How to persuade engineers to abandon silly tolerances? 3

[616Nickel]

I'm the Senior Manufacturing Engineer for a mid-size electronics manufacturer, and our Mechanical Design Engineers have recently started to put GD&T tolerances in all sorts of strange places:

- 0.02" perpendicularity specs on PEM studs up to 1" long.
- 0.005"-0.015" flatness specs on machined 14" x 10.5" aluminum bezels.
- 0.010" parallelism specs between the front and rear faces of a bezel.

Now, these bezels are assembled to a sheet-metal PCB mounting plate and enclosure, and the entire thing mounts into a console. We have never had problems with poor flatness on our parts, and there is simply no reason to ask our machine shops to hold parallelism with a surface that is effectively just waving in the air. As for the PEM stud perpendicularity spec, I am told that this is to allow us to reject parts if they are damaged in shipping, and to ensure proper packing by our vendors!

Now, I used to be a Mechanical Design Engineer not so long ago, so I didn't come down in the last shower. But the trouble is that these engineers are largely insulated from the cost implications of these specs. Plus, they're quite jealous of their territory, and it's notoriously difficult to persuade them to change their drawings. So what I think I need to do is demonstrate in a pretty unambiguous way the cost impacts of adding these kind of superfluous tolerances. Are there any good sources for analyzing the cost impacts of tighter tolerances, or is it really just a matter of asking our vendors for case-by-case estimates?

 

[MikeHalloran]

That discussion would go better if you had a pile of perfectly good 'scrap' parts to put on said Heavy Hitter's desk.



Mike Halloran
Pembroke Pines, FL, USA

 

[616Nickel]

So, if it's _just_ a tin shop, one has to wonder what GD&T brings to the party, and why the engineers started using GD&T.
You won't likely change the engineers' minds; you have to find and re-educate whichever Heavy Hitter got talked into it.

I'd never thought of it that way, but "tin shop" just about covers it. The parts in question are machined bezels, but everything else is sheet, and nothing is high-precision. It's actually a little painful: I designed this stuff when I was still pretty junior and had no experience in sheet metal, so it makes me wince to see that these frankly rather poor designs just keep getting repeated over and over with slight variances.

These two guys are the only mechanical design engineers in the company (the other mechanical engineers are in Project Management or Manufacturing). The "heavy hitters" supervising them are mild-mannered electrical and software engineers who don't tend to question mechanical design decisions, since all the really critical and interesting work is in the electronics. If their equipment sits in a box that meets all the client's requirements, they're pretty happy. On the other hand, now that I've demonstrated the cost implications of these new specs, they'll be more likely to ask questions the next time the drawings come round.

 

[gdtdave]

"Why is it so important that the cosmetic, non-functional front face of a console-mounted display is everywhere within 0.010" of the rear?"

Good question. Is parallelism of .250 too much? .500? Flatness of .375? For certain, most suppliers hate having rejectable characteristics added to drawings, so they jack the price. Or they don't know what it means, so they jack the price. Pick a value at which you'd throw the part out and give the engineers a hint. Also, start looking for suppliers that aren't either untrustworthy or ignorant.

 

[KENAT]

I've definitely seen the other side of this. Minimalist drawings with lot's of assumptions not backed up by the relevant drawing standard (if that standard was even properly invoked) such as features shown in line being related etc.

Made fine for X period...

Then, one day some parts don't fit, but they're to the letter of the print.

If they've done the calculations to justify the tolerances (seems unlikely on the bezel paralellism) then are they really wrong to specify it? Or is your view point that they aren't taking into account the slightly flexible nature of the parts - or that the shop floor guys are used to parts not quite fitting so they'll "Hammer to shape, File to fit, Paint to match"? Or is it just that you think it's unlikely to be a problem?

Posting guidelines faq731-376

http://eng-tips.com/market.cfm?

(probably not aimed specifically at you)
What is Engineering anyway: faq1088-1484​

 

[KENAT]

Oh and as to the pem studs specifically I'm not sure my first thought would be to use perpendicularity control. My first thought would be to rely on position control applied to the major diameter of the studs.

Posting guidelines faq731-376

http://eng-tips.com/market.cfm?

(probably not aimed specifically at you)
What is Engineering anyway: faq1088-1484​

 

[truckandbus]

I've seen two approaches to dealing with the problem of over-specification of tolerance.

One would be to have a parts quoted with tight tolerances and the same part quotes with open tolerances. Procnc.com has a tech newsletter from September 2010 that goes through the cost benefit of optimizing tolerance call outs. Quantifying the savings goes a long way toward inspiring people to change their habits. Setting the part cost as a design goal should also be a motivator.

The other approach is to have workshops with the vendor and the design team to go over the limitations of the vendor's tooling. This usually comes across as the design group coming to the rescue of the poor vendor rather than a critisism of the designer My company has avoided countless wholesale redesigns for 'cost reductions' simply by letting the vendor do what they are capable of instead of pushing them to meet fictional requirements

 

[duk748]

sorry to be so late to this thread but i have one for you - i recently checked a drawing of a shaft that the centerline was marked w/ a flatness datum of .03mm - go figure

 

[EMorel]

I came from working a machine and am now a design engineer. I find that a lot of designers don't have a proper understanding of two things: 1) How the parts they are designing are made in a machine, and 2) What role the part they are designing really plays in the final assembly. These two things will make or break a designer. A lot of young designers (like me) will be looking at previous designs to gather inspiration, or basically steal the design and change what they need to change. Now, not knowing the limitations of the machines exactly or how the parts play a role in the assembly, they (we) tend to over-tolerance things.

An example would be the face bezel you are talking about. Does it need to be relatively flat? Yeah sure, cosmetically, it does. Does it need to have flatness and parallel tolerances? Yeah maybe, to an extent. Do they need to be within a few thousandths? No, likely not.

I would say, bring a few of these designers into the production pits and show them how their designs are used, put together and functioning, how they are made, what tools are used, the capabilities of the machine. Its likely that they have had very little experience with these things. Maybe it would simply take SHOWING them that the face bezel is just there for looks and doesn't mate with anything that requires it to be so flat. Just a thought.

 

[nobog]

Where tolerances come fromfrom a text long, long, time ago)

“…It looked as if some of the tolerances were assigned much closer than should be necessary, and I started to try to find out how they had been fixed. To accomplish this, on every occasion when I met an engineer I asked him how he decided the tolerances in his branch of the subject; I fear I bored a great many people at this time. I got a variety of answers which sometimes explained things a bit, but often not at all, and though I discussed it with quite a number of men, many of them occupying prominent places in different branches of the profession, I came away with the impression that scarcely any of them were really interested in the subject of tolerances. To exaggerate the picture which I got as the result of my inquiry, I concluded that in designing a new machine the chief engineer drew it freehand with dimensions to the nearest inch, and sent it to the draughtsman to work out the detail to the nearest thousandth, who then gave it to his junior assistant to mark in the tolerances. Instructions were certainly always given that tolerances should be as easy as possible, but only lip service was done to them, and the junior assistant, anxious not to get himself into trouble, would, as a general rule, think of the smallest number he knew and then halve it.”

JK