Control systems engineers tell us that, in order for a system to be controllable, it must be observable. This statement is true and compelling in its simplicity. However, just a little thought reveals that there are nuances of observability.
If you want to hook a servo to the throttle of your car in order to control its speed on the freeway, the first thing you’d like to observe would be your car’s speed. And the first thing you must admit is you don’t really know your car’s true speed. Perhaps your best estimate would be from your GPS device. Grabbing that number would require too much complicated engineering. Monitoring the speedometer needle could be another way, but hooking some kind of device on that slender needle is problematic. A camera to track the needle is equally silly. Tying into your engine’s tachometer will only give you your engine’s speed. I am sure you can think of other potential solutions that may or may not give you your car’s speed.
Devices you could buy as kits in the 1970’s involved placing magnets on your drive shaft and mounting a sensor near-by. The kit worked pretty good since the drive shaft is easily accessible and its speed is directly related to the drive wheels’ speed, thus the speed of the car.
This kind of control system, seemingly simple, can get pretty fancy. Really good factory-installed car speed controls have an estimate of the vehicle dynamics built into the controlling algorithm. Some will have an optimization algorithm to save fuel or be more responsive or give a smoother ride on hills. But ultimately, it depends mostly on observation of a key factor: speed.
In our paradigm, in order for a system to be sustained it must be adequately (and affordably) observed. In this case, readiness factors, not speed, are the key metrics we are attempting to observe. The outcome is not speed control, but risk mitigations employed lead time away.
So, for sustainment, where in the system is reliability observed? What needle will we look at to observe availability? Just as in the car speed control, these factors are not directly observable. And sometimes, like with a speedometer needle, observation can influence the reading. And sometimes, like with a tachometer reading, we realize we are looking at the wrong thing.
An organization can easily expend millions on testing and tracking a voltage only to find
- the batteries were not representative of the deployed system batteries
- the data got scrambled when it was recorded
- current was the real parameter that mattered
- the test director failed to enforce test discipline
- testing equipment was switched out without ensuring consistency
- everything was probably done right, but paperwork is missing to prove it
- you tested way too many batteries to get the data you needed
This is why an entire volume is dedicated to this part of the complex system sustainment management model.