Hacker News Reader

Comment by taneq | Hacker News Reader

That's why you take your maximum design stress and double it, right? /s

I don't know what happened here, but if I remember correctly, in rockets, safety factor is typically smaller than 2 or the design would be too heavy. But I guess the safety factor for pressure testing is 4.

Source: article 6.8.2.2 of this document; https://www.faa.gov/about/office_org/headquarters_offices/as...

Edit: I didn't see the "/s" before writing this comment.

> in rockets, safety factor is typically smaller than 2 or the design would be too heavy

Safety factors feel like p-values. Why .05? Because. What if it's not good enough? We'll move the goal post.

They are, somewhat. You can theoretically make a safety factor 1.0 if you're really absolutely certain of the input criteria.

Safety factor is really just a (somewhat blunt) method of managing risk. It's just admitting that we don't know the true load spectra that a design is exposed to, so we take what we believe is the max load and then slap a multiplier on it to manage how much uncertainty we expect, be it from estimates of that load, or dynamic factors, etc. etc.

We then also consider the consequence (since risk = likelihood x consequence). If the consequence of a failure is that our balsa-wood model bridge falls over, we shrug and keep it low. If it's that our pressure vessel undergoese a BLEVE [1] in the middle of a population centre then you jack it right the hell up.

There's nothing that specifically requires a safety factor. We could spend millions of dollars and thousands of hours understanding exactly the load spectra a design experiences, but that may be prohibitively expensive in the case of designing a bridge, so we instead accommodate more risk by overdesigning the item. In a space application where every kg of launch mass represents big $$$, then spending that extra time and money to understand the load specifics in more detail makes sense.

[1]: https://en.wikipedia.org/wiki/BLEVE

Safety factors also mitigate stupidity (your bridge is rated for load X, but someone drives 1.1 X through it because they're in a hurry) and cascade failures (a component upstream of yours fails and sends more load down your way).

Those both sound like kinds of ways the real maximum load could be higher than what you've called your maximum load

Only as long as your "real maximum load" != "announced maximum load". For the type of stupidity mitigation I mentioned, you want the real maximum load to be greater than the max load you announce to the customer/users. For mitigating cascade failures, you want the real maximum load to be greater than the maximum load value meant to be used internally by people designing other components in the system (though here, a smarter way would be to do a system-wide analysis of load flows to prevent cascade failures under user stupidity; however, here we're rapidly approaching the point at which I just talk out of my ass, having no real expertise on the topic).

This is getting into the "really absolutely certain of the input criteria" category your initial parent was talking about

There's also some benefit in throwing a moderate safety factor on something if it doesn't actually compromise the intended form or function of the component, just as future insurance. For example, if you design a bit of equipment to take a certain load, but there's no size or mass constraints on it, and you can throw a 1.5x safety factor on, then it's not necessarily a bad thing. Even if your 'true' load as measured over several years of service only ends up being 1.2x, then the extra 0.3 or so can come in useful if a future engineer has to make modifications to the function of the device, or if they are asked to evaluate a life extension, etc.

In less words, I'm eternally thankful that I work in an era where much of the older equipment I used was designed in an era of slide rule. This means there's a bit of extra 'meat' in the designs which often means that when I do a more precise computational analysis, I can deal with 10% material loss through corrosion, or extend out the life by some period of time because it's not been designed precisely to the material limits.

> I'm eternally thankful that I work in an era where much of the older equipment I used was designed in an era of slide rule.

Yes, I feel this way too. The knowledge and tooling of modern engineers is amazing, and can do miracles when applied to tasks that seemed impossible just a century ago. But more often than not, it's used to pinch pennies from old designs. This makes me wish for infrastructure projects to be forced to be designed with slide rules again - the less precise your determination of maximum load is, the less the beancounters can "optimize" it.

This is doubly sad in our era, where matter is cheap and labor is expensive - adding extra safety margin can be almost free, but it's an easy target for cost reduction.

It's more like a re-usability factor in mechanical design. Higher safety factor tends to equal fewer cracked welds down the road in anything exposed to vibration and shock. I've only ever heard of one lifting fixture weldment that used a safety factor of 1.0. Every single welded joint was taped instead of painted, and peeled off and fully mag particle tested between each use.

Keep in mind that weld strength is not one single number, it is a wide range. You can inspect a weld with X-ray, UT, MT, and PT methods to narrow the range, but there are still variable heat input, soluble hydrogen content, pre-heat and post-weld heat treatment factors that ALL affect the final hardness/brittleness and ultimate strength of a welded fabrication.

Note that p<0.05 isn't the arbitrary threshold for something to be true, it's the threshold for something to be interesting. If your p-values are greater than 0.05, you're wasting your peers' time trying to show off your work.

It's been a while since I did pressure vessels but it's usually 2.5 times for proof and 3.5 for burst. Then with added safety you usually need about 4x to actually see a "planned" failure. This is most likely a construction fault.

How else do you think I keep my reputation as a miracle worker? -- Scotty

"The tank can't handle that much pressure."

"Where'd you get that idea?"

"What do you mean, where did I get that idea? It's in the impulse engine specifications."

"Regulations 42/15: 'Pressure Variances in IRC Tank Storage'?"

"Yeah."

"Forget it. I wrote it… A good engineer is always a wee bit conservative, at least on paper."

- La Forge and Scott, Star Trek TNG "Relics"

Typically the factor of safety is not that high in the aerospace industry.

Reply on news.ycombinator.com