I've had some work in designing vessels to take external pressure.
It's fundamentally a nonlinear problem. A lot of new(and some not so new) engineers don't understand the process of solving it properly
For instance, euler buckling is the simple eigenvalue problem of solving a column in compression. But, it misses: initial defect because nothing is truly straight, material nonlinear effects if the part approaches yield strength, etc. I've seen it overestimate compressive capacity as much as 3-10 times depending on how geometry and material. You have to use more advanced techniques than just force over area.
Metallic vessels are difficult enough. You have to set up the problem properly and run nonlinear solves in the right way, and account for the right (or at least bound it conservatively) initial imperfections and analyze. And provide margin. Lots of compute time for a realistic analysis. And you probably still want to test. And you want to know your material very well, as the higher it's loaded, the more nonlinear that behavior could be too.
And that's with the simplicity of metal.
Composites are different. And that's not accounting for complexities of this problem. A lot of things get brittle at lower temperatures. Steel does, and I'm not sure about this particular material but at depth I would want to know specifics of temperatures and material behavior and select the right material for the job. Matrix and fibers themselves. Either failing would be fatal. I don't know enough here to say anything other than there are a lot of variables and I'm far enough away from the problem that it would take a lot of data to convince myself that I understood what could possibly happen.
I remember that being the mindset from industrial robotics: you're not modeling the device, you're modeling a device you can prove is weaker than the actual device.
Pity this isn't a top level comment, but you're spot on. Materials science for such designs is a really hard problem. And when you start cycling them the problems get harder still.
That’s sounds crazy complicated. In engineering is it like software design where you can get very far just being bad? Like do companies hire just crap engineers who don’t know all this stuff and call it a day? (Kind of like hiring offshore to build mvp for cheapiest crappiest quality possible)
Software failures don't usually visibly lose money, make a mess, or kill people. (Sometimes they do, see RISKS). So it's possible to have the usual sort of software project failure where something is 200% over budget and time, without anyone really waking up. But if a pressure vessel goes bang people tend to notice.
You can do that but you'll spend a lot of money breaking things. If you're in a non-safety critical field then that is a possibility. But, you really need good engineers, or to make them through these mistakes, to figure out what is really going wrong.
It is crazy complicated. In my experience, most companies have a few highly competent engineers who get funneled the work of lower level engineers with a review process. These engineers will also set up some domain-specific guardrails/guidelines that can help keep the lower level engineers from causing catastrophic issues.
Combine that with a healthy safety factor to cover the unknowable (internal material structure, etc) and you're generally safe.
There is a relative recent trend by fire fighters to replace old steel/aluminum cylinders with breathing gas to cylinders made out of carbon (300 bar). They are also used under water in diving.
The intend benefit is the reduced weight, but I wonder a bit how the failure mode might be a bit different from steel.
People deal with potential gas cylinder failure by occasionally testing them at higher than their usual pressure. You fill them with water and pressure to say 350 bar rather than 300. At least that's what they do with scuba tanks. Even if designed well they can be dropped, corroded etc.
The advantage of testing pressure vessels with water is that the failure mode is usually less bad than gas (because it doesn't expand through any breach like gas does).
Unfortunately a sub is the opposite of a pressure tank (the pressure is on the outside) so you need a pressure bottle bigger than the sub to test it in. I don't think these are common or easy to come by.
I've been in the reverse, an autoclave large enough to hold an entire spacecraft. That was already engineering on a scale that defied my imagination considerably, doing the same at the level where an entire sub could be pressurized to 400 atmospheres is engineering on a different plane. I don't think you could do this any cheaper/easier than just strapping it to a tether and dropping it overboard in a very deep part of the ocean, then winch it back in to see if it survived.
Gosh, that's one big autoclave. As you say, a sub tank would be even more of a monster bit of engineering. They do exist for more normal depths but I'm not sure if there are civilian / rentable sub test tanks for this kind of ultra deep stuff.
I guess in practice for something like the titanic sub, you could program it somehow to dive unmanned to some depth deeper than normal use and then resurface. Presumably with some radio beacon so you can find it when it does. Or a long cable connected to monitoring devices. If it didn't come back you'd have lost a sub but not lives.
Yes, that is what they do with steel and aluminum cylinders. After over pressuring them you measure how much the metal flex, and that way you get an indication of how much fatigue and thus risk there is with that cylinders.
I am unsure if they do that with carbon versions. Do they flex with over pressuring, and is that flexing indicative to failure? If they do fail, is it like an explosion where the whole thing just unwrap, or is it more like a leak?
Everest oxygen cylinders, made of titanium and kevlar usually sometimes get dropped off a cliff by accident, hit a rock and explode. Apparently they go off with quite a bang.
It's fundamentally a nonlinear problem. A lot of new(and some not so new) engineers don't understand the process of solving it properly
For instance, euler buckling is the simple eigenvalue problem of solving a column in compression. But, it misses: initial defect because nothing is truly straight, material nonlinear effects if the part approaches yield strength, etc. I've seen it overestimate compressive capacity as much as 3-10 times depending on how geometry and material. You have to use more advanced techniques than just force over area.
Metallic vessels are difficult enough. You have to set up the problem properly and run nonlinear solves in the right way, and account for the right (or at least bound it conservatively) initial imperfections and analyze. And provide margin. Lots of compute time for a realistic analysis. And you probably still want to test. And you want to know your material very well, as the higher it's loaded, the more nonlinear that behavior could be too.
And that's with the simplicity of metal.
Composites are different. And that's not accounting for complexities of this problem. A lot of things get brittle at lower temperatures. Steel does, and I'm not sure about this particular material but at depth I would want to know specifics of temperatures and material behavior and select the right material for the job. Matrix and fibers themselves. Either failing would be fatal. I don't know enough here to say anything other than there are a lot of variables and I'm far enough away from the problem that it would take a lot of data to convince myself that I understood what could possibly happen.