Data on Cost Bias for Geometrically Dimensioned and Toleranced Parts? 5

[solid7]

I would very much be interested to know if anyone can produce any tangible data, to suggest a correlation between GD&T on a drawing, and an increase in vendor procured parts?

I'm not trying to lead a revolt. Clearly, GD&T is the way to go for a great many parts. But for low cost drivers, I believe that many vendors will arbitrarily increase the quoted price, even at the mere mention of the most primitive tolerancing. (or even datums) I cannot prove this, and I can't get anyone to commission a study of the matter. But as a 24+ year engineer, and former business owner, I have firsthand knowledge of this issue. Which does no good, when I can't reliably communicate it to others.

The company that I work for, is currently trying to implement some tolerancing rules, that are broken down by product level. Main product, tooling, electrical, GSE, etc. I believe that we need a very clear delineation in these product lines, as none of them demand the exact level of rigor as any of the other.

 

[JohnRBaker]

But what do you offer as an alternative? I mean, you have to have some sort of quality control, at least in the sense of being able to specify, and subsequently verify, what it is that you're getting from your suppliers.

Now if you're talking about holding part parameters to too tight of a tolerance, inappropriate for either the intended use, or the expected methods of manufacturing, I might be willing to be concede that often poor application of tolerances, whether they're documented by GD&T or some other scheme, can significantly increase the cost of a component without making it any more suitable for its intended use.

As a young engineer I had to learn the hard way that forcing the machine shop, or a supplier, to meet very close tolerances, just to make sure that parts will always fit together, is often the best way to guarantee the highest possible cost for a part what will work as expected. It took time, and the mentoring by people with years of, in my case, machine design experience, to learn the tricks of the trade where you can specify part designs that will fit together, and perform as expected, even if you use less stringent tolerances by simply thinking smarter about how parts fir together and which dimensions are most critical and which are not, and so you learn to loosen the tolerance where it has little or no effect on fit & function, while holding those that do. Also, you learn to understand the manufacturing processes which are appropriate, again, to assure the best fit & function without wasting time and money on overly complex tooling and work processes. For example, don't specify a tolerance which will require a milling operation when they could have used a drill press to produce a hole to your component, and if the set-up and quantity are such, you might even move it to a punch & die operation, which could reduce the cost even further. As an old timer once advised me, "Don't force the shop the use a 3-axis mill, if cutting it with a torch would have still given you what you wanted."

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[solid7]

The alternative is a tolerancing scheme that doesn't use GD&T. If you notice my initial post, I suggested that there were multiple product streams, all varying in degrees of sophistication.

This isn't about inter-office revolutions, or radical new ideas. It's about trying to extoll the virtues of allowing simplicity, where simplicity dictates. We currently have a QE mandate that ALL parts must have a 3 datum system, regardless of their complexity. In addition, their policy is that if a part has datums, it SHALL implement tolerancing which relates to said datums. So, we're held hostage to this methodology. I could certainly have it overturned, if I could provide data that would illustrate an unintended negative.

I'm not going to get into a drawn out discussion over the specifics. I'm very confident that I know how to ensure the produceability of parts. The issue here, is whether or not we can prove a cost bias. I have literally hundreds of pieces of very low level tooling that benefit nothing by having anything more than an overall tolerancing scheme. In stark contrast to the hundreds of pieces which deserve more elaborate tolerancing, up to, and including SPC.

My problem is not lack of experience - at least not from my end. I'm 25+ years in the design business, with a machining, R&D, and prototyping background. In addition, I am a former business owner, who was responsible for design, development, prototyping, and delivery of custom products and product lines. In that time, I met more than one vendor who openly admitted that geometric tolerancing was a cost driver. I also established relationships to thwart the price premium. (i.e., work with the vendor in the design process, to reduce the cost, and understand why they build it in, in the first place)

But... surely somebody, or someone that somebody works for, must have conducted an applicable trial to ascertain the pricing effects of product definition... I don't believe anyone who would make the kind of claim that I am making, and can't corroborate it.

 

[3DDave]

No one will have conducted such a trial because the end result is not verifiable. There won't be a way to ensure that parts will meet some set of requirements that are not there.

However - I think there are a large number of cases where the general precision available produces sufficient accuracy (not the QA "accuracy" vs "precision") that resulting variations are likely to be rejected only in the case of very small tolerances. Case in point - I saw an entire program omit all angle tolerances on 90 degree implied dimensions and not a single complaint seem to have arisen to correct that oversight. If even 1/2 degree out parts would fail to fit.

Likewise I've seen projects where the allowed tolerances would result in 20-50% failure rates of the final product, but the supplier simply made the product not to fail; all tolerances could have been omitted and the parts would still work because the supplier would not want to lose the contract; nor would they complain or ask for clarification.

In contrast I "worked" with a supplier to create an ICD to ensure that parts that would be interchanged (covers to transit cases) would be interchangeable. I made up the drawing, said it was preliminary, and asked if they had different dimension origins to mark those up. Their mark-up was to cross every dimension from the drawing so that their QA would not have to inspect them. Their "solution" was to add a note that the lids for any one order would be interchangeable, but no guarantee that they would fit from the next order.

 

[IRstuff]

Well, of course, it's a cost driver; it's got to be cheaper to just crank out crap without regard to any acceptance criteria. The question isn't the cost of THAT piece; it's the cost of trying to get crap to fit together, hunting and pecking for replacement parts, and rework and rejection at the end of the line, when all the mechanical parts have to work together and work smoothly. The Japanese auto industry didn't eat the US auto industry's lunch in the 70s and 80s by building loose tolerance engines and transmissions; they did the exact opposite. They made their parts tight where needed and loose where needed, so their engines ran without problems, were cheaper to build, and required less repair because parts didn't wear because they were fitted and toleranced well. In the meantime, companies like Ford would allow their machining, even with tolerancing to bang against the limits because it was cheaper to machine, but they burned tons of money trying to get one engine and one transmission to play nicely together. And, the cars were ludicriously unreliably.

While I have no direct experience with that type of manufacturing, I was party to second-sourcing a Japanese integrated circuit to be run on a US manufacturer's production line. We had similar "tolerances" and we had test chips on each wafer to ensure compliance to the semiconductor processing requirements, but it usually cost 5 chips in the prime real estate on the wafer. We got Hitachi's mask sets, and no test chips were to be found; we asked and they said, "Don't worry about that." We asked what processing parameters (think surface finish, etc.) were needed, and they said, "Don't worry about, just run it in your standard process." So, we did, and the first batch yielded better than our own products that were specifically designed for our own process, and the second batch yielded even better.

The moral is that the notion that adherence to tolerancing "costs" more is a myth propagated by lazy manufacturers who don't want build quality product. They're willing to do the same thing their grandfathers did and suck up the rejects and poorly fitting parts as the cost of doing crappy business.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[solid7]

I'm not really looking to turn this into a rant. And for any one individual to just adamantly declare that "no one will have conducted" such an experiment, is - to put it politely - patently uninformed. There are various small scale studies that we regularly refer to in our industry/operation, that are inherently more pedantic than what I'm asking for. And if you knew who I worked for, you'd know that what I speak is true. (no, I won't tell you who I work for in this forum)

Try to step out of your own experiences for a minute. Of course, by asking that, I realize that this immediately disqualifies you from being able to answer the question. But many modern workplaces are data driven, and do not rely on such anecdotal and presumptuous logic. (as what you've put forth, declaring that this exercise has no merit)

I have absolutely no idea where "crap" enters into this discussion. Never did I say that we have any issue producing the product that I'm implicitly referring to. I did not mention tolerancing, fitment, mating conditions, etc. None of that. That's a nice story, but it has no relevance to what I'm asking about. But if that's not sufficient, let me be clear - I'm 100% positive that the requirements that I specify, are sufficient to get a quality product delivered. Without knowing what my product or process is, you have no means to undercut me on my assertion. Nor is there a need to. My issue was that I'm being constrained by a department outside of my own, who is overriding my design intent, with their own internal processes. (i.e., it is MY job to specify criteria, theirs to inspect to what I define - I trust everyone to do their job, and ask the same) In turn, it may potentially mean less money in MY coffers, to work with my entire product line for the year.

So, back to the title. Does anyone have any data to suggest a cost bias for geometrically dimensioned and toleranced parts? I'm not looking for opinions about whether it's the right thing to do, or whether I'm doing it right, or anything else that's not directly associated to this thesis. If the answer's no, so be. We'll close shop, and move on to the next topic.

 

[IRstuff]

Now, the current crop of CNC machines probably will build parts to some level of precision, regardless of the spec, since it's built into the lead screws, and tooling, etc. If you can live with that, them that's great, but it's not because you didn't do GD&T, it's just that the machining supported a level of accuracy that was better than you needed. But, there are lots of applications where that could be doomed to failure, if, say, the fit requirements were close to the limit of the unattended manufacturing equipment.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[solid7]

There are implications to specific tolerancing schemes that have nothing to do with the manufacturing operations, themselves. Consider such things as one vendor's own internal processes, in-process check requirements, final inspection, interpretation of statements of work, or, or, or... There are many facets to this discussion that don't entail the actual part production. Every one of them may get a $ for reasons that don't even appear tangible in the most straightforward parts of the conversation.

 

[3DDave]

"I believe that many vendors will arbitrarily increase the quoted price, even at the mere mention of the most primitive tolerancing. (or even datums) I cannot prove this,"

You can prove this. Send the drawing out for bid with no FCFs and then send out an update with FCFs.

What you might have difficulty in proving is the drawings are identical in their requirements.

Please share those example studies. We cannot know what is applicable to your products if you don't tell anyone what they are.

 

[solid7]

You can't prove this unless you've sent the same basic thing to the same vendor, and asked for two (or more) bids, in some unrecognizable derivative form. Otherwise, you've let them onto the game. One could optimize a part, that has identical volume, identical machine run time, (based on subtractive surface area), and a few other factors. Then, you apply the different tolerancing methods, to each part, see what comes back. Valuable data rarely comes so easily as formulating a sophomoric "just do this" strategy. And what I've described, is a concerted effort. It's actual research.

The problem is, most people don't have the time or scale for this experiment. This is squarely in the realm of data gatherers, not engineers. I do not have a properly established channel for this data.

With all due respect - I'm not under any obligation to provide you with any information about unrelated processes, simply because I've mentioned it. At best, it only serves to take this conversation in another direction. If you feel that I'm flat out lying to you, about a company using intelligent data and metrics to establish pricing strategies, then by all means, please just say so - and I'll ignore you properly. Otherwise, understand that I don't work in the back office at the local foundry. We actually do have some fairly sophisticated practices, that might not be "the way it's always been done"

 

[solid7]

Also... when I say that I can't prove something, it means that I don't currently possess the facts. And unlike the way some things have been presented in this thread, I will not try to pass opinion as fact. What I know from experience, is still just heresay without data. This does not mean that it's absolutely unprovable - only that it is, at this point, unproven. Please don't make the mistake of taking my comments out of context.

 

[CWB1]

Sounds like you have your mind made up to prove the overwhelming majority of the world for the past 40(?) years wrong. While I have no doubt there are studies proving you correct, it will largely be bc there's a "study" proving pretty much every opinion correct including the existence of a flat earth and aliens. As far as I know, most every GD&T and quality text written mentions studies proving GD&T's value simplifying and clarifying prints, as does common sense. I honestly cant think of a part print which GD&T doesn't simplify, unless you don't fully define the part.

I see vendor prints occasionally with overly complex tolerancing that a few simple modifiers could easily define. Much like an unprofessional attitude, they're usually a good indicator that we don't want to do business with that supplier and more often than not those prints are trashed and we move on to other companies. Growing up in a job shop it was much the same, if a customer asked you to copy or repair something without a print, we did it. If a customer provided a lousy print we usually didn't even consider the work.

We currently have a QE mandate that ALL parts must have a 3 datum system, regardless of their complexity. In addition, their policy is that if a part has datums, it SHALL implement tolerancing which relates to said datums.

So how does that actually complicate anything? One view defines three datums.

 

[solid7]

Sounds like you have your mind made up to prove the overwhelming majority of the world for the past 40(?) years wrong.

And it sounds as if you are taking what you want away from this, rather than what's clearly printed on the page. I'm beginning to see that it's impossible to have discussions here, without endless tangents, unsolicited criticisms, and needless lecturing, or citing of personal tales. My question was simple: do you have data? I didn't ask you to question my premise, to tell me how I should be thinking, to provide me with comedic distraction, or to reiterate the value of a system that isn't even being questioned.

We are trying to establish if there is a particular psychology to the quoting process. Every sample has a use case. I have many uses cases, and a finite budget. I know what I am asking, and why.

As to your experience, please do not presume that I have control over who is selected to do business. Again, not related to the discussion.

 

[handleman]

A lot depends on what you need.

A simple prismatic type part, in which standard machine shop practices and default dimensional tolerances per ISO 2768 will yield usable parts, and part qty less than 5 or so, will add cost to add GD&T due to the added time to both interpret and inspect it. Probably not much more. How much more probably depends a lot on the shop.

The tighter and more critical your tolerances become, and the higher your part quantites become, the more benefit you get from GD&T in precisely defining what you need while allowing maximum flexibility elsewhere.

Also, the higher your quantities, the more the up-front print interpretation and inspection setup costs get spread out, making them less of a factor.

I would certainly agree with the assertion that GD&T can always better communicate and clarify the requirements for parts. And if you're quoting thousands or even tens of precision parts, it's essential. But if you're quoting qty 1 welded angle bracket to mount an air cylinder to the side of a machine using clearance holes, a drawing referencing ISO2768 with no explicit tolerancing at all will, in almost all cases, be less expensive overall than one in which the amount that you really don't care about most dimensions is explicitly laid out in GD&T.

 

[solid7]

I have hundreds of "qty 1" parts, with relatively zero sophistication, which will NEVER be made twice. In fact, where possible, these parts get 3D printed.

Nobody is getting rid of GD&T for the vast majority of our processes.

 

[drawoh]

solid7,

Your question depends a lot of who the vendor is, and who is applying the GD&T to the drawings. If everybody understands the GD&T, there ought to be much opportunity to identify non-value-added tolerances, and make the parts cheaper. It is not just the fabricators who do not understand the drafting standards.

There was a good discussion in the Drafting Standards forum on thread1103-322065.

--
JHG

 

[solid7]

That's kind of my point, drawoh. If we say that "it depends on the vendor", then we understand that there is some bias. There are many reasons. But the buying process can be completely blind to the underlying factors.

I have constant discussions with some of my colleagues, regarding the necessity of frugality, and the avoidance of "machine shop prices for weld shop parts". We're more or less at bare bones for drawing information - but we have no problem with product definition.

For those who have their boots on the ground - i.e., our purchasing agents - and I mean the ground level, not operational level people, there is an explicit understanding that not all shops are equal. However... the conundrum is, that some of the shops who may be a little less sophisticated, and more prone to charge a premium for GD&T, are also the ones who provide the best price for short lead time. And they deliver the product we need, 100% of the time. We understand how to leverage the strengths of different entities. We just aren't always able to communicate that to our other team members, however. (just as we can see in discussion, people think/say/do as they prefer, not necessarily what's being asked)

 

[3DDave]

"We are trying to establish if there is a particular psychology to the quoting process"

Should have led with that.

And it seems like your problem statement is:

"We just aren't always able to communicate that to our other team members"

After all, you state that you can communicate with suppliers to get the best price, best time, best performance, but not your own people. Why are they going to believe random people on the internet when they have you?

 

[solid7]

"After all, you state that you can communicate with suppliers to get the best price, best time, best performance, but not your own people. Why are they going to believe random people on the internet when they have you?"

That is NOT what I said. I said that in a former capacity, that was my firsthand experience - negotiating and developing pricing. Apparently, you don't have a grasp on how disconnected, and full of blind spots, organizations can be in large companies. If you think that you can solve that problem, then you'll be my personal hero, and certainly, you'll be a captain of industry. But for me, I'm working to be part of that solution in my organization.

If it's unclear that I'm looking for a psychological angle, when using terms like "bias", then I'm not sure how much more I can clarify this for you. Let me point out, however, that it was you who suggested sending the same quote out for bid twice. So that certainly implies that you have at least somewhat of a grasp on what's being said.

"We just aren't always able to communicate that to our other team members" means exactly what has been demonstrated in this thread: I can tell you all day long what I want from you, but if you are determined to go your own way, that's exactly what you'll do.

Case in point: I ask for research/data, or some reference to it. What I get, is anything but research/data. Translate that into the office space. As someone in a position to do so, I ask you for something that means something to me, but not you. What I get, is having to put up with a lot of BS - some of it self-righteous, some of it uninformed, all of it anecdotal, and none of it helpful. But somehow, it's thought that I need to hear your official take on the matter, as a whole. If I just tell you my rationale without fact, I get all of it, reloaded, and with ever-gaining forceful momentum. So I like to speak to facts. Everyone's got a story. I'm much less interested in those, unless you're buying me the next round of a nice single malt scotch. Then, we can talk about the other stuff, in a more casual, and less consequential manner.

Back on track - I can suggest to QA why I don't want to add a specific callout. Rightfully, it's well within my boundaries to do that. In no scenario should QA be dictating engineering requirements. It can be negotiated, and it can be pointed out where error occurs, or where improvement is warranted. But in our case, we are transparent, and collaborate amongst groups. Sometimes, that gives a little too much boldness to those to whom it is not entitled. But it helps us all get better at what we do, if approached with the right mindset. But we are fact merchants, first and foremost - not speculators. This is an attempt to prove or disprove, that my past experiences, hold credibility. If I'm correct, there is a direct economic impact. If I'm incorrect, we can state with certainty, that there's no reason not to accommodate an outside request.

 

[solid7]

For the record - the QA inspector wants to know the same thing that I want to know. So that you know that it's not just me, on some rogue quest to "undo the last few decades of progress." We are working together to define processes for all who produce product definition. He and I are NOT in disagreement that we can do without the tolerancing, from a pure definition standpoint. Just whether or not the more valuable purpose is to create a homogenous approach for all product lines - or delineate our methods, to achieve value, relative to function. His angle is more to the effect of, "creating a habit".