Thursday 13 September 2007

PSLV Crash -Failure Analysis

Business India, January 17-30, 1994

What went wrong?

The Failure Analysis Committee's report

Shivanand Kanavi

The Failure Analysis Committee headed by N. Pant, to probe the failure of the PSL V -Dl flight launched from Shriharikota, on 20 September 1993, has submitted its report. Three factors have been identified as having caused the failure.

First is the time gap between switching off the second stage engine and switching the third stage engine. Second, some retro-rockets failed to get fired after the second stage separated from the third stage, leading to an imbalance. And third, there was an error in the control software. How these three factors com­bined to plunge the Rs.45crore PSLV ­D1 into the Bay of Bengal is a revealing tale of the technological complexities of such a mission.

Theorising, modelling, and simu­lating possible scenarios started in ISRO in earnest right after the mission failed. Simultaneously, the Space Commission launched an independent probe, through the Failure Analysis Committee, ploughing through the 100,000 pages of telemetric data from the launch vehicle.

Control systems
To understand what went wrong it is essential to understand how the vehicle is controlled. The rocket is under con­trol only when the main thrust motors are firing. Any deviations from the flight path due to disturbances in the pitch, yaw or roll of the vehicle, are then set right by firing the control systems.

In the period between one stage being shut down and another being ignited there is no control. This gap is unavoidable due to the number of operations required for a clean separation of the stages.

The second and third stages are sepa­rated by exploding a ring of explosive embedded in the casing that shears the alu­minium alloy. At that time, though sepa­rated, the second and third stages will still be moving with more or less the same velocity. If there is even a small imbalance in velocities, the jettisoned second state can hit the third stage.

To prevent such a mishap retro-rockets are fitted to the jettisoned part and are fired along with the separation explosion so that the jettisoned part is slowed down slightly. It is in order to complete these operations smoothly, that a small gap is kept between shutting one stage's engine and firing the next. If liquid propellants are used, the engine can get hiccups when the fuel is over. So it is advisable to shut it off before the fuel gets exhausted.

Earlier ground simulations had given ISRO grounds to believe that if there was a three-second gap between the separation of stage two and three, nothing drastic would happen. But now, in the light of PSLV-Dl having failed, it is felt that the three-second gap may have been too large, allowing errors to multiply dangerously. Hence, the committee suggests this gap should be reduced.

The second flaw was that all the retro-­rockets did not ignite to slow down the jettisoned stage. This caused a slight imbalance in the jettisoned stage, and could have hit the third stage before ignition. It is conjectured that the firing circuits of the two retro-rockets got so dis­turbed by the explosive separation that they did not fire. Hence, further isolation and protection of these circuits from the shearing explosion has been recom­mended.

Error compounded
The third problem with PSLV-D1 was a software error caused by the 'overflow' in a control parameter. What it means is that the control software in the mother console was designed to handle variations in a par­ticular parameter, between, let us say, plus (+) or minus (-) 99.99. Now when that parameter crosses, say, -99.99 and reaches -100.00, the seven characters in ­100.00 could not be recognised and so the software ignores the bit representing the ‘-’ (minus) sign. The result was that in the flight a control command geared to correct a parameter of say -99.99 was suddenly changed by default to that required for + 100.00, while the system was actually suffering from a deviation of -100.00.

Thus the control command from the computer instead of correcting an error, actually compounded it. ISRO is debug­ging the control software to remove any other such glitches. Some believe that despite the problems created by the retro­rockets, etc, the vehicle could still have been controlled if the correct command had reached the control systems. (Incidentally, such software errors are not unusual. NASA's space shuttle mis­sion had to be grounded in 1988 when similar software errors were found and all the five on-board computers had to be debugged.)

The rocket motors for the second flight, PSLV-D2, in 1994, are under con­struction. The corrections required will not lead to any major design changes. In fact, but for this mishap in the separation of stage two and three, all other systems (including many new technologies) have worked remarkably well. Thus, despite the truism in space flights anything less than 100 per cent success is a failure', the PSLV-D1 flight is considered a 90 per cent success.

1 comment:

Anonymous said...

Thou iam reading your post quite lately...the causes of failure are very well explained. I was in 7th grade when i heard about PSLV failure on radio.