A sample of one is a pretty difficult place to start when trying to predict future failure modes. A space launch vehicle, for instance, can harbor many un-thought-of failure modes within its hundreds of subsystems and components. And it takes only one surprise to fail a mission or end a life.
Sometimes you have a few “copies” of your system. For instance, the B-2 Spirit Bomber production only yielded 21 copies and not all survive to this day. That means there is a bit more data you can gather as each sortie, aircraft generation, and repair is scrutinized. When this is done well, there is a little bit better chance to detect and avoid emerging failure modes that can doom a system and its mission.
Then there are systems that might have the numbers, but not all systems are active all the time. ICBMs is the perfect example of this. Four hundred guidance systems, for instance, operating 24 hours a day, 7 days a week create a lot of data. But the stage 1 rocket propellant is only seen in its full operational mode 3 or 4 times per year in flight tests at Vandenberg AFB or special static tests laying on its side in a desert in Utah. Special tests and age surveillance are used to fill the gap in data.
And then there are systems that are operating all the time. If you have verifiable operational and maintenance data from a fleet of one thousand operating systems (Today’s commercial airliners? Tomorrow’s Google Loons?), it should be much easier to spot emerging failures in one or two items that may clue you into issues that will overwhelm your fleet in a year or two.
No matter which complex system you find yourself responsible to sustain, the Fundamental Theorem of Sustainment still directs your efforts: “An effective sustainment organization will always find ways to affordably detect threats to the system in time to correct them before the mission is impacted.” The methods described in this blog, that is, the methods used by ICBMs, stand out as the best. They had to be if they were to support the nation’s critical mission of deterrence for over half a century. And a weapon system buried and scattered across the northern tier needs a great deal of lead time to plan maintenance and mod activities. Predictions have to be accurate and made many years in advance.
Fortunately, in all the examples above, with the right assessment program, fairly strong statements can be made about the current and future reliability, availability, accuracy, loiter, or whatever readiness factors are important to your system and its mission.
For your consideration, outside of definitions intended for acquisition rules and laws, there is really no specific day where “sustainment” begins. Every launch vehicle has a first launch, every aircraft has a first flight, every complex system has a first operation that “really counts”. The designers expend a lot of effort ensuring that first time will succeed. For instance, they look closely at their suppliers, they impose quality and acceptance tests for parts and components, they perform extensive inspections throughout the assembly and final countdown, and much more. Sometimes you have a bit of history to draw from. During the Titan-Gemini Program when using a known booster like the Titan ICBM, extra care is taken with a specific vehicle, especially when the payload includes astronauts.
When a launch vehicle, or any complex system, emerges as pretty reliable, changes are made with caution and past successes help simplify the intensive examination for each launch. In this sense, even unique, supposedly “one-off” systems can find themselves entering a “sustainment phase” of their lives. And each newly-minted “sustainment organization” comes to realize that even in the case of unique, low production, systems, you still have certain subsystems, components, sub-components, and parts in the dozens, hundreds, and thousands that you can evaluate from yesterday’s and today’s data to attempt an aggregate understanding of your system’s current and future readiness factors.
Confidence can be higher and error bands narrowed for estimates of aggregate reliability across hundreds of copies. It’s just that sometimes those copies are subsystems, components, and parts. And, consider this, sometimes some of your data can come from other systems using those, or similar, pieces.
All of this is to make a critical point. Inherent in a sustainment phase of any system is the absolute need for extremely disciplined configuration management. For observations and testing to be useful, you must be able to know to what extent your fleet of x-number of systems, or your historical database of launches, or your bill of materials, are exact copies, or not.
“Well, yes sir, that part is serialized, but we haven’t been tracking them. I can’t find the bad actors. I can’t trace lot numbers, I can’t tell if our newer parts are more reliable that our older parts. I do know that some of our systems use a different part. No, I don’t know which ones. No, nobody else is tracking that component in their systems either. I can’t really tell you much at all, sir, except we seem to be having a few more failures and it might have something to do with that part.”
If you’d like to learn more, try using the search box next to any post, read one or more of my technical papers under the “Presentations and Publications” menu, leave a comment, or buy my ebooks on Amazon by searching for “Charles Vono”.