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

The most expensive number in engineering

surjan.substack.com

Post reply on HN
Print view View original

121–130 of 176 posts

Re: The most expensive number in engineering

#121

Post by robbmorganf »

> For example, a NASA document is explicit in saying that a factor of safety only covers #1 and manufacturing tolerances and does not cover #2 - #5

Why do different fields have different definitions? Because they're different fields! Aerospace doesn't really worry about material imperfection because they do very intensive quality inspections that aren't feasible for bridges (e.g. X-rays that can't effectively be done outside). And early nuclear design use focused on "imperfect theory" because (of course) the theory at the time was somewhat uncertain.

I think this article is overly dismissive of a proven way for a whole industry to learn over time about risk management. e.g. Boeing doesn't want to share the distribution of their material strengths. But they're happy to share some safety factors that don't reveal a lot about their business but help out the whole industry.

Re: The most expensive number in engineering

#122

Post by redis_mlc »

ben_w wrote: post #58
glitchc wrote: post #49

Earlier quoted context omitted.

Seems disingenuous to lump a science fiction movie where the focus is future technologies to an action movie with exaggerated physics and lack of real damage. In action movies, bad guys die with one bullet while the hero finds his way to safety (and survives!) despite having 10 or more lethal bullet wounds. For the record, Star Wars is not science fiction nor has it ever been portrayed as such. It's very much an acti…

> For the record, Star Wars is not science fiction nor has it ever been portrayed as such. FWIW this is literally the first time I’ve encountered someone claiming that it isn’t science fiction. ( Soft SciFi, to be sure, but still: https://tvtropes.org/pmwiki/pmwiki.php/Main/MohsScaleOfScien... )

> FWIW this is literally the first time I’ve encountered someone claiming that it isn’t science fiction.

This is HN, where pendantry is their job. Featured daily.

Re: The most expensive number in engineering

#123

Post by bigger_cheese »

ErikVandeWater wrote: post #119

My intuition is that a safety factor is neither as safe, or as expensive as it seems; since people know they are working with safety factors, they start cutting corners. It might be a good idea to lie to the contractors for a project about the margin of safety so they take better care to make it right. But that may not be possible.

Professional ethics is taken very seriously by practicing engineers. Not to mention that there are very serious liability consequences for falsifying this sort of thing, taking shortcuts is a good way to end up in prison.

In my Country (Australia) as an Engineer there are potentially very serious legal consequences if I certify something or sign off on work some other engineer has done (like the contractor you suggested in example) without doing appropriate due diligence.

In my workplace an industrial plant, for example before we adjust any safety factors there is at minimum a documented risk assessment process carried out. My work justifying the change will need to be reviewed and signed off by two other engineers.

Re: The most expensive number in engineering

#124

Post by bigger_cheese »

CharlesW wrote: post #76

Is there a software equivalent of safety factor? How do you/would you calculate it?

My work uses SiL https://en.wikipedia.org/wiki/Safety_integrity_level

But I think this is more an electrical engineering thing (for control systems and interlocks and such) not sure how applicable it is to general software.

Re: The most expensive number in engineering

#125

Post by H8crilA »

This is so, so, so similar to equity risk premium / default spreads, and even the interest rates themselves. There's no particular reason why the equity risk premium should be at 4% or 6% or 8%. But we do know that if it dips too low bad shit happens. Taleb wrote quite a bit on this topic, most clearly about the specific case of realized-vs-impled volatility gap, aka the VIX is too expensive and at the same time you can't short it to make money, you'll lose money, and yes did I say it's too expensive at the same time.

And yeah, the mentioned finance "risk factors" also generally keep going down over the decades/centuries. In a similar fashion: the markets dare to use a slightly lower number, over time nothing too bad happens, more people jump the bandwagon.

The "cost" of using higher than needed "financial risk factor" is easily in the trillions per year.

Re: The most expensive number in engineering

#126

Post by whack »

codeflo wrote: post #35

The article suggests using a probabilistic failure model instead of a large a safety factor, and explains how a safety factor established in the 1930s affected the cost of the Space Shuttle. But spacecraft might be a special case, where any additional weight is so expensive, and you also expect the models to be especially accurate and manufacturing to be extremely precise. For more everyday civil engineering, I think…

> The article suggests using a probabilistic failure model instead of a large a safety factor

That wasn't my takeaway from the article, and the article discusses the same problems you've brought up. To quote:

> On the flip side, it requires more information as well. There must be good data or the result will be as arbitrary as the factor of safety, without the benefit of decades of experience... A bigger issue, and the one I think has prevented more widespread adoption, is that probabilistic design doesn’t account for fluke events -- the unknowables. If you don’t know what could happen, you obviously can’t assign that event a probability... The ideal approach might be a hybrid. Probabilistic design could be responsible for covering simplifications and a reduced safety factor could cover the unknowables. Of course, there’s no simple way to determine how much of the current factor covers simplifications, so reducing the factor would still be a risky endeavor... For my projects, I intend to embrace the empirical nature of safety factors and not think too hard about it. If a factor already exists for the area I’m exploring, I’ll use that

Re: The most expensive number in engineering

#127

Post by zmmmmm »

it's interesting that it seems to be sort of unitless.

So, if the unit of your design is linear but the unknowns operate on the square or the cube of that then the effective safety factor is much smaller.

For example I could design a 12 inch bucket to hold water and then say I'd better make the handle twice as strong in case somebody fills it past the line indicating it is full but of course that will be super-linear in its effect and my safety margin is way less than a factor of 2.

Re: The most expensive number in engineering

#128

Post by srer »

auxym wrote: post #118
wiredfool wrote: post #2

"""A non-empirical alternative to the factor of safety has been around since the 1940s, but still doesn’t have widespread adoption. I think the image below describes the concept, called probabilistic design, best. """ This is _exactly_ LRFD (Load Factor Resistance Design) which has been in the Civil Engineering building codes since the mid 80's, and became common in use the 90's when I was an Engineer (in training).…

Did you do civil? I did mechanical and barely just heard of LRFD and probalistic design. Not very common in mechanical, maybe in aerospace.

A large concern in civil design loads is rain, wind and earthquakes.

These are all probabilistic in nature, you do not generally seek to build something to be flood proof. You instead design it to survive perhaps a 1 in 100 year flood, or perhaps 1 in 1000 if it's important.

There is a trade-off being constantly made, between the price of a project and the (estimated - of course) chance of it still being standing in a year.

Re: The most expensive number in engineering

#129

Post by johnwalkr »

sobriquet9 wrote: post #13
vortico wrote: post #9

You could just square your quantity before applying the 1.5 factor. Instead of "Our shuttle is safe up to 150% of the required speed!" design for "safe up to 150% of the required kinetic energy (1/2 m v^2)". Then you only need to design up to sqrt(1.5) = 123% of the required speed. (My point is that the scaling of the importance of quantities is arbitrary so a single safety factor doesn't make sense to be applied to…

The definition is > breaking force divided by the expected force Note force being used here, not energy or speed.

It depends on the application, but usually yield force (the point where deformation starts to become permanent), not breaking force.

Re: The most expensive number in engineering

#130

Post by pixl97 »

TeMPOraL wrote: post #98
gerdesj wrote: post #95

"Safety factors started being formalized in the mid-1800s for bridge building, where factors as high as 6 were used to cover for the massive inconsistencies in the quality of early cast iron." I'm not a real Civil Engineer but I was a graduate one from 1991 - I'm now a IT bod. Anyway, Civ Eng uses established factors of safety or safety multipliers or safety factors or whatever. Structural steel uses 1.2 I recall - s…

> If I ran a Space Agency I would say something like: "Soz, we are going to make some cracking firework displays first and then we will know what to avoid." This is the modus operandi of SpaceX - they just keep tweaking their rockets and launching experimental tweaks as much as they possibly can without risking bankrupcy, as failures teach them more than successes.

When it comes to the Artemis program it's kinda funny listening to all of SpaceX competitors.

"All of SpaceX's prototypes have blown up"

Meanwhile they don't even have a design that's gone past paper.

Post reply on HN