Frequency/Level of Tol analysis

[KENAT]

Out of interest, how frequently do people do tolerance analysis (all drawings, only for precision/complex parts etc) and how thorough are they.

I've been getting a lot of parts/drawings to check with interferences at worst case (typically a long way before worst case) and am wondering what most people do with tolerance analysis.

I admit I don't do a full analysis on every interface of every part but I at least make sure hole patterns match etc.



KENAT, probably the least qualified checker you'll ever meet...

 

[PSE]

My opinions for what they are worth.

Features critical to proper safe performance should always be analyzed for interferences or stackups.

If the component/assembly is sub-contracted out and becomes a procured item, it should be analyzed. This can save tremendous headaches down the road in the event of quality issues.

If you have confidence in manufacturing's capabilities to do certain processes with accuracy and repeatability then you might consider letting your diligence "slide" a bit. Here you start gambling with the odds of something going wrong.

Please keep in mind that it is less expensive to discover errors at the drawing level than when you have a fabricated part on hand that you now need to diposition.

Regards,

 

[Heckler]

It depends on the complexity of the assembly and what the project contract states. I typically do Worst-Case or Root Sum Square. The more complex statistical tolerance anaylsis really depends on getting information from manufacturing and inspection. I'm often greeted with the phrase "we don't make bad parts" when I ask for SPC data from our manufacturing engineers. I personally think this type of data is really helpful when designing critical parts....I really dislike throwing my designs over the fence and letting someone else figure out if it works or not.

Heckler
Sr. Mechanical Engineer
SWx 2007 SP 3.0 & Pro/E 2001
XP Pro SP2.0 P4 3.6 GHz, 1GB RAM
NVIDIA Quadro FX 1400
o
_`\(,_
(_)/ (_)

"First they ignore you, then they ridicule you, then they fight you, then you win." - Mahatma Gandhi

 

[JsTyLz]

Agreed PSE, the cost of design mistakes on the drawing board (or computer screen) are astronomically less than that after production. Problem with the world today is that the customer orders a product tommorow and expects is yesterday. Management has eyeballs full of dollar signs and agrees, all the while engineering has to shuffle design time into a small block before tooling is released for production. ahhhhh

Back to the original post:
I do it for all female, male interfaces. Problem is when someone pulls the tolerance off old prints, or vice versa forgets to they don't know what they are applying. I am sure to look for interferences when I build the asssembly model. anything close I analyze. If it is to complicated for stackups, I sometimes build the max/min models and assemble them for a quick visual check of interference.

 

[MikeHalloran]

The new designs with which I've been involved were mostly checked, or at least eyeballed, for tolerance, by experienced designers, with mostly acceptable results. When I see tight tolerances, I try to find a way to reorganize things so they're not necessary, or are localized, and I encourage my designers to take some extra time to do stuff like that.

Lots of times, I get dumped off the turnip truck into a swamp of stuff that has reccurent manufacturing problems, needs hammers or big wrenches at assembly, or fails in strange ways. Usually I find that no one has _ever_ done a detailed tolerance analysis, or designs have been adjusted in a direction that minimizes the screaming, without consideration of side effects. That's when I start measuring things, myself. I usually find stuff like blind threaded holes that are too short for the screws, o-rings crammed into a cavity that's too small, stuff like that.

A lengthy revision list is often a clue that there's a problem that goes beyond the extant part.



Mike Halloran
Pembroke Pines, FL, USA

 

[KENAT]

Thanks all, I was trying to make sure I wasn't being too demanding or something.

Someone gave me a print on Wednesday pm that was a very slight modification of an existing part that they wanted checked fairly quickly. Given it was based on an existing drawing I knew that if I wanted to change much they'd say 'well we've been making to the existing drawing with not trouble' so was fairly careful on it.

The moment I looked at it I saw what appeared to be implied coaxiality and asked a few pertinent questions.

I looked more closely at one of the parts it mated with and discovered the the CSK holes matching threaded holes in the other part had a massive potential fastener interference.

Also that two of the diameters I'd already spotted not only had coaxiality issues but at max tol had a greater diameter than the mating female feature on the other part.

I've seen similar problems on a number of other prints and was starting to wonder if expecting a basic worst case tol check on simple/obviously close features was too much to ask.



KENAT, probably the least qualified checker you'll ever meet...