Posted by m463 8 hours ago
> Right at transsonic, which is like just before engine startup, one of the chine covers ripped off. Which is something we were - we were worried about these spot weld margins on chine skin right before flight, we wouldn't have predicted the exact right place, but this cover that ripped off, was right on top of a bunch of like single-point-failure valves that must work during the landing burn. So thankfully none of those or the harnessing got damaged, but, we ripped this chine cover off over some really critical equipment right as landing burn was starting.
So, they had single-point-failure valves - which didn't fail - and they had protective cover - which failed. Overall they didn't really had single point of failure, and - while two specific failures in a row could cause an extensive failure - those two-in-a-row failures didn't materialize.
> We have a plan to address that.
Good, since it looks like the first failure was considered of a rather high probability.
Coming back to the 1 second from the abort situation. The first order approximation of Super Heavy mass distribution is relatively simple, and with such model predicting how it would behave in the air when responding to random external forces is rather possible. Terminal control, which has a goal of getting to caught by Mechazilla situation, can reasonably determine the necessary difference in forces which needs to be applied, with reasonable margins - for future random forces, for variations of performance of actuators, for model errors (keep the ship closer to the middle of trajectory so more subsequent flight control is possible). Given reasonable excess in hardware - that is, there should be multiple sensors, multiple RCS engines and e.g. duplicated intermediate systems - and assuming the hardware won't fail too fast, this first order model would allow control of last seconds of the flight with reasonable assurance. The more realistic model could be built using real data on the Super Heavy dynamics accumulated both in previous flights and in this flight. And the failures in the algorithms could probably be guarded against by a sort of digital "check list" - I understand that checking something could be error-prone, so computers should really help here.
This all is of course from the proverbial armchair of an interested and empathizing observer.
Valves are the single most difficult part of space. Valves can fail if the air was too humid for too long six months ago. They can fail if a single assembler's eyelash falls out and finds its way into the valve during assembly. They can fail if they're shaken wrong. They can fail if they're too hot or too cold or not enough of either.
If I was smart enough to solve valves I would be a insta-billionaire.
Almost every single abort is due to valves. Sometimes telemetry but almost always valves. When a countdown stops a safe question to ask isn't "what happened?" it's "which valve didn't valve right this time?"
The cover being ripped off and some malady afflicting the valves due to its absence is only one mode of failure for that single point of failure-- there are always others.