How to Predict the Tolerance for Manufacturing Parts?

Hey Del....if i am paying for the material to make that spoon it will not be + or - it will just be -

Anthony @ ALNO

www.alnoproductservices.com.au

.
A while back I was explaining the whole manufacturing / sales process to my Daughter by showing her all the costs involved in selling 'nothing'. If you had to package, label, advertize, ship etc, 1" cubes of 'nothing' the costs still add up pretty quick even when the actual part cost is zero.
In fact you could probably give away a free spoon to scoop out the 'nothing' (after all you wouldn't want to leave any in the tub)
Del

Like this company perhaps?

http://www.sealedair.com/products/protective/default.htm

milo

That's a complex question. It depends on function, fit, cost, prediction of tool wear, inspection processes, standards, etc.

Find a design draftsman with about 20 years experience and watch him for a week. Then ask him about the tolerances he put on.

to summarize from the above post, I get

1. tight tolerance is costly and the prediction should be mainly based on end use application, designing techniques, manufacturing processes and materials used.

2. practical way of predicting tolerance is also one of the best way to define tol.

3. defining tolerance is complex but should be prioritized for the final requirement

cheers

In sound practice, tolerances are set by the usable amount of variation that will still allow the part / component / assembly to function as designed within the application. In practice, because engineers really don't trust manufacturing / production types, the tolerances are tightened further. Manufacturing / production types knowing this, tended to let things drift into the fringes until the screaming started down stream.

Hence the arrival of SPC and GDT. SPC to satisfy all parties that the processes involved are under control and therefore the product is in spec, and GDT to give manufacturing / production additional leeway in producing products that still comply with form, fit, and function, even if individual features are somewhat out of spec.

Not to say that this has eliminated the senario laid out in the first paragraph.

Engineers fighting tolerance fighting production fighting tolerance. Too funny.

While working at GM tech in Pontiac, I discovered that a tolerance stack didn't work and the frame of a truck should not be able to be assembled to the suspension. However they were currently building prototype frames daily. So a gaggle of engineering folks (you know how GM likes to waste money) wandered down to the moch-up assembly line a few streets over to investigate.

After a very interesting tour of other assembly moch-ups we eventually discovered that the production people were using a de-burring tool to oversize a pair of holes at assembly. They had planned to implement this as part of their production assembly.

When questioned about why they didn't communicate the issue back to engineering the answer was simple. It just didn't occur to them.

I have run into this many times where production use drawings and engineering instructions as a "guideline".

So much for tolerances and engineering......

Geometric Dimensioning and Tolerancing (GDT) is a wonderful tool......when used correctly. When I first learned to use GDT in the '70's, I found it really helped manufacturing. When applied corectly, you really knew that if the parts met the tolerances, the assembly would work. You also knew that anything outside those tolerances really had the potential to not fit. It seems from my experiences that most designers in recent years lack even a basic undestanding of GDT, but they still try to use it! They specify datums that make no sense, call out true positions with no basic dimensions, reference datums that cannot be used for the dimensions specified, etc. I think the most frustrating is when they specify, like in a recent example, a profile tolerance of .003 inch on a NURB (non-uniform rational boundary) surface, which is formed with multiple curves, no defined radii. That means +/-.0015 inch from being theoretically perfect, on the outside of a hand-held case! Totally impractical, not easily measureable, and absolutely non-sensical.

I'm not so sure I agree with "prediction" of tolerances. Tolerances are a function of a requirement or an allowance for process. It is to ensure parts fit or function as designed. It is also the less understood communication of design intent.

To develop tolerances a few things that help:

-an understanding of production process (machining v.s. water-jet v.s. flame cut) and their resulting finishes and accuracy.

-fits and clearances required to allow assembly or function (braze gaps, bearing seats, shaft fits)

-product or part use or function (like Del's plastic spoon v.s. a high speed gear).

Anyone have anything else to add to this list??

Properly applied SPC practices can be predictive of what a machine / process / system can reliably produce. That is why capability bell curves are so usefull. Develope the curve and continue to narrow the tolerances down until the resulting CPK numbers are at the minimum aceptable level. This is the tolerance for that machine running that job, which can be applied towards other similar jobs.

This only works for the machine if it is not being constantly 'adjusted'. With a profusion of adjustments, you now have a tolerance for a system.

I once worked on a project that involved a hydraulic motor turning a shaft which, in turn, is supported by two bearings. After several months of getting criticized for the shaft not fitting properly, I found that the machine shop (which used calipers to check NC produced parts) was making the ID and OD of the shaft with about .060" TIR. Needless to say, this was NOT per the drawing.

Then the shop was insulted when I said they were doing it wrong.

Unclear to what it is you are manufacturing and what your preferred method is? ( stamping, moulding, fabricating ) Also quantities are a major factor?