Launch failures: an Atlas Groundhog Day
by Wayne Eleazer
|The General Dynamics and Pratt and Whitney team figured out what went wrong, and it was a real doozy, the kind of problem that made the term “rocket science” so legendary.|
It was only the second flight of the new Atlas 1 version of the venerable booster. While the Atlas Centaur had been one of the workhorses of the US space program, with the Surveyor missions to the Moon as its first real payloads, there had been some recent changes. The mad scramble to rebuild US launch capabilities following the loss of the Space Shuttle Challenger in 1986 had led to a need to upgrade the Atlas Centaur and squeeze out additional performance from it, as well as most other US ELVs. Atlas, the oldest US booster, no longer had to compete with the government-subsidized Shuttle, but instead faced a host of later vehicles offered by US competitors, as well as France, China, and soon, Russia.
The Atlas booster and sustainer stages performed well, and then it was time for the Centaur second stage to take over. The Centaur was one of the most reliable upper stages ever developed, and its RL-10 engines had a nearly flawless flight history, with no failures for well over 20 years. All that was about to change.
The Centaur separated from the Atlas and was to start its burn at T+361 sec. Then came a problem, a big one. The Centaur and its payload spun out of control, and less than a minute and a half later, at T+441 sec, Eastern Test Range Safety sent the destruct command that destroyed the vehicle.
Very quickly, examination of the telemetry data revealed that one of the Centaur’s RL-10 engines had started normally but that the other had failed to do so. There was no debris available to examine, so figuring out what went wrong would have to be based entirely on analysis and ground testing of similar hardware.
The General Dynamics and Pratt and Whitney team figured out what went wrong, and it was a real doozy, the kind of problem that made the term “rocket science” so legendary. The RL-10 was unique among US engines in that it was turbopump-fed but burned no propellant to power the pumps. Instead, some of the hydrogen fuel itself was allowed to heat up on its way to the combustion chamber, and the resultant expanding gas was harnessed to drive the turbopump. This was incredibly clever and marvelously efficient, but during start-up there was not a lot of heat available to warm the hydrogen. It did not take much to stop the start-up process. At one time boost pumps had been included in the system, but the company that made them had gone out of business and that feature had been discarded years before.
So what had stopped the start-up? The answer was astonishing. The cleaning procedure for the propellant ducts involved scrubbing them with plastic scouring pads, pads exactly like those used to clean pots and pans in the home kitchen. The investigation concluded that small particles from the pads had gotten down into the expansion bellows for the ducts and lain there, unknown, until propellant began to flow. The particles had then gotten stuck in the turbopump, and while not offering much resistance, it was enough to stop the critical start-up procedure.
The corrective action adopted was to change the procedures to require baking the propellant ducts at an elevated temperature after cleaning but before final installation on the vehicle. The plastic scouring pad particles would be vaporized by the high temperatures and thus no longer present a problem during engine start-up. Problem solved!
|But of even greater significance than the two failures was that the first failure investigation had itself failed: a remarkable situation, and one unequaled within recent memory, at least in the US.|
The next 14 months was a busy time for Atlas—almost unprecedented, in fact. There were two launches of converted ICBM Atlas E boosters from Vandenberg AFB and five from Cape Canaveral, the largest number of Atlas vehicles launched in over a decade. Even more significant than the sheer numbers of launches was the introduction of still more new versions of the Atlas Centaur. In addition to another Atlas 1, there were the first three launches of the later Atlas 2 and the first of the Atlas 2A. Then, on August 22, 1992, Atlas 1 AC-71 was launched from the Cape, carrying the Galaxy 1R commercial communications satellite.
Just like AC-70 from the year before, all looked good during the first stage burn. And just like AC-70, the Centaur stage spun out of control and the destruct signal had to be sent. And just like AC-70, one of AC-71’s RL-10 engines failed to start.
A new investigation was convened. Clearly, baking the inlet ducts to eliminate any plastic particles had not solved the problem. Perhaps it was a new problem that yielded a similar effect? Ground tests of an RL-10 engine initially failed to yield an answer.
Then one day the test team broke for lunch, came back, and duplicated the failure. It turned out that because of the lunch break they had left a valve open on the test stand. The open valve allowed atmospheric nitrogen to leak into the engine inlet and when the much colder hydrogen hit the nitrogen a plug was formed that prevented the engine from starting.
Further tests showed that one of the valves used on the flight hardware was prone to leak and allow nitrogen to enter the ducts during ascent, but that was not enough to cause the failure by itself. The extra factor that tipped the scale was due to the heavier payloads that Atlas had to carry in order to remain competitive. GD engineers had figured out a way to wring more performance out of the Centaur by increasing the cool down of the engine prior to the actual start. The air leak likely had been there for a long time, but it took the more efficient cool-down process to make it a mission killer. Sophisticated thermodynamics modeling confirmed this was the cause of both the AC-70 and AC-71 failures.
But of even greater significance than the two failures was that the first failure investigation had itself failed: a remarkable situation, and one unequaled within recent memory, at least in the US. Normally, corrective actions for launch failures tended to use a “shotgun” approach, with multiple possible causes all addressed. And it was not uncommon for failures to spur a basic review of the management processes of the organization as well. The Chinese and the Russians both managed to repeat the exact same launch vehicle failures but that kind of thing just did not happen with US boosters.
However, the AC-70 failure was different in a very basic but non-obvious way. The failure investigation was conducted by a private firm, GD, rather than led by the government. This was a first, and it produced another first, a failed investigation.
At the time of the AC-70 and AC-71 failures General Dynamics was the most vocal advocate of drastically reduced government involvement in the space launch business, even for government missions. Frustrated by the government’s disastrous “Shuttle Only” space policy prior to the loss of the Shuttle Challenger and stung by a series of competitive losses for government missions since the revocation of that policy, GD had decided that the government was the real problem, for both the company and the industry as a whole. GD did, in fact, ask for government involvement in the AC-70 failure investigation, but it could not ask too loudly—and putting the government in charge of the investigation was unthinkable. Not surprisingly, the government was somewhat more involved in the AC-71 investigation.
The bad times were not over for GD, either. There was not another Atlas launch until March 25, 1993, designated AC-74, and that one failed as well. It was not the same problem; instead of a Centaur failing to start, improper torquing of a first-stage sustainer engine set screw had caused the thrust level to degrade during flight and place the payload into a useless orbit. Ironically, AC-74 was the first of the UHF Follow On missions, launches that used an approach that emphasized reduced government involvement. The payload was procured under a “delivery to orbit” approach, in which the US Navy bought a satellite that would be delivered to its final orbit rather than to a loading dock. Failure to deliver the satellite to the required spot meant the government did not have to pay for it, but the Navy also found itself without a needed capability.
|Rocket science involves much more than just the work of rocket scientists.|
GD and the government both had to develop a new relationship based on changed responsibilities, and GD had to figure out how to take on many of the responsibilities that had been done by the government, not just let them fall by the wayside. Every other launch company faced this same challenge, and every one failed at it initially. Martin Marietta lost one of its commercial Titan 3 missions because the payload wiring harness had been hooked up wrong. A government inspector very likely would have caught that error, but since it was a commercial mission there was no such inspection. McDonnell Douglas (and later Boeing) saw its entire Delta 3 program fail commercially because the company did not undertake the kind of mission assurance program that the government had done for the earlier Deltas. And ironically, the failure that broke the program’s back involved an actual failure of the ultra-reliable RL-10 engine. Various smaller companies suffered their own versions of the AC-70/71 failures when faced with the necessity of operating independently from government control, even as they touted that independence. Even today, many companies are still struggling with the challenge.
Rocket science involves much more than just the work of rocket scientists.