Saturday, April 24, 2004

Operation "San Luis":

I recently read the accident investigation report for the Brazilian rocket explosion that occurred on August 22 last year, killing 21 people. The report was released in February, and is available here (in Portuguese, as a 130-page PDF). I posted a note last year on the intended launch date (Aug 25), but much of that post was dedicated to a similar launchpad accident in the Soviet Union.

I will pick up the story from Brazil here. All the pictures included below are from that report, and most are linked to larger versions so that you can see details. The report itself contains many more pictures and diagrams.

The report is a solid piece of failure analysis, done in 172 days by a group of 23 commissioners, 3 of of which were from the science community, 2 as representatives of the victim's families, along with 6 Russian specialists, and 4 other collaborators. The commission took the basic approach mandated by the Brazilian Aeronautical Accident Investigation and Prevention Center (CENIPA), guided by the triptych: man - machine - environment.

The commission looked at several different factors:
  • The Meteorological Factor;
  • The Materials Factor;
  • The Operational Factor; and
  • The Human Factor

This is obviously a cursory skip through some very serious material, but I'm not about to translate the whole report, and Babelfish won't do the trick for those of you who are not lusophones. So bear with me.

Brazil is not without experience in rocketry. The country had hosted NASA tracking stations since 1956, and had built its own tracking stations immediately after the launches of Sputnik and Explorer I. By 1964 a national rocket program had been established, and Brazil launched a U.S.-made Nike Apache in December 1965 with Brazilian personnel trained at NASA Wallops and Goddard. By 1967 a completely Brazilian-designed and -built rocket, SONDA I, was capable of taking a 4 kg payload up to 65 km. Two-hundred twenty five SONDA I's had flown when the program ended in 1977.

Building on this experience, and on Canada's Black Brant III, the SONDA II program developed a single-stage 370 kg rocket capable of taking a 20 to 70 kg payload to 100 or 50 km. Sixty-one SONDA II's were launched (no dates are given for the program - the report does note the SONDA II program was "loosely structured" and "poorly documented").

The SONDA III program, begun in 1971 and still operational today, was a large leap. This two-stage 1,590 kg rocket can boost a 150 kg payload to a 500 km apogee, and has been launched 31 times (the latest launch was 12 May 2002). However, as large a technical achievement as this apogee and payload represent, they are still not orbital, and putting a Brazilian satellite into orbit with a Brazilian rocket remained a national goal.

The SONDA IV program provided the intermediate step, with an enormously more complicated system for management of the design, testing, production, assembly and launch. The report, for whatever reasons, does not give any figures for the SONDA IV launch weight, payload, or apogee capabilities.

I gathered the following figures from elsewhere: first launch, April 28, 1989; Height: 11 m (other sources have 9,2 m); Wt. 1656 kg; payload 300 - 500 kg; apogee 1000 - 700 km.

Here's a schematic of the SONDA IV:


With an intermediate program, the VS-40, used to test the proposed 4th stage engines in a vacuum for an orbital rocket system , the Brazillian program reached the point of producing the VLS, or Veiculo Lancador de Satelites ("Satellite Launch Vehicle," proving that Brazil needs to come up with better names).

Here is a schematic of the VLS (Coifa is shroud, Propulsor is engine, others are left to the reader...) :

VLS1 schematic

The VLS is theoretically capable of putting 100 to 300 kg payload into a 250 - 1000 km circular orbit, with equatorial to polar inclinations.

Here is a mission profile, indicating the time for each stage's burn, its altitude on ignition, and the velocity:

Mission profile

Two previous launch failures with the VLS have occurred:

VLS-1 V01 on December 2, 1997 self-destructed 29 seconds into flight when the non-ignition of 1st stage engine D caused an excessive and unrecoverable angle of attack. Faulty pyrotechnic ignition systems were identified as the root cause (Portuguese report available here).

VLS-1 V02 on December 11, 1999 was destroyed by the range safety officer after 189 seconds of flight. An explosion occurred on ignition of the second stage that caused the thrid stage to separate prematurely - this stage ignited after the correct elapsed time, but of course the vehicle was by this time off-course on a ballistic trajectory. When the predicted impact point neared the edge of the secured area in the Atlantic, the RSO sent the self-destruct signal. Unexpected and uneven ignition of the forward section of the solid propellant in the second stage was identified as the cause of the explosion (Portuguese report available here).

By 2003 the VLS program had been restarted and a third launch attempt was scheduled for August 25, 2003. The various stages and components were all airlifted to the launch site, Alcantara, in the State of Maranhao by the Brazilian Air Force:

C-130 loading stages

A view of the Alcantara site, with the long engine storage and conditioning building in the center, and the vehicle assembly building in the rear.

Alcantara complex

Here is a view of on of the first stage engines in storage inside the engine building:


Here is a schematic of the assembly sequence, indicating that it was expected that an entire capaign could be carried out within 60 days. The accident occurred within the last frame, during vehicle testing, but before the (2) simulated count-downs.

Vehicle assembly sequence & timing

Here is a view of the assembly sequence, with engine C of the first stage being attached. According to the above sequence, this must be between day -43 and -32 in the assembly sequence. Note that the equipment on the top of this stage is visible here (we will return to one of these pieces below...).


Here is a view of the fourth stage being mated to the stack. According to the above sequence, this must be between day -23 and -21 in the assembly sequence.

Fourth stage assembly

A view of the completed stack with the vehicle assembly building rolled back from the launch pad during the verticality check. According to the above sequence, this must be between day -15 and -2 in the assembly sequence.

Verticality test

here is a view from the exterior of the vehicle assembly building, showing the VLS enclosed, and the platforms that allow technician access.

VAB exterior w/ VLS enclosed

Here is a schematic of the vehicle assembly building. Pay particular attention to the location of the monitoring cameras looking down on each of the access platforms.

VAB layout

This is a frame from the security tape at the instant the explosion occurred. Each quadrant is labelled, and represents the image from the respective camera. At the instant of frame capture, quadrant (and camera 2) had just refreshed, so images 3 and 4 are from a few milliseconds earlier. Note the light from the explosion showing through the gap on the access platform in camera 2. Also, on camera 4, note the yellow plastic covering over the shroud that was used to blow refrigerated and dehumidified air over the shroud and enclosed satellite.

Camera frame 1

This is the next frame. Cameras 1 and 2 have been destroyed. The light from the explosion flames is now visible in camera 3.

Camera frame 2

This is the view from the monitor camera mounted on the roof of the engine storage building, looking down the access road to the vehicle assembly building, now consumed by the fire. Stages 1,2 and 3 are alight at this point.


This is a view of the aftermath, showing what was left of the assembly building, which collapsed on itself as the flames eroded one complete side. People near enough to hear, but far enough away to survive testified that they heard the sound of several normally functioning engines. This was also supported by the wear patterns on the launch pad and the location of debris. Of course, the locking bolts had not been released so this was in essence an unintended partial static test.

VAB aftermath

Now to the culprit. This is a schematic of the detonator/initiator/ignitor at the top of the second stage engines.

The most probable cause for the unintended premature ignition of second stage engine D is thought to have been an static electrical charge built up by the cold, dry air blown over the shroud that caused a spark somewhere in the ignition system. All other electrical circuits appear to have been properly grounded, but the problem of static electricity is not common in the humid tropics, and had been overlooked.

Detonator details

A view of the detonator assembly on the top of the second stage engines (this is part of what can be seen in the assembly photo I referred to above). Engine and ingitor pressure sensors are labelled, as are the detonators and the ignitor head.

Detonator in place

A paragraph from the preface struck a particular chord for me:

Acidentes, como ensina a longa e frequentemente sofrida experiencia humana, raramente sao obras do acaso. Ao contrario, costumam ser o ultimo elo de uma cadeia de eventos, razao pela qual formouse a consciencia de que as comissoes constituidas para investiga-los nao devem ver a investigacao como um fim em si mesma, mas como um poderoso instrumento de diagnostico, por meio do qual e possivel atingir niveis de desempenho operacional mais seguros.

Accidents, as taught by the lengthy and frequently suffered human experience, are rarely chance events. On the contrary, they tend to be the last link in a chain of events, reason for which that the commissions sonstituted to investigate them should not see the investigation as a goal in itself, but as a powerful diagnostic tool, through which it is possible to attain safer levels of operation.

OSL Patch




Post a Comment

<< Home