Thursday, January 6, 2011

Fault tree analysis and reliability

I will now get deeper into my project using a method which is probably unknown to the car or any major consumer industry: First see what can go wrong, assuming the product is finished.

Thankfully the nuclear and space domains used this way, and they had to. Or else the cost of even a small issue would annihilate the finished product, and its crew.

Nasa wasn't as diligent and Challenger/Columbia resulted... which gave the doomed shuttle a reliability on par with the Kamikaze planes used by the Japanese in WWII.

I exaggerate a bit since the later had a 100% pilot kill rate but 14 dead in less than 150 flight is not something to brag about.

Don't get me wrong, the Space shuttle is an awesome feat of engineering, that was put in the most plain looking shape ever for such means that were dispensed. It is a showcase of how to make human beings travel in a shaking nuclear bomb-sized potential energy accumulator.

Story Musgrave once said how much he grew scared at every missing that the thing would indeed blow up. His quote, hard to find online but which I read in a book about space once, expressed how flying something is supposed to inspire more and more confidence as you learn about it... not the damn reverse, which is what was happening in a vessel where everything was near its boiling/explosive/rupture point in one way or another.

It is once again the scourge of the custom and low volume. Separate elements which are made to order and unique will have reliabilities hard to judge. On the opposite scale using an element produced in huge quantities will garantee its specs are always the same and its reliability will be easy to estimate inside a more complex system.

My first operational decision is make the screws used to build the mesh of the spaceship the standard for everything that needs to be screwed. Another one is use metallic tubes of exactly the same size.

It will be easy to make a fault tree analysis on a small mesh subset and scale it up. Also I will be able to optimize the whole structure using linear programming.

No comments:

Post a Comment