Hacker News Reader

Let's put it another way.

Suppose you've got a gearbox, with a shaft in it, and a stack of parts on the shaft. The shaft with its parts must fit in the gearbox. The gearbox has a max and min size, and each part has a max and min size. The engineer divides up the tolerances so that if they are all at max size, and the gearbox housing is at min size, the parts will fit.

Now, if a gear comes back from the manufacturer larger than its max size, but enough of the other parts are not near their max size, the stack will still fit and it's ok. (But engineering still has to approve the fit.)

This is not in any way a crisis. It simply costs extra money because the out-of-spec part gets bounced to engineering for review.

Yes, I did these exact calculations on the 757 stab trim gearbox.

TIL Walter Bright is actually an Aerospace Engineer.

I had always assumed that (as a compiler guru) you had a CS background. :)

> Bright is the son of the air force pilot Charles D. Bright.[4][5] Bright graduated from Caltech in 1979 with a Bachelor of Science in aerospace engineering.[6]

* https://en.wikipedia.org/wiki/Walter_Bright

I'm guessing that the CS stuff is more intellectually challenging.

Why would you guess that?

If you have friends in mechanical engineering, ask them to join you on a visit day at their workplace or any other place with engineering. That will change your mind.

When I was at Stevens everyone piled into CompEng and EE, kind of looking down on ME (and CS actually). To this day my ME friends are some of the smartest, coolest people I know.

Yep same here. Also people doing ME ah a hobby (so I guess not engineering in the legal sense)

Yeah, you need ME to do any useful things in the real world. Like try repeatedly making the same robot without any MEs on your team.

The best thing about being an ME at Boeing is working with the old guys who knew what they were doing.

I don't know. Look at self-driving cars. The mechanics is trivial compared to the CS stuff.

CS has interesting things going on but I doubt it's more intellectually challenging than designing airplanes and spaceships... The expression was never "this is like computer science" for a good reason.

So often I come to HN and just read random comments from random people with out realizing that I'm listening to ACTUAL experts. It is rather humbling to know I get to talk to awesome people.

Here is Walter's wikipedia: https://en.m.wikipedia.org/wiki/Walter_Bright

He went to Caltech, studied Aerospace Engineering, wrote many books, has given lots of presentations

For my own notoriety, I am a large clam that has slaved away at data for 10 years in Financial Services. I'm about to quit my job for a TOTALLY AWESOME job somewhere else

It seems like the manufacturers of these parts aren't producing in bulk, why aren't the tolerances exact and have a min/max? Are they not able to measure the parts out of the factory better? Is the measuring equipment on their end and the gearboxes company difference? Or are these like nanometer tolerances which we just can't manufacture consistently?

> why aren't the tolerances exact and have a min/max?

Because time and money are not infinite and +/-0.005 is a hell of a lot cheaper than +/-0.0005.

This is an assembly of stamped and sheared aluminum and composites. Machine shop level accuracy gets real expensive real quick with those construction methods and this isn't intentional defense industry pork so the cost does need to be kept under control.

Here's an example for something as simple as a hobbyist using a metal lathe in their garage:

For a rough part, you can measure how much material you need to take off, do it in a few passes without stopping to remeasure, and you're within 5 thousandths of an inch or so.

For a shaft that is going to slide or rotate in a bore without too much force, you need to be within a thousandth or two, so for the last pass you will stop, measure carefully, and only take a little bit of material off at a time so that the cutting forces are low and there is less deflection in the tool to throw off the size of the part.

For a press fit (two parts are sized precisely enough that they can be pressed together with a hydraulic press and then never come apart), you need tolerances in the tenths of a thousandth of an inch. For this, you'll dust off your special expensive micrometer, and it's important to let the part cool before measuring, because the heat from machining can cause the part to expand and cause you to take off too much material, which would ruin the part.

So even for the same person on the same machine with the same material, the effort (and therefore cost, if you're doing this commercially) can vary quite a bit based on the tolerance needed.

I don't know aerospace, but I imagine the required tolerances require vary from "looks good from here" (seats in the cabin) to "must ride on a film of oil this many microns thick when spinning at 10,000 rpm." And the manufacturing processes can be anything from "intern with a saw" to "specialized metrology lab with most expensive machines in the world" depending on what engineering specifies.

I don't know much about machining but this was a really cool explanation for the significant/difficulties of tolerances, thanks for explaining mrfredward

I forgot the details, but manufacturing costs go up by an order of magnitude with a halving of the tolerance, or something like that.

Gears are usually slightly smaller than design size for clearance sake (and to make cutting the teeth easier). I imagine this tolerance disappears to almost zero in aerospace, though.

If the outer diameter of a gear tooth is too large, it can bind with the root of the matching gear, and this can cause rapid wear, and an unplanned disassembly at some point very early in the service life.

I can say that the aviation gears are far better cut and fitted than the ones you'll find in a car transmission! I once brought to work some gears from my car tranny, and the other engineers were horrified at them.

My understanding is that it's not about airflow, but about unplanned for localized stresses rather than having them distributed evenly. This means the joint does not meet limit load requirements. While it's not so bad it could immediately structurally fail under expected loads, over time the stress concentration could weaken the structure at those points, leading to premature failure or more likely, significantly shorter service life than expected.

The FAA said it wasn't a safety issue. Not meeting limit load requirements is a safety issue. Though I wouldn't be surprised if the CNN reporter got it all bollixed up. I wouldn't take anything seriously from CNN.

If it is the issues from last year, this article has a lot more detail: https://seekingalpha.com/article/4374235-boeing-dreamliner-m... As I understood it, there are two issues, neither of which is a problem by itself, but if they occur simultaneously, it would not meet limit load requirements. But maybe CNN is reporting on another new issue - hard to tell.

It looks like this isn't exactly the issues from last year, but is a new related issue. This Seattle Times article has slightly more detail: https://www.seattletimes.com/business/boeing-aerospace/a-new...

Last year's issue related to joints around the rear pressure bulkhead. This article indicates the new issue is a similar problem related to the forward pressure bulkhead.

I think this (the FAA involvement) stems from the previous rear fuselage issues, which was a safety issue, and the FAA is still directly overseeing that issue. It sounds like the FAA have not restored Boeing's delegated authority for final airworthiness sign-off on new 787s due to not having accepting a new inspection plan for that issue yet.

That would would explain why the FAA remains involved on an issue it does not consider to be a safety concern, especially as it implies there may be even more quality control issues than previously realized.

Perhaps “no immediate threat to flight safety” just means that no aircraft with the defect have been delivered or are in service?

The phrase "Boeing said 787 planes already in service do not need to be grounded" kind-of implies that there are aircraft with this defect in service, but it need not necessarily do so.

> There are always going to be leaks.

As this seems to be something that's given quite a bit of thought, do you have an idea on what the maximum diameter hole through the fuselage to safely maintain sufficient cabin pressure during cruise flight at, say, PL390? More specifically, what's the maximum rate of pressurization that the cabin pressurization system is designed to support?

I have no idea, as I didn't work with the tin benders (structures guys).

But I do know that there can be a hole the size of a quarter and nobody will notice a pressure drop. You're not going to knock an airliner down by punching holes in the fuselage, either (some people think a bullet hole will cause it to pop like a balloon, nothing of the sort will happen).