark means that the speed is <=300 m/s. At an angle of 45 deg (v_hori = v_verti = 210 m/s) that gives only about 60 s of ballistic flight time before hitting the ground and the respective impact radius is <=17 km. That is the distance covered during powered + ballistic flight. The city of Baikonur is clearly outside this radius.

There is some truthiness in your statement above, as I forgot that the Proton first stage is underpowered. However, there still would be 102 seconds of flight time under your assumptions above and that pretty much gets the stack into Baikonur in a worst case scenario (~ 30 kilometers flight path).

[Danderman]

How about examining the success rates of Russian Federal launches vs ILS launches?

Does ILS management have a positive impact on successful launch probabilities?

[mdo]

Anyone claiming a vehicle cannot reach the city of Baikonur within the 45 second FTS delay is forgetting that the vehicle continues flying after the engine thrust is terminated.

Let's throw in some numbers based on the Proton Launch System Mission Planner’s Guide. It states that Qmax occurs at T+66 s with v = 465 m/s. Therefore, an engine shut-down at the given T+42 s mark means that the speed is <=300 m/s. At an angle of 45 deg (v_hori = v_verti = 210 m/s) that gives only about 60 s of ballistic flight time before hitting the ground and the respective impact radius is <=17 km. That is the distance covered during powered + ballistic flight. The city of Baikonur is clearly outside this radius.

There is some truthiness in your statement above, as I forgot that the Proton first stage is underpowered. However, there still would be 102 seconds of flight time under your assumptions above and that pretty much gets the stack into Baikonur in a worst case scenario (~ 30 kilometers flight path).

Again, 17 km is the distance covered during powered PLUS ballistic flight phases. And that doesn't even take drag into account. So 17 is conservative, an actual range of 12 - 15 km is more likely.

[Remes]

Elon has tweeted.

https://twitter.com/elonmusk/status/352221585010212864

from the link in Elon's tweet:

However, telemetry from the doomed mission shows that a steering mechanism on one of the engines apparently moved it into an extreme angled position

http://www.popularmechanics.com/science/space/rockets/inside-russias-spectacular-rocket-disintegration-15653112


Edit: Are the 7.5° swivel range mentioned above the nominal range? Then there is typically some reserve of 10%..20%. So if one actuators moves to the mechanical end stop it might have 8.5...9°, which makes it even worse for the nominal operating TVC actuators to counteract.

[Liryc]

Hi all,
very interresting thread here !

I'm trying to find where the rocket has fallen...
I've found a website giving a location south of the launch pad (cf. attachment) but I find this quite surprising. Have any idea ?
(there are absolutely no photos of the crash site on the internet atm..)

Cyril

[Remes]

If it is software issue how would you end up with it producing that kind of result in a launcher?

Quite easy, I would say.

There are a few dozen closed control loops on the rocket. All go from
physical system->sensor->"computer"->actuator->back to physical system.

I put "computer" in quotation marks because the computer doesn't have to be a electronic computation device. It can be also mechanical, hydraulic, etc. E.g. the mechanical control loop on the Space Shuttle TVC are mechanical, part of the damping is hydraulic.

There are loops closed on the TVC actuators (most likely position, velocity and acceleration, perhaps jerk), there is the state of the overall thrust vector which is controlled (2 degree of freedom for thrust direction + most likely derivatives of it + thrust itself). There are control loops for the direction (2 degree of freedom + derivatives + velocity). There are other systems like fuel/oxidizer valves, the state of the pumps, tank pressurization, ... some of it are controlled open loop, some closed.

If there is such an abnormal situation, that loops reach their maximum actuation values, you can get results as we observed. Some sensors might not work as intended (e.g. the IMU might have left their nominal operation range).

Many sensor signals are not usable as they are and must be filtered. The filtering might work good at nominal operation, but might be false under extreme conditions. (There was a case with a US launcher, where paramters for a filter where typed in with errors. Leading the system to believe, that there is no movement and leading to extreme actuator value requests. I think it was somewhere in the end of the 90s).

If extremes are detected by the system, the system might try to reset the control loops or try do switch to alternative (more robust but less performant) control schemes.

Then there is the point about damping. Many controlled systems tend to oscillate if you excite them. Exciting might be done through disturbance forces or through required actuation forces. In order to avoid oscillations an artificial damping is introduced. That is a science on its own, as to much damping gets your system instable. E.g. everything with hydraulic tends to oscillate (as the hydraulic fluid and the moved mass create a spring/mass system). The rocket itself tends to oscillate, as there is nearly no friction in roll/pitch/yaw. Especially in abnormal situations this damping can be very tricky.

And finally: As it is a system of many many closed loops, many of them coupled/interacting/... the error can occurr at any place, and it will spread throughout the whole system if not compensated. So it doesn't make to much sense without telemetry to try to find the root cause. It might be software, it might be mechanics, it might be sensors, ... in the end it is the overall system which failed. But I understand, that speculating makes fun.

[Liryc]

I'm trying to find where the rocket has fallen...

Trying to find the crash zone, I was wondering where the guys were standing on this video and I found something weird....



If you take a close look on the fist images of the video (Attachment 1) you can see the service tower in front of the launcher and it is clear the tower is seen from the back.
-> This would mean the cameraman is standing on the South side of a line going South-East / North-West (Attachment 2).

Then, you can see the second launch tower on the right side.
-> This would mean the cameraman is standing on the northern part of the previous zone.

Problem: The pad on the north (81/23) is retired and the only pad used on area 81 is the southern one, 81/24....

If we flip the image horizontally, we get Attachment 3 and then we get something easier to understand (for me at least) and a view site located right at the south of the pad, close to the crash area spotted above in my previous message.

This could mean :
- the crash site could be located North-East of the pad
- the video has been flipped horizontally... (why??)

[Remes]

http://microsat.sm.bmstu.ru/e-library/Launch/Proton.pdf

On page S-5 (page number 58) is a picture of the rd-253 without protective cover. I think the remove before flight article in the upper right point (cylindrical) is where the TVC is assembled. So it would be a tangential movement.

Edit: If I make a simple calculation (and if I'm not wrong), then I don't see a way how to compensate for one TVC going to end stop:

Picture one shows the nozzles and the deflection direction. Let us assume, nozzle 5 goes into X+. Then Nozzle 2 must go into X+ too in order to compensate roll.

No we have to compensate for the extreme forces into "X+" by generating "X-" forces with nozzles 1, 3, 4, 6. They would deflect into the direction of the red arrows.

In the second picture I calculate the forces for Nozzle 6. Nozzle 6 is at 330°, so it can produce thrust into 330°+90° and 330°-90°. We assume that the Force F is generated by TVC. We split up F into FX and FY which is FX= F cos60° and FY=F sin60°.

We can ignore the Y-Forces, as nozzles 2 and 6 doesn't generate Y-Forces, and the y-components of 1/3/4/6 cancel out. We end up with

FresX = 2 * F - F * 4 * cos 60°

(2*F for nozzles 2 and 5
4 * F * cos 60° for nozzles 1, 3, 4, 6.)

cos60° is 0.5, so the resulting forces are 0.

So theoretically the forces of one fully deflected nozzle could be compensated, but:
1) We don't have any margin any more. If there is a small disturbance on our system we don't have any forces to counteract, as all nozzles are deflected
2) As I said, if one nozzles goes to end stop, and all other nozzles are limited to their nominal value (which is end stop minus a few percent) the above calculation doesn't work
3) The system must be able to recognize this abnormal situation (i.e. Move Nozzle 2 to a extreme in order to compensate defective nozzle 5)
4) As all nozzles are deflected by 7.5°, we are loosing thrust in flight direction. The mission might be lost anyway.

[R7]

4) As all nozzles are deflected by 7.5°, we are loosing thrust in flight direction. The mission might be lost anyway.

Aren't there two other option, both involving 4 nozzle (including the broken one). In former the reduced thrust in Z axis is greater (2 + 4*cos(7.5))/6 = 0.994 SWAG it's still enough to do the job?

In the latter nozzles 1 & 4 don't need to go full swing but asin(sin(30)*sin(7.5)) = ~3.7 deg. That's no quite accurate because distance from nozzles 1&4 and 5&6 to the CG is different, and there's still the thrust imbalance from reduced thrust in Z-axis from 5&6 to cancel, but angle for that is smaller because 1&4 has longer moment arm to CG to do it.

edit: added third, which is variant of second. If I imagined the vectors correctly both second and third solution force the rocket to slip slight in the sideway, or in other words fly with AoA. Is it too much for the rocket to handle, beats me.

Of course control logic to continue flight changes dramatically, very theoretical at best.

[LouScheffer]

http://microsat.sm.bmstu.ru/e-library/Launch/Proton.pdf

On page S-5 (page number 58) is a picture of the rd-253 without protective cover. I think the remove before flight article in the upper right point (cylindrical) is where the TVC is assembled. So it would be a tangential movement.

Edit: If I make a simple calculation (and if I'm not wrong), then I don't see a way how to compensate for one TVC going to end stop:

Picture one shows the nozzles and the deflection direction. Let us assume, nozzle 5 goes into X+. Then Nozzle 2 must go into X+ too in order to compensate roll.

I don't think this is right.   You do not need to use only #2 to compensate roll.  Instead move each of the 5 others 1.5 degrees in the opposite roll direction.  This stops the roll.  Then you need to stop the translation, which can also be done by some combination of the remaining 5.  I'm on the road and cannot do the details, but I'm sure both roll and translate can be cancelled without driving any of the remaining 5 engines to the limits.

As far as the total thrust loss, it's at most cos(7.5 degrees).  Again, I'm on the road, but if we call this 0.1 radians, then the loss is roughly 0.1^2/2, or 0.5 percent.  This is probably within margins, if nothing else goes wrong.

[Prober]

How about examining the success rates of Russian Federal launches vs ILS launches?

Does ILS management have a positive impact on successful launch probabilities?

How about Proton fully Russian owned and operated vs Partnered with LM?