I need to vent 2

[ALK35]

I need to vent a little here.

As a company we put a lot of time and effort in the design, engineering, analysis of our products. We check to make sure the drawings are within the standards of the ASME Y14.5M. We do extensive interactive hi/lo’s on the assembly and all the components. We do all the analysis work, including FEA, ADAMS, and specialized engineering programs to optimize the design. We do all these things to ensure that we have the most optimized, cost efficient product. We pick our suppliers and they produce parts that are checked through their inspections houses, and we receive the PPAP inspection sheets claiming that the part passes all the dimensional checks. On closer review we discover, time after time, these submission reports are incorrect. Not sure if it is from a lack of knowledge, carelessness, or something a little more devious. Some things are obvious mistakes, but a lot are not so obvious. Just a warning go through your inspections reports very carefully, most are not correct.

 

[MintJulep]

Yep.

thread404-330149

 

[dgallup]

I recently received a PPAP that said every dimension was in spec. We over checked just the 7 dimensions that had critical characteristic symbols, most of them were out. You could just glance at the part and tell it was a POS. Where do these supposedly ISO 9001 certified companies get their quality certificates from, the internet?

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[ALK35]

The way the system is running right now seems to be a conflict of interest. The supplier is checking their own part through their own inspection or one that is hire by them. Every inspection report seems to have obvious errors, errors that are calculated by their own results on the report sheets. So I believe this shows a lack of knowledge of how the GD&T on the drawing is interpreted. Now for the not so obvious issues, most every inspection house that I have had communication with has the CMM set for average. Average plane, average hole, average does not mean what the assembly will see. This is a huge issue that I have been addressing for some time now. Are there any inspection standards out there? I know I am in the process of creating a set of checking standards for my company, something that will go the suppliers stating our expectations on how to check the part, the use of “averaging”, and how to communicate the results. How is this issue handled in your companies?

 

[pmarc]

I am not trying to be sarcastic or something like that, but if your problems with suppliers boil down to averages and that kind of stuff, I would say you are still lucky - there are suppliers having troubles in understanding much more fundamental GD&T things.

 

[ALK35]

pmarc,

If that is the case then how do we trust them to check the parts that they are responsible for correctly? Do you have your suppliers go through an inspection house that your company has certified? We can design and engineer a part to death but if we don't know what we are actually getting then I see this as a big problem, especially in the case of a failure within an assembly.

 

[dingy2]

I believe that the biggest problem is the method of inspection. There just isn't any common method so a supplier may measure a feature one way and the customer another and we get a difference.

Just take the measurement of flatness as an example. The supplier may use a CMM. How many points should the Operator take? The larger the number, the more confidence in the result. Where should the points be taken? Should we include the edge burr? Should we have a 3 point set up or let the CMM calculate the "best fit" from the points taken? Where should the 3 point set up (datum targets) be taken? Should we use a "scan" mode or take the readings at certain points?

The customer may use a dial indicator on a base with a granite table. Should the Operator use a spherical contact or a flat? Is the contact tight? The Operator must take a 3 point set up but where? Include the burr? When "sweeping" the surface, is there a common track to follow or just travel anywhere? Usually, the dial indicator method will achieve a larger value than the CMM. Why? More contact points.

Can you see the problem? I don't think the discrepancy is caused by someone "fudging" numbers. I think it is caused by a difference in measuring methods.

Dave D.

http://www.qmsi.ca

 

[KENAT]

Even when pushing most inspection out to vendors to self certify parts, I'd think you still need to routinely do some kind of sampling. Or at the very least first article inspection to qualify the vendor & their process or some such.

As to the method aspect, I suspect dingy has a point. I always wonder when ever anyone is basing everything on CMM if they're used an appropriate technique.

Posting guidelines faq731-376

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

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

 

[pmarc]

ALK35,
I fully understand your concern and as a matter of fact I have been through this a couple of times. My experience is that in order to have inspection done in a way that you imagined, one would have to participate in each and every measurement operation at supplier site, otherwise - no matter how well the supplier is trained in GD&T, how great inspection guideline was provided, what kind of certificate they can show and how experienced they are - there will always be a place for discrepancies. There are factors beyond customer's control. Of course such guideline, if properly written, will help to a certain extent, but I am pretty sure it will not solve all potential issues.

 

[dgallup]

In cases where a dimension is critical and difficult we have provided suppliers gages identical to the ones we use. This usually resolves discrepancies.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[bobslo]

I must point out some points regarding upper statements (and I do work in a company as TIER1 or OEM supplier)

1. when you do PPAP report, measured parts MUST be available to customer if he wants to do measurements by him selves. Confirmed PPAP parts need to be stored as an evidence.
2. Such topic as ding2 mentioned regarding flatness is a typical cause of problems => flatness require to measure a surface and surface has mathematically looking infinite number of points. How to deal with it?. In practice are two solutions => a) real surface measuring (costs??) or b)yes, you take 3D device and pick defined number of points on defined position (number and position must be tuned prior)
and
3. I am in this business for more then 20 years and what I noticed is that drawings are becoming more and more filled with dimension and notes,important dimensions, statistical measurements, etc. Please take in mind that a drawing is a message which must clearly pointed out core-issue (german guy says "schwerpunkt"). If you receive a drawing of small general-purpose part which has 130 dimension and almost all are marked as important,some of them are redundant, then ...

 

[fsincox]

Welcone to my world!

 

[fsincox]

bobslo,
An engineering drawing is more than just solely a manufacturing or inspection document of the critical features. What you want is some kind of inspection operation plan, IMHO. Go talk to a manufacturing or process planning department for that.

 

[DeanD3W]

There are no certifications for metrologists, so no easy way to determine the knowledge level of those performing the measurements. Since suppliers generally get paid the same for bad data as they do for good data (since the customer of the data usually can't usually tell bad from good) the data can generally be expected to be bad.

If the data is audited and compared to data for the same parts from a trusted source, then some correlation and possibly deficiency-directed supplier training could be done, but that all costs money, of course. This work should be done if the cost of dimensional problems will be high... A judgment call that all businesses need to make.

With the generally low level of GD&T knowledge and dimensional metrology knowledge, it seems that we're still in the "Wild West" stages of dimensional specification and measurement. It should get better over time as dense measurement data point clouds, improved GD&T standards, and very good software to evaluate the data become more common. Education and awareness in industry, and, I hope, awareness of the problems by mechanical engineering education departments, that may lead to ME graduates with better knowledge of how to specify a mechanical part's acceptable limits, should help too.

Quite a few people involved with this forum are working to make things better.

If suppliers knew that machine driven CMM measurement requires that they iterate as they establish a datum reference frame in order to gather repeatable data, that would reduce a significant part of the problem, I believe. The points they use to establish the datum reference frame need to be located relative to that datum reference frame in a repeatable manner and that cannot be done without iterating. The third iteration DRF is usually repeatable enough to be considered valid. The more imperfect the parts are, and the tighter the tolerances are, the more important it is to iterate when establishing the DRF.

Much of the training provided by measurement equipment representatives teaches methods for establishing datum reference frames, and methods for evaluating tolerances, that are incorrect. Their job is done when their customer can make the machine move and give numbers. The supplier's inspection department may consider their job done when the numbers they send don't end up raising red flags from their customers. It's up to the OEMs to first educate their own people, develop their own metrology capabilities, do correlations and probably then supplier training... That's all if they are aware enough, and care enough about their quality.

Thorny problem, I believe/agree.

Dean

http://www.d3w-engineering.com

 

[ALK35]

First of all I would like to thank everyone for some good conversation on this subject.

Dean hit on the subject that we are looking at, and that is the point cloud, white / blue light scanning. We are currently working with a place that scans the parts and gives us a cad model of it. We then put it on a “virtual gage” and check the part to the drawing specifications. Accuracy does not seem to be an issue, and it takes care of the CMM random point picking. I believe it is a step in the right direction but I’m sure there are some drawbacks that I have not come across yet.

 

[fsincox]

Dean,
Thank you for you insight, by iterate do you mean, basically just "do it more times than once", right?
Attempting the same points or differenent areas intentionally?

 

[DeanD3W]

Hi Frank,
Yes, by "iterate" I do mean do it more times than once... Since the points probed are located relative to the DRF that they are being used to establish, the process must be repeated, usually 3 times, to be able to repeat those point locations again some day.

The first iteration the point locations, especially on the primary datum feature, are not much better than "eyeballed" locations. For the second iteration all the points probed can be located relative to that first, rough DRF. For the third iteration the points are located relative to the DRF established from the second iteration. I believe the third iteration's DRF that is established from those points can generally be considered repeatable enough. If there's any doubt, a forth iteration could be completed in order to verify that the change in DRF orientation and location from the 3dr to 4th iteration is small enough to be insignificant.

The repeatability becomes a greater concern the more imperfect the parts are, and the tighter the tolerance values are.

I've been to quite a few suppliers who looked very doubtful when iteration is first explained to them. I don't think the measurement equipment companies generally train people to do this. Whether it's worthwhile or not can be determined by the difference in measured values between a "no iteration" DRF and a "3 iteration" DRF.

Dean

http://www.d3w-engineering.com

 

[fsincox]

Dean,
Thank you, I had never thought of that as an option. I figured that is why they are working on scaning probes and optical methods to reduce these issues through quickly obtaining a larger amount of data points.

 

[dgallup]

There should be gage R & R done on any inspection equipment. You basically have multiple people measure the same batch of parts repeatedly. If you don't have repeatability, you don't have a measurement system.

----------------------------------------

The Help for this program was created in Windows Help format, which depends on a feature that isn't included in this version of Windows.

 

[dingy2]

dgallup:

Your R & R study would go wild on most GD&T applications such as the flatness example unless there is a set documented measuring method including set up. If the application is considered "critical" then no more than 10% of the process or tolerance specification is allowed. On other characteristics, no more than 30% is allowed. I have seen R & R studies in the past that reflected 100% of the tolerance range.

Most companies use 10 pieces, 3 Operators and 2 or 3 reps each following the AIAG method.



Dave D.

http://www.qmsi.ca