Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why isn't Titan classed? (2019)

web.archive.org

Post reply on HN
Print view View original

271–280 of 302 posts

Re: Why isn't Titan classed? (2019)

#271

Post by kergonath »

rossjudson wrote: post #130
spencerflem wrote: post #109

Earlier quoted context omitted.

Its a big hackernews meme that NASA spends way too much on safety/beurocracy instead of innovation. Plenty of other submarines companies do the same (and get classed). I don't think this is a very insightful comparison.

I have rarely seen large groups of humans be accused of excessive consistency over time when it comes to matters of fact and science.

It’s not large groups of humans posting on a given HN story. It is entirely unsurprising to have different moods in different threads, because different people with different opinions are driven towards different stories. It’s not actually the same people saying both black and white, it’s just that HN has a polarised audience on some topics, with both sides quite vocal.

Re: Why isn't Titan classed? (2019)

#272

Post by brandall10 »

cycomanic wrote: post #266
gerdesj wrote: post #234

Earlier quoted context omitted.

You guys are tall ship masters. Your vessels are generally well understood these days but let's not forget some comparable disasters: Vasa and Mary Rose. Yes both are from a long time ago but both were based on top of a lot of good solid engineering but both pushed the envelope somewhat and perished. Vasa design basically forgot to sort out the various CoGs for a ship - it was top heavy. Mary Rose loss is a bit more…

I recommend watching the interview with Cameron. You say deep sea diving history is pretty sad, well actually there have been amazingly few incidents (and even fewer deaths) in the history of deep sea diving. And the engineering challenges are pretty well understood. They were just ignored in this case.

Cameron went so far to say the design mythology itself was inherently flawed at the principal designed of his own vehicle. That said, his vehicle wasn’t certified either, rationale being it wouldn’t carry passengers.

Re: Why isn't Titan classed? (2019)

#273

Post by kergonath »

swayvil wrote: post #29
dpedu wrote: post #3

Patent for their hull monitoring system: https://patents.google.com/patent/US11119071B1/en

With the right waveform analysis algorithm you could monitor stress, cracks and the progress thereof. Get more warning.

Which is helpful if the material is somewhat ductile. That’s what they do with aircrafts that are checked for microcracks. In a brittle material such as carbon fibre composites, the step between “a crack appears” and “the whole piece shatters” is typically measured in milliseconds. So monitoring crack propagation is not that useful. Both stresses and strains would look completely normal up until the point of failure.

It’s the same for the bond between the titanium and carbon composite bits. These things deform differently and at different rates. Stress monitoring is much more difficult than in a single piece.

Re: Why isn't Titan classed? (2019)

#274

Post by flaminHotSpeedo »

phkahler wrote: post #162
xupybd wrote: post #79

Earlier quoted context omitted.

I've heard the same argument against unit testing and strong typing.

If I recall correctly, even NASA found unit testing to be of limited value, finding a small fraction of software bugs. That's not to say "don't do it" but my stance is that the developer should do some ad-hoc testing for everything they write, just don't formalize and document it or you'll land on the wrong end of cost/benefit. I can point to some subtle logic bugs that no sane unit test would have caught.

"even NASA" implies that NASA is the objective paragon of software development. There are many ways to develop software, and while I have never worked for NASA I can virtually guarantee that their development process is not the same as some random web service.

I've found that it's better to think of unit tests as a contract. You don't write them to check your code for bugs, you write them to declare an explicit (though incomplete) contact that's checked in to source control.

Re: Why isn't Titan classed? (2019)

#275

Post by flaminHotSpeedo »

stickfigure wrote: post #73
cbzoiav wrote: post #16

> The vast majority of marine (and aviation) accidents are a result of operator error, not mechanical failure. Could that be because the vast majority of mechanical designs go through processes designed to validate them and catch flaws? They also use SpaceX as an example. SpaceX ran a large number of unmanned launches to prove out the design before putting people on board. Several of those ended in loss of the vehicl…

The first space shuttle launch was crewed. It ended up with a ~2% failure rate. The Titan made about 50 dives. So, from a reliability perspective, on par with the shuttle. I posit that half the people on HN would volunteer for a hypothetical next shuttle flight without thinking twice.

The space shuttle was known to be a borderline death trap (the platform was temperamental at best, the abort maneuvers for several segments of the launch were laughable, it was considered to be a flying brick, it couldn't do go-arounds... Etc) that never took paying passengers.

A) I expect many HN readers are familiar with how stupid and bureaucratic the shuttle program was and would never agree to fly in one

B) comparing the shuttle to titan is a pretty obvious false equivalence

Re: Why isn't Titan classed? (2019)

#276

Post by flaminHotSpeedo »

ShroudedNight wrote: post #138
_yb2s wrote: post #103

Earlier quoted context omitted.

This is a really good point... from what I've seen they estimated the strength with Finite Element Analysis in Solidworks, and went for a ~2x margin of safety. There seems to have been no attempt to experimentally determine the expected number of load cycles to failure, they only considered the theoretical case of a perfectly flawless composite with no wear and tear.

Is 2x abstractly reasonable in this context? My [probably flawed] understanding in contexts such as hoisting is that safety margins of 5x are considered a bare minimum. Even crappy "Don't use this to lift things" chain is sold with 3x margins.

Depending on the context 2x is within the realm of reasonable for things like aircraft.

But yeah for stuff like bridges you're probably looking at something closer to 20x

Re: Why isn't Titan classed? (2019)

#277

Post by noduerme »

robjan wrote: post #262
noduerme wrote: post #139

Earlier quoted context omitted.

One thing I'm curious about, as someone who knows nothing about the engineering of these things... why wouldn't anyone building something like this choose to over-pressure the sub once people were inside it and it began descending? Wouldn't 2 atmospheres inside represent a 50% reduction in the pressure differential on the hull at depth?

The internal pressure would have to be hundreds on atmospheres to offset the pressure outside. They would have to carry a lot more air at very high pressure on on the surface with more complicated and robust equipment to handle it. It would also make buoyancy management much harder. There's also the additional problem that there is a risk of gas narcosis above 3 atmospheres of pressure.

So remember "The Abyss"? (James Cameron)... the aquanauts were living in a habitat that was pressurized at the bottom of the Mariana Trench or something. Was that just done before he realized it was physically impossible for humans to live in that kind of hyperbaric situation...? Or is it possible at any pressure with the right gas mixture?

Re: Why isn't Titan classed? (2019)

#278

Post by elcomet »

gerdesj wrote: post #234
Gwypaas wrote: post #188

Earlier quoted context omitted.

As another tall ship, and for that matter in-shore commercial captain I found this quite telling. > While classing agencies are willing to pursue the certification of new and innovative designs and ideas, they often have a multi-year approval cycle due to a lack of pre-existing standards, especially, for example, in the case of many of OceanGate’s innovations, such as carbon fiber pressure vessels and a real-time (RT…

You guys are tall ship masters. Your vessels are generally well understood these days but let's not forget some comparable disasters: Vasa and Mary Rose. Yes both are from a long time ago but both were based on top of a lot of good solid engineering but both pushed the envelope somewhat and perished. Vasa design basically forgot to sort out the various CoGs for a ship - it was top heavy. Mary Rose loss is a bit more…

> It isn't the same as running up Everest or going into space. Deep sea is way, way more dangerous

That doesn't seem right to me. Deep sea and space seem equally dangerous

Re: Why isn't Titan classed? (2019)

#279

Post by bradknowles »

gerdesj wrote: post #234
Gwypaas wrote: post #188

Earlier quoted context omitted.

As another tall ship, and for that matter in-shore commercial captain I found this quite telling. > While classing agencies are willing to pursue the certification of new and innovative designs and ideas, they often have a multi-year approval cycle due to a lack of pre-existing standards, especially, for example, in the case of many of OceanGate’s innovations, such as carbon fiber pressure vessels and a real-time (RT…

You guys are tall ship masters. Your vessels are generally well understood these days but let's not forget some comparable disasters: Vasa and Mary Rose. Yes both are from a long time ago but both were based on top of a lot of good solid engineering but both pushed the envelope somewhat and perished. Vasa design basically forgot to sort out the various CoGs for a ship - it was top heavy. Mary Rose loss is a bit more…

With respect, the "DevOps attitude" is not "move fast and break things". That's the Facebook attitude.

DevOps is, in part, about using repeatable methods of automation and extensive testing throughout the software development lifecycle to try and make sure that the software you deliver in production will do exactly what it should do and nothing else, and so long as it is properly designed, should be able to withstand significant scaling stress but just getting slower and slower but not less reliable.

From what I can tell regarding the Titan incident, there was no part of a proper DevOps mindset or approach that those people adopted. They people in charge were just damn stupid cowboys out to make a fast Buck from people who were too stupid to know how much in danger they were of losing their lives.

Re: Why isn't Titan classed? (2019)

#280

Post by Mizoguchi »

yreg wrote: post #164
Mizoguchi wrote: post #143

If their $30 joystick wasn't a red flag their RTM system should had been.

What's bad about a $30 Logitech controller? It should be more reliable than what they can build on their own.

Nothing is wrong with a consumer grade, wireless, AA battery operated, gaming controller as a steering control interface for a submersible taking you down 10K ft under the ocean to maneuver around a 10 story high entangled mess of steel in pitch darkness with a flash light.

And there's absolutely nothing wrong either with the decision making process of the engineers thinking that was a good (and fun) idea.

Post reply on HN