Geometric Dimensioning and Trouble

apone_gdtRight behind Stonehenge and an improbability drive, Geometric Dimensioning and Tolerancing (GD&T) is somewhat of an engineering enigma, if there ever was one.  Developed as a language for precisely communicating design intent, the ASME standard has been in existence for over half a century.  The benefits of GD&T are very real, providing a reliable and verifiable means to ensure parts function and interface as intended, while reducing scrap rates normally driven by unnecessarily restrictive rectangular tolerancing schemes.  Yet, at the same time, GD&T is largely not institutionalized in engineering curriculums, nor  is training widely delivered to engineers at companies of all shapes and sizes.  Some companies hold an utter disregard, and don’t utilize GD&T at all in their engineering.  Model Based Engineering (MBE) promising to annotate GD&T directly into 3D models as Product Manufacturing Information (PMI),  continues to suffer serious adoption problems.  Certainly, something is wrong.

Sometimes it’s easy to characterize GD&T as an absolute, the fanfare can resemble the Sarge coming out of hypersleep:  Every day in GD&T is like a day on the farm.  Every feature control frame is a banquet!  Every cost savings a fortune!  Every virtual condition a parade!  I love GD&T!  But such enthusiasm is not misplaced.  Reducing design intent miscommunication is paramount when dealing with global suppliers.  Part interchangeability or scrappage problems can be quite costly, especially in higher volume environments.  GD&T is specifically tailored to alleviate these issues.

However, there’s certainly an issue of complexity.  Even with substantial and frequent training the concepts involved are fundamentally straight forward, but ultimately arise to rather complex situations in practice.  You can still have furious chair-tossing debates over whether that tertiary datum should have been referenced at max material condition (depending on which version of ASME Y14.5 the company has adopted).  Such arguments could result in questionable references to parentage and unwelcome advice on where to place your datums.  In many ways it’s quite literally a system developed by rocket scientists to be used by ordinary humans.  As fundamentally perfect as the system is, it’s not terribly accessible for many and that’s a problem.  Most experts are apt to react negatively to such criticism, pointing out that the standard by its very nature is necessarily complicated.  Tolerancing complex parts can never be “dumbed down” and to even attempt such is an unacceptable concession.  In other words, if you can’t deal with complexity, then the problem is youLook into my eye.

So if the system is perfect, then the problem must be those pesky humans.  Too many of them don’t know what they are doing.   What do we do?  Resort to genetically-engineered border collies?  Non-adversarial Cylons?  We could try training the humans and vaporize them if they complain too much.  If their management doesn’t want to pay for said training, then we’ll vaporize their management.  What could possi-bly go wrong?

Most engineering curriculums are limited to a single engineering documentation course – many of which have been more aptly focused what little time they do have on Computer Aided Design (CAD) and other technologies.  There’s not much room left in a 4-5 year program unless we dispense with, say, Chemistry.  I don’t think that’s a viable option.  So it seems most GD&T training happens on the job or at least within the realm of the employer.  And employers are reluctant to spend ample dollars on training despite apparent costs savings – often because such cost savings aren’t adequately measured or the cost barrier of training everyone.  The result is few engineers are trained.  Many have exposure once to a 3-4 day class, and the result, without continuous training, is sometimes more disruptive than total ignorance.  A designer wholly ignorant of GD&T is one thing, they are obviously easy to spot.  A designer who has had some training, but may have a misunderstanding or a specific skill gap, will more likely result in undetected errors down the road.  That hidden cost damages the overall value proposition, and becomes a barrier.  So you see, complexity is a problem, regardless of your ideology.  Hudson, come here.

But what about all those companies that aren’t using GD&T?  Shouldn’t they be going out of business?  Shouldn’t GD&T be a necessity for competing on quality products in this global economy?  Many of the companies who skirted GD&T in the past tended to be low volume, vertically integrated, or reliant on a +5 sheet metal alchemist with a forge at the local renaissance faire.  But many are observing the trend of companies giving secondary consideration to GD&T seems to be increasing, rather than decreasing.  Is there another factor at work?

Let’s think about scrapping parts.  One of the rather correct tenets of GD&T is that you can never manufacture a perfect part at nominal tolerances.  Too much variation from that perfect part results in scrap.  GD&T optimizes the management of those variances.  But what if manufacturing in general advances to a point where variances are rather small for most applications regardless, with little or no added cost?  Perhaps we’re starting to see this manifest itself.   Case in point a quote from Lean Startup:

“One often hears commentators lament the loss of manufacturing jobs in the United States over the previous two decades, but one rarely hears about a corresponding loss of manufacturing capability.  That’s because total manufacturing output in the United States is increasing (by 15 percent in the last decade) even as jobs continue to be lost.  In effect, the huge productivity increases made possible by modern management and technology have created more productive capacity than firms know what to do with.”

Since we’re already talking Lean Startup, let’s take the entrepreneurial perspective to GD&T as an interesting exercise.  GD&T is a product that (as far as you are concerned)  is perfect in every way.  Perhaps you can even prove conclusively how perfect it is, by how it measurably performs, improves yields, or saves money.  Trouble is, too many companies don’t seem to be interested.  Do you blame your flawless creation – or do you blame the consumer of that technology process?  That’s the tricky part of entrepreneurship.  Just being right is not enough.  You have to be right, and make your solution compelling.   And as compelling as GD&T is, in today’s continually improving manufacturing environment, perhaps the luster of an evolved 1965 standard has dulled somewhat.

Does that mean GD&T needs immediate evac?  No.  But it’s clear that something has to change, and MBE represents that opportunity.  However, up to this point, MBE is merely lifting GD&T and the standard as is into 3D as appended annotations.  Worse yet, the transition is losing further focus on drafting skill, merely because drawings are seen as an anachronism, regardless of the universal versatility of their presentation.  MBE is an opportunity to leap forward in technology and provide true visualization for dimensioning and tolerancing in a way previously unimagined – and therefore increasing the accessibility of the principles that drive GD&T.  The question remains: will anyone truly seize that opportunity?  What do you want me to do fetch your slippers for you?  Gee sir, I’d like that.

  • Hans Erdmann

    Hi you hit a problem almost no one wants to see, there is even more: Parametric dimensioning is not connected automatically to dimensioning. The companies get really confused by extra communication costs due the drawing is in many cases the basic legal ducument. Dimensions and parametic dimensioning is a pain to tranfer between CAD systems in spite of STEP or IGES. Standards are hard to translate between ASME and ISO for normal engineers. In the Electrical area we have a total standarddiseaster just open a pc or smartphone.

    • Thanks for your input, Hans. Some tools do have connectivity/inheritance between 3D annotations and traditional drafting annotations but it’s often not 1:1, and the problems with translation you mention still apply. JT, now an international standard, is supposed to address some of these issues…
      Also, in your opinion, why do you think no one wants to confront the problem?

      • Zevzek

        Let me put my two cents here.
        There are, and always were, new technologies vs. old ones.
        Today’s CAE is a continuation of drafting board and excel sheets engineering, but you’ve always had good and not so good engineer(ing). I wouldn’t expect any new technology to completely replace knowledge you need to create value added content in addition to the capabilities of new equipment. Knowledge will always be needed to choose and utilize different new technologies, read $$$ per product unit.
        When machines start making all decisions, there will be a cyborg world, no humans, anyway.

        • Michael


          To underscore your point, I wrote this earlier

          The misconception seems to be that “we can get rid of GD&T through increased technology.”
          Well…GD&T is not married to 2D drawings…nor any technology. Some folks believe GD&T began in the auto industry in the 1970’s…other folks believe it originated from the US government. I believe the truth is that GD&T at least goes back to when Roman generals decided to bring along spare parts for their catapults as they prepared to conquer a city.
          My view is that dimensional engineering will always be around…the question is whether you want to do it better than your competition.

          I’ll take this and tie it back to Ed’s article now.
          It seems no one wants to do it well because we think we’ve approached “tooling capability utopia.” Well, several thousand years from now…people will think of our technology as crudely as we today think of the…then state of the art Roman Catapult.
          …And dimensional engineering will still be around.

      • Hans Erdmann

        Let us guess.. It keeps people to one CAD develloper offering internal communications and the customers have no speach. It is like medical industri.. A leak of economical incitement.

  • I really don’t understand the problem. I thought CAD was going to make things easier.

    GD&T is Feature Control. Instead of defining a location you define the feature envelope.

    It is only for inspection. The drawing is an inspection document.

    Form, fit, function and dimensioning was in the realm and responsibility of drafting. It was always checked by two people the drafter and the checker. The checker was an experienced drafter. Engineers never got involved and it looks like when engineers do get involved simple things get complicated. What seems to be missing is the Checker. There seem to be many new positions being created to replace this position.

    Today with CNC hitting .005 do you even need GD&T unless you need a tighter requirement. 90% of the current process is probably covered by a profile anyway.

    I am a big believer in doing complete drawings. Yes, PMI can do the job to get the information to manufacturing. But the drawing give a second check for the designer, engineering or drafter and an easily re-viewable document..

    I have 47 years as a drafter and when I do a complete drawing from my solid model design I aways find an error or maybe see a better design.


    Redefining 2D/3D

    • Thanks for you comments, Joe. Drawings have value for all the reasons you mentioned. Anything that comes next must at least match if not exceed that value, and be as or more accessible. Reinventing without acknowledging that value is what has soured MBE.

      Regarding your comments on CNC accuracy – I tend to agree. Do you think that improved machine accuracy has been progressively marginalizing the benefit of GD&T as a means to more precisely characterize a required tolerance?

      • I put a bit more thought in CNC and a base tolerance. You could just set a profile tolerance of what ever the CNC people agree. The machines standards are probably tighter than .005. Unless otherwise specified .005 is the tolerance and we design to that.

        You know that suppliers are really smart and can figure things out! They have CAD systems. Not just CNC, but actual CAD systems. The difference between engineering CAD and Manufacturing CAD is that the manufacturing CAD has direct edit capabilities. At least they do in the NW. All have a seat of CADKEY/IronCAD/ZW3D. When a supplier tells me he has Solidworks, I laugh probably losing a sale, knowing he is working twice as hard as he has to. All of the systems will read all of the native files and bring the assembly into a single file.

        We still need complete drawings. All dimensions go in without a tolerance except for a few tight fits. How easy would that be???

        • Michael

          There is a fundamental problem with adopting common tolerances for types of features….or interfaces.
          When I use a tape measure, I immobilize one end of the tape with my left hand….then extend my right hand to measure. Both hands are required to do their jobs…otherwise my measurement will suck.
          By adopting a common tolerance for all features based on tooling certification capability, we’re concentrating on our right hand…and not paying attention to what our left had is doing.
          Our right hand is the controlled feature…our left hand is datum strategy.
          In the last several decades, engineers and designers have focused on controlled features…but programs have succeeded or failed…and companies have succeeded or went bankrupt due to datum strategies.

      • Michael

        Ed and Joe,
        Yes…A CNC can make an extremely accurate tool.
        So…let’s say we’re talking about a primary tool.
        The CNC will cut it…and then the actual tool is inspected to within +/- 0.25mm to CAD (I’m one of those metric guys).
        I agree this should be done…But, in doing this, you’ve only certified your tooling…you have not guaranteed all the parts coming off of that tool will be +/- 0.25 mm to CAD…or +/- 0.25 inches to CAD for that matter. And, we will have no idea whether the process is in control or out of control.
        The main idea…do the actual parts you pull off of your CNC-cut tools vary enough to possibly affect function? Do you want to monitor your process for some kind of statistical quality control or statistical process control?
        If so, you need some kind of GD&T.

        • Michael – completely agree with you with respect to tooling (that is a critical case for GD&T), but I was talking about just parts in general that are machined without tooling of any kind (more now than ever)- in many of those applications the machines are unfailing accurate now. Wasn’t always the case – and with the advent of additive manufacturing processes, more and more parts end up being created with processes that are impressively reliable from both accuracy and repeatability – all by just supplying a model. Surely this broader trend is having an effect?

          • Michael

            I’d argue that the machines are capable of greater precision now…I haven’t looked at accuracy. Accuracy is a very complicated animal. We’d have to look at things like bias and stability to determine accuracy.
            I do agree though that the science of tool-making has improved…and that additive manufacturing (3D printing) is on the rise.
            I’d also like to add that the practice of dimensional engineering has the ability to help companies realize the benefits of those new technologies.
            The misconception seems to be that “we can get rid of GD&T through increased technology.”
            Well…GD&T is not married to 2D drawings…nor any technology. Some folks believe GD&T began in the auto industry in the 1970’s…other folks believe it originated from the US government. I believe the truth is that GD&T at least goes back to when Roman generals decided to bring along spare parts for their catapults as they prepared to conquer a city.
            My view is that dimensional engineering will always be around…the question is whether you want to do it better than your competition.

          • I would certainly agree with you Michael, dimensional engineering is a fundamental consideration, despite whatever name we assign to it. As GD&T is an evolution of what the Romans did, is it time for the evolution to take another step?

  • Designing
    and defining tolerances based on CNC capability as discussed below demonstrates
    total incompetence in both design and how CNC works. CNC CAN cut tools and
    parts within .005, but it does not have to. CNC has cutting speeds, steps, and
    tool changes, which includes tool maintenance (sharpening) not to mention
    machine calibration. There is this thing called CNC programming. Based on all
    the above, not to be completely inclusive, CNC MAY cut more acceptable parts at
    greater speeds with less cost if the true requirements for tolerances is stated
    on the 3D model.

    Drawings? Those are history. Any company of any significant size
    still producing 2D drawings is wasting money. It is as simple as that and I
    don’t care to be politically correct about it. I am in the business of showing
    companies the benefits of both using GD&T right and annotating 3D models
    the right way and I guarantee my services save the company money. I’ll show you
    the metrics that so can be measured.

    Who says everything is made by CNC anyway. Total foolishness.

    For the companies who get along without using any GD&T and
    even produce good product, I have a simple question for the CEO of any such
    company. Wouldn’t you like to make a higher quality product, have more
    satisfied customers, and do it with a higher profit? Any competent CEO will
    answer yes to that question.

    Some companies have had so called trainers (draftsmen) come in
    and teach GD&T only to see no benefit. They see their people argue more
    about the GD&T then producing parts and moral go down. The flaw is simply
    bad training and bad mentoring. That of course starts with any clown who thinks
    CNC does it all for you and GD&T is unnecessary.

    There are too many so called GD&T trainers out there that
    should be out of work. In fact, many of them are out of work. That is why they
    start training on a part time basis in hopes to get an offer to go direct for
    some customer as their internal so called expert.

    To you managers and CEOs. If you are looking for real GD&T
    training and consulting, the very first thing you need to do is VERIFY that the
    individual is ASME GD&T certified. That can be verified on the ASME web
    site. If their name does not appear, you can call ASME and ask if someone has
    ever been certified. (Not all certified people renew their certification for
    numerous reasons.)

    AND, if anyone coming on your site can’t show you the real money
    in 2 days or less, let the door hit them on the way out.

    This is a good article. It gets people thinking and I believe
    has the undertone of the necessity of GD&T and how model based definition
    can greatly improve the application of GD&T. Am I wrong about that? Good

    • A couple of points here:

      CNC technology is rapidly becoming highly available at very inexpensive costs. There are CNC machines used for students that cost $1500

      I’m not saying the $1500 machine is making your part today… but don’t be surprised if it does in 10 years. The larger point here is CNC is getting so ridiculously cheap, it’s starting to render the cost savings from more precise dimensional management irrelevant. That’s why the companies not using GD&T aren’t suffering the punishment they would have in the past. Of course, not every part is CNC’d, but with the increasing advent of additive manufacturing, more and more parts are departing from more traditional tooled processes. Expect that trend to continue.

      I agree that it is time to abandon the drawing… however the technology is not where it should be. The usability and accessibility of that technology has not been perfected. It’s not compelling enough. Users should come flocking, but instead they try to run away. Case in point, companies are still desperately hanging on to drawings and MBE adoption continues to be low, despite that the basic technology (and standard) has been around for a decade now. That tells me there’s a problem. A resolvable one, but a problem nonetheless.

  • I fail to see how MBE can somehow replace or change the physical reality described with correct GD&T. The standards for GD&T are not perfect. It continues to evolve. But the physics do not change. What MBD/ME does offer is great simplification of not the application of GD&T (it isn’t hard in the first place.) It does a great job of simplification of adding the GD&T to the part definition. For example, the time consuming useless BASIC dimensions can be eliminated with MBD. A single profile tolerance can be applied to many surfaces, not the same as all around, in a single view a single feature control frame, which adds clarity over the numerous feature control frames needed in the old drawing paradigm.
    Although manufacturing capability continues to improve beyond what some people think are diminishing returns, try explaining that to P&W and the F-35 engine that catches fire when you start it up. You might guess that some of the best manufacturing technology was used. You might also guess that the GD&T or lack thereof was very questionable. Will P&W go out of business. Maybe not. But customers like tax payers pay for the product. The same is true for other successful companies not using GD&T. My question is always the same. Would the company like to improve?

    • Norman, I’m not suggesting the MBE needs to displace physical reality but rather that how dimensioning and tolerancing is applied needs to fundamental change into a richer more visual format that only 3D can provide. Feature control frames and annotations are constructs created to fit within 2D paradigms. We need more out of the box thinking here.

      With regards to manufacturing accuracy and the P&W failure, I would argue very few failures are due to out-of-spec parts these days. Again, this is because of the extraordinary dimensional accuracy afforded by today’s processes, especially the processes used for high cycle multi-contoured parts like turbine blades. Failures will instead be analytical, materials, or errant design. In fact, this particular incident was traced to vibration interacting with parts made from a bad batch of titanium with compromised strength. GD&T does nothing for internal material quality.

      • Ed, I am all for more out of the box thinking. But we always want to see benefit too. At this point, I think the symbolism of GD&T (FCFs and modifiers) do an outstanding job. So, I guess I am not seeing how it could be any better, yet. I do know that if someone ever shows me an example of better symbols or attributes ( I tried attributes back in 1982-1984 on CADDS3), I will be all eyes and ears.
        As for the potential for out of spec parts referenced above, the extraordinary accuracy manufacturing processes do not guarantee parts that meet true functional requirements. And the inspection of the parts are very much in question. One might think that those multi-contoured parts are defined with appropriate tolerances and datum reference frames. Let me assure you that I have had some very interesting discussions over the tolerancing of some turbine blades and I can also assure you, I was stunned. Let me leave it that.
        Certainly other factors may come in to play. A big one is material quality. No question about that. But just because someone somewhere says that was the problem, I don’t buy it. Perhaps I have seen too much. 🙂
        Certainly a problem with the dimensions and tolerances (GD&T) can be found to be a root cause, because to do that, one needs to know GD&T in the first place. Get my drift?