Gauge R&R for standard measuring instruments- calipers, micrometers...

I agree. There are times that specifications are set tighter than necessary. In some organizations, slight widening of the tolerances are less costly or feasible than putting too much effort in correcting the machinery, tooling or gauges.

 

But when you have one organization responsible for the design and one for making the thing it becomes difficult. Theses days designers have little understanding of the manufacturing process used to make their widgets. So getting more tolerance is very difficult.

 

Yeah. That's why designers should be knowledgeable on the principles of DFM (design for manufacturability).

 

Or, you drag them there, kicking and screaming if you have to.

 

I'd also reverse calculate the tolerance required to achieve the R&R objectives.

 

We're a long way from than, Bev.

Often, the designer copy/pastes some tolerance from some similar part that was arrived at by atmospheric analysis. I've never seen intentional testing of dimensional limits during a launch myself. Well, maybe a few very key ones. But, if in a production investigation, say for some lean study or root cause analysis, we actually ANALYZE some statistics (because we also all know THAT's being done on a regular basis. Anyway, in this study, if I find that "this whole time" we've been making parts to a larger tolerance and there's no warranty or test data to indicate it has been a problem - should they then not be opened up? Or at the least scrutinized with a study demonstrating why the NEED to be so tight?

(For the discussion, we will assume safety critical things are off the table, I'm talking general tolerances ...)

 

Good discussion. My counterpoint ....

I agree. Specifications (tolerances) should be engineered. There is a large body of work that allows this, but it is not a COMPLETE body. Example: There's all kinds of work calculating tolerances for steel pins to be press fit into steel holes. But nowadays, we use a lot of different materials. There's not so much work for press fitting steel pins into magnesium or composites. Yet it's done.

So when we launch a new product, yes, we should engineer the tolerances. But today, there's tremendous pressure to get new designs to market. The consuming public has an attentions span of seconds these days. Consequently, and hopefully with a DFMEA process, we pick and choose the critical tolerances to be investigated and engineered. This isn't every one ...

The product goes into production and we start making components. And hopefully, SPC is done with due diligence. Reality is (hopefully via a PFMEA) that a lot of attention is given to key characteristics. Other dimensions - not so much. Fast forward a year after start of regular production and we have an assembly issue of some relatively innocuous feature. And we can't make parts. We start the investigation and find that we aren't capable, and never really were. But ... we have a years worth of production that worked just fine.

That IS testing and engineering to some degree. It's a lot less controlled than a design validation test, but it has a LOT more data points. Most people measure key characteristics hourly or continuously. But EVERY dimension is measured at least daily, if not every shift. (Typically). So just looking at historical measurements, that's a lot of data points.

I'm not convinced that I can't analyze that data with the knowledge that it represents a population that works and use it to adjust tolerances - either widen or shift them - and get a better product for it.

If I go back to my press fit example, I know the pins diameters from history, I know the holes diameters from history, and I know the force required to push the pin in from my assembly process. If I suddenly crack a composite part from excessive press fit where things were working before, I would argue all this data would be better than calling the engineer and having him say "Increase the lower limit of the hole" to reduce the risk of cracking. To me, that's the lazy answer. Typically that later in production answer isn't arrived at by testing. The engineer says "meh, just move it 5 microns". Which reduces the tolerance overall by 5 microns ...

 

My comment wasn't intended to imply that you set tolerances to pass a GRR but reverse calculating the R&R to determine what gages are appropriate for what tolerances. (Just another bit of information that can be gleaned from the study)

 

We're diddlers. We like to poke a system from different directions and see what happens. Sometimes it even WORKS !!!