Latest / Mechanical Engineering Made Simple / (#35) GD&T Deep Dive Mastering ASME-Y14

Transcript

(#35) GD&T Deep Dive Mastering ASME-Y14 · Mechanical Engineering Made Simple

• 0:00
  Welcome back, design junkies, to Mechanical Engineering Made
• 0:02
  Simple. Today we are opening the master
• 0:05
  blueprint, that secret language of the design specification.
• 0:09
  It's what we use to make sure complex machines actually fit
• 0:11
  together perfectly. And this isn't just about, you
• 0:14
  know, making things. It's really about setting the
• 0:16
  geometric requirements. These requirements determine
• 0:19
  well the success, the profit, or maybe even the failure of an
• 0:23
  entire assembly line. So this deep dive, think of it
• 0:26
  as our shortcut to understanding that contract you sign with
• 0:30
  manufacturing. That's absolutely the mission.
• 0:32
  Today, we are navigating the authoritative standards that
• 0:36
  govern these geometric requirements.
• 0:39
  Specifically, we're diving into some key updates in ASMEY
• 0:43
  14.52018, looking at the sort of universal fundamentals of fits
• 0:47
  and limits from ISO 286, and then tackling the really crucial
• 0:51
  mathematics of tolerance stacking.
• 0:53
  That's what dictates whether your manufacturing process
• 0:55
  actually succeeds or just stalls out.
• 0:57
  So moving away from desk work? Exactly.
• 0:59
  We're moving decisively into defined geometric control,
• 1:04
  getting away from that tribal knowledge approach.
• 1:07
  OK, let's unpack this right away then.
• 1:09
  When that design sheet, you know, the perfect CAD model,
• 1:12
  actually hits the shop floor, right?
• 1:14
  What happens? Because often there seems to be
• 1:16
  this massive, sometimes expensive disconnect right
• 1:20
  between the ideal theory and the measurable functional reality.
• 1:23
  Oh, absolutely. It's almost unavoidable.
• 1:26
  The engineering drawing, it captures the design
• 1:28
  requirements, sure. But a standard like ASMEY 14.5
• 1:32
  for instance, it only defines the theoretical tolerance zones,
• 1:35
  just the zones. So developing an exact
• 1:38
  measurement method that aligns with reality and you have to
• 1:40
  factor in business needs like cost, liability, speed, well,
• 1:44
  that requires making assumptions that go beyond just the words in
• 1:47
  the standard itself. Technically, no measurement can
• 1:50
  truly conform perfectly to Y 14.5, but the methods we choose,
• 1:54
  they have to be heavily influenced by it.
• 1:56
  And sometimes, I suppose the human element just overrides the
• 1:59
  standard entirely, which reminds me of that great anecdote you
• 2:02
  shared about the contract inspector.
• 2:04
  Oh yes, the one who basically said I know what the drawing
• 2:07
  says, but function matters more. That's the one, the inspector
• 2:11
  who said so. What story?
• 2:15
  This happened during an audit. They were checking
• 2:17
  perpendicularity, and the inspector deliberately chose to
• 2:21
  measure surface B as the datum, even though the drawing clearly
• 2:25
  specified datum. A deliberately why?
• 2:27
  Well, when they challenged him, he successfully argued that
• 2:31
  because the tolerance was frankly massive .4mm the
• 2:36
  deviation caused by picking the wrong datum, it was negligible
• 2:40
  for how the assembly actually functioned in that specific
• 2:42
  case. Wow, .4 mil is pretty generous.
• 2:45
  It is so it really highlights the ambiguity engineers are
• 2:49
  constantly trying to eliminate. Now, his approach was radical,
• 2:51
  maybe not by the book, but it shows that sometimes the drawing
• 2:54
  really needs to capture the functional manufacturing
• 2:56
  reality, not just some abstract ideal.
• 2:58
  So that brings us to geometric dimensioning and tolerancing GD
• 3:01
  and T That's supposed to be the fix, right?
• 3:03
  To eliminate that kind of costly ambiguity, acting as a sort of
• 3:07
  universal translator across Borders machine shops.
• 3:10
  That's exactly its core function.
• 3:12
  GD and T is globally recognized really as the only effective way
• 3:18
  to properly define geometric requirements and control part
• 3:21
  geometry for reliable meeting functionality
• 3:24
  interchangeability. It works like a concise
• 3:26
  language. It's got symbols, rules, precise
• 3:29
  mathematical definitions. Without it, defining clear
• 3:32
  design specs is, well, it's practically impossible.
• 3:35
  You end up relying on expensive, slow, you know, tribal knowledge
• 3:41
  just to force parts to fit despite a poor drawing.
• 3:44
  OK, so GD and T is the language. Now let's talk about some major
• 3:47
  technical revisions that hit the industry back in 2018 with a
• 3:50
  SMEY 14.5. These weren't just minor tweaks,
• 3:53
  were they? They seem to fundamentally shift
• 3:55
  how engineers should approach design intent.
• 3:57
  That's a good way to put it. They weren't minor.
• 3:59
  The main goal of those 2018 revisions was really about
• 4:02
  eliminating gaps in interpretation.
• 4:04
  Clarifying with a tolerance zone actually covers.
• 4:06
  We can probably boil it down to three really crucial takeaways.
• 4:08
  OK, what's the first one? First, based on clause 4.1 Q, we
• 4:13
  formally moved away from the idea that the as designed
• 4:16
  dimension value, the number you put on the drawing establishes
• 4:19
  the manufacturing target. Wait, hold on.
• 4:21
  If the drawing says 32mm engineers, machinists, they've
• 4:26
  been trained for generations that 32 is the perfect target.
• 4:29
  How does the 2018 standard change that?
• 4:32
  Yeah, it's a mindset shift. It dictates that the statistical
• 4:35
  center of the tolerance zone is actually the target.
• 4:37
  The middle of the zone. Exactly.
• 4:39
  So if your nominal size is 32, maybe with a plus minus
• 4:42
  tolerance, the best part, the ideal part, is the one precisely
• 4:47
  in the middle, Let's say 32.01 if that's the midpoint.
• 4:50
  This officially kills the assumption that the nominal
• 4:53
  dimension must be the process mean.
• 4:54
  It's a really vital clarification if you're
• 4:57
  implementing statistical process control.
• 4:59
  That is a huge change. It sources a complete rethink of
• 5:02
  process entering OK. What was the second major
• 5:04
  clarification? The second one from 4.1 deals
• 5:07
  with surface elements. Surface elements.
• 5:09
  Elements of a surface now explicitly include things like
• 5:12
  surface texture and flaws. You mean like burrs scratches?
• 5:15
  Exactly. Burrs, scratches, any roughness,
• 5:19
  all of it must now be contained within the specified tolerance
• 5:22
  boundaries. This forces designers, QC
• 5:25
  personnel, everyone to view surface finish not as some
• 5:28
  separate requirement over there, but as an integral part of the
• 5:31
  overall size and geometric tolerance.
• 5:34
  OK, that makes sense. And the third major update, you
• 5:36
  said it tackled two of the most well, notoriously confusing
• 5:40
  symbols in the GD and T toolkit. Yes, concentricity and symmetry.
• 5:44
  The standard completely eliminated the term, the
• 5:47
  concept, and the symbol for both with concentricity and symmetry.
• 5:50
  At least got rid of them. Gone.
• 5:51
  The official reason was to eliminate the confusion that
• 5:54
  surrounds these symbols and their misapplication.
• 5:57
  So basically we spent decades trying to master these really
• 6:00
  complex symbols, only for the standards body to essentially
• 6:04
  say, yeah, these are maybe too confusing or redundant.
• 6:06
  That's pretty much it. They were often redundant, yes,
• 6:09
  and frankly almost impossible to measure reliably without very
• 6:12
  specialized, often expensive techniques.
• 6:14
  Position and run out are considered clearer, more
• 6:17
  reliable alternatives for controlling coaxial features.
• 6:20
  They're just easier to verify consistently.
• 6:22
  OK, fair enough. Another area that always felt a
• 6:24
  bit messy and older standards was radius tolerances.
• 6:27
  How did 2018 clean that up? Yeah, radius was ambiguous
• 6:32
  because just calling out R didn't define the center point,
• 6:37
  and R doesn't necessarily mean half the diameter, right?
• 6:39
  So the new rule under 5.16 clarifies it.
• 6:43
  When the center of the radius is located via dimensions, the arcs
• 6:46
  are concentric. So locating the center is key.
• 6:49
  Precisely, if you dimension the center, you create a profile
• 6:52
  tolerant zone that is completely different and much clearer than
• 6:55
  just specifying a radius with an unlocated floating center point.
• 6:59
  These revisions definitely show ASME trying to tighten things
• 7:03
  up, reduce ambiguity, but it does raise that point about the,
• 7:07
  let's call it the global standard situation.
• 7:10
  We've got a SMEY 14.5, mostly North America, and then ISO 11
• 7:15
  O1 dominating elsewhere. Are they still speaking the same
• 7:18
  language fundamentally? That's a great question.
• 7:20
  They're speaking, I'd say 2 increasingly divergent dialects
• 7:24
  of the same language. Divergent.
• 7:26
  How much? Well, while they might have once
• 7:27
  been, say, 90% or even 95% in agreement, fundamental design
• 7:31
  principles are continuing to drift apart.
• 7:34
  Some estimates I've seen suggest maybe only 65% agreement today
• 7:37
  on certain core concepts, so the choice of standard really
• 7:41
  impacts how a drawing is interpreted.
• 7:42
  OK, give us a really clear conflict point, maybe on
• 7:44
  something basic like size. Size tolerance itself is a
• 7:47
  perfect example. ASMEY 14.5.
• 7:50
  It controls both the actual mating size, how it fits, and
• 7:53
  the local size at any 2 points. Both mating and local.
• 7:55
  OK, but ISO 11 O1 by default controls only a 2 point local
• 7:59
  size measurement unless you specifically add the envelope
• 8:01
  symbol, the circle E. So you need the extra symbol in
• 8:04
  ISO to get the same control as Asmes default.
• 8:07
  Exactly, that's a huge functional difference in how
• 8:09
  you'd verify a part for FIT. And don't forget concentricity.
• 8:13
  Even though ASME removed it, the symbol still exists in ISO, but
• 8:17
  it means something different. In ASME it meant control of
• 8:19
  opposing medium points. In ISIL 1101, the same symbol
• 8:23
  refers to a 2D control of a center point, more like
• 8:27
  position. Same symbol, radically different
• 8:29
  geometric requirement. Wow.
• 8:30
  OK. A minefield if you're working
• 8:31
  across standards. That critical distinction
• 8:34
  between controlling local size versus mating size, that leads
• 8:37
  us perfectly into the science of defining size variation itself.
• 8:42
  If the standards even differ on what size is, they must have a
• 8:45
  really robust shared language for tolerance, right?
• 8:48
  Well, yes, so let's shift to the global code for tolerances on
• 8:51
  linear sizes, ISO 286. What's the core vocabulary we
• 8:55
  need to get our heads around? Here.
• 8:56
  Okay, yeah, ISO 286. We need precise terms to talk
• 8:59
  about size variations mathematically.
• 9:01
  First up is the basic size. Think of this as the ideal
• 9:04
  theoretical size right out of your design calculations.
• 9:07
  Maybe 50mm. The perfect number.
• 9:09
  The perfect starting number, yeah.
• 9:10
  Then there's the tolerance. That's the total wiggle room,
• 9:12
  the permissible variation, the difference between the maximum
• 9:15
  and minimum limits. And this tolerance can be
• 9:18
  unilateral, meaning variation is only on one side of the basic
• 9:21
  size. Like plus tolerance only.
• 9:23
  Exactly. Or it could be bilateral, where
• 9:25
  the variation is allowed on both sides, plus and minus.
• 9:28
  OK. And the boundaries of that
• 9:30
  variation. Those are the limits of size.
• 9:32
  Correct, you have the upper limit of size uls and the lower
• 9:35
  limit of size LLS. And crucially, we need to talk
• 9:38
  about limit deviation. Deviation OK.
• 9:41
  This is the algebraic difference between one of those limits of
• 9:44
  size and the corresponding basic size.
• 9:47
  We use specific symbols ES for the upper deviation and EI for
• 9:52
  the lower deviation on internal features like holes and we use
• 9:56
  lowercase ES and EI for external features like shafts.
• 9:59
  Why the different letters? Uppercase for holes, lowercase
• 10:02
  for shafts? Seems a bit fussy.
• 10:05
  But it's key, it's designed to prevent calculation errors when
• 10:09
  you're figuring out the fit between parts by using separate
• 10:12
  symbols. Capital for internal, lowercase
• 10:15
  for external. If still, the ISO system
• 10:17
  basically locks the dimensioning into a specific, non ambiguous
• 10:21
  relationship. The symbol itself tells you
• 10:23
  immediately if you're dealing with a hole or a shaft relative
• 10:26
  to that basic size. Got it.
• 10:28
  Smart. So the ultimate goal of all this
• 10:30
  precision, this careful terminology, is achieving a
• 10:33
  specific fit between mating parts.
• 10:35
  How do we classify those relationships?
• 10:37
  There are three main types, right?
• 10:38
  That's right. We classify them based on
• 10:40
  whether they guarantee a gap, guarantee an overlap, or might
• 10:43
  do either. Let's use the shaft and whole
• 10:45
  example number one, clearance fit.
• 10:47
  This always provides a gap or clearance between the parts.
• 10:51
  Always we can calculate the minimum clearance.
• 10:54
  That's the smallest gap you'll ever get by looking at the
• 10:56
  difference between the minimum hole size and the maximum shaft
• 10:59
  size. And where would you use that?
• 11:00
  Thick precision sliding parts. A common example is an H7H6 fit,
• 11:04
  nice smooth movement. Or maybe an H8E8 for an easier
• 11:07
  running fit with more clearance. OK, clearance means always a
• 11:10
  gap. What's next?
• 11:12
  #2 interference fit is always results in interference, meaning
• 11:17
  the parts have to be forced together like a press fit or a
• 11:20
  shrink fit. The allowance here is actually
• 11:22
  negative. The minimum interference is the
• 11:25
  difference between the maximum hole size and the minimum shaft
• 11:28
  size. Examples.
• 11:29
  Think heavy drive fits like an H7 to six, or even shrink fits
• 11:33
  like HU8 where you heat one part of cool the other.
• 11:36
  Right, so always interference. What's the third type?
• 11:39
  The third is transition fit. This is interesting with the
• 11:41
  middle ground. Here the tolerance intervals of
• 11:43
  the hole and shaft actually overlap.
• 11:45
  Overlap, so it could go either way.
• 11:48
  Exactly. The final assembly might result
• 11:50
  in a clearance, or it might result in interference, just
• 11:52
  depending on the random luck of the draw with the actual
• 11:55
  manufactured sizes. A light press fit, maybe an H7 N
• 11:59
  6 falls into this category. Could be snug, could have a tiny
• 12:02
  gap. OK, that makes sense.
• 12:04
  And you mentioned those H&H symbols to make the system
• 12:07
  efficient. Engineers usually prefer one of
• 12:09
  two basis systems. Correct.
• 12:11
  The whole basis system is generally the preferred choice,
• 12:14
  pretty much standard industry. In this system, the lower
• 12:17
  deviation of the hole symbol H is set to 0.
• 12:20
  Why is that preferred? It simplifies tooling, reduces
• 12:24
  cost. You can use standard drills,
• 12:26
  standard reamers for the holes. Their size is fixed at the basic
• 12:29
  size as the lower limit. Then you just vary the shaft
• 12:32
  tolerance, make the shaft slightly smaller or slightly
• 12:35
  larger to get the specific fit, clearance, transition
• 12:38
  interference you need. OK, standard tools for the whole
• 12:41
  custom shaft. What's the alternative?
• 12:43
  The alternative is the shaft basis system using the symbol H
• 12:47
  where the upper deviation of the shaft is 0.
• 12:50
  This is usually only preferred in specific situations, like
• 12:53
  maybe if you're using standard pre finished shafting material,
• 12:56
  or if you need to mount several different components with
• 12:58
  different types of fits onto one single shaft of a standard
• 13:01
  nominal size. Got it.
• 13:02
  Whole basis is usually easier. OK, so we've defined parts, we
• 13:05
  defined fits. But getting the individual parts
• 13:08
  right is just Step 1, isn't it? Ensuring the final assembly
• 13:11
  actually works functions correctly.
• 13:13
  That's where the complexity really seems to peak.
• 13:17
  Which brings us to tolerance stack analysis.
• 13:20
  This is it. This is the moment of truth for
• 13:22
  a design really Stack up or tolerance Stack analysis deals
• 13:27
  with what happens when the cumulative dimensional
• 13:29
  variations from all the parts in an assembly add up, especially
• 13:33
  when they all happen to be manufactured at the worst end of
• 13:36
  their allowed tolerance range. And that accumulation can
• 13:39
  prevent assembly or stop it from working.
• 13:42
  Absolutely. It can prevent parts from going
• 13:44
  together at all, or it can mean a critical gap disappears, or
• 13:47
  maybe a required alignment is lost.
• 13:49
  Designers have to calculate this cumulative error upfront to
• 13:52
  guarantee that the assembly functions within its limits.
• 13:55
  So how do we calculate it? I understand there are two main
• 13:57
  methods and they represent a kind of direct trade off between
• 14:00
  maybe certainty and cost. The first one is worst case or
• 14:04
  arithmetic tolerancing. Yeah, arithmetic or worst case
• 14:08
  tolerancing. It's definitely the most
• 14:09
  conservative approach you can take.
• 14:11
  It makes the absolute worst assumption that the deviations
• 14:15
  of every single part in the chain will align themselves in
• 14:19
  the absolute worst possible way to cause the maximum possible
• 14:22
  assembly error. So everything is at its Max or
• 14:25
  min limit in the worst direction.
• 14:27
  Everything. And because that total error
• 14:30
  grows linearly with the number of parts in the stack, well,
• 14:33
  this method usually forces you into incredibly tight and
• 14:36
  therefore expensive detail tolerances on the individual
• 14:39
  components. But the upside?
• 14:41
  The huge advantage is certainty. It guarantees mathematically
• 14:44
  that every single assembly produced will fit and function,
• 14:48
  dimensionally speaking, 100%. Certainty OK, guaranteed fit but
• 14:51
  potentially very expensive parts.
• 14:52
  What's the other option? The more popular, cheaper route?
• 14:55
  Statistical tolerancing root sum of squares or Rs.
• 14:59
  That's the one RSS or root sum of squares.
• 15:01
  You could think of it as the the Pythagorean shortcut for
• 15:04
  dimensioning. Pythagorean shortcut, yeah.
• 15:06
  Statistical tolerancing assumes that the variations in
• 15:09
  individual parts are random and independent.
• 15:12
  Because they're random over large batches of assemblies,
• 15:15
  they'll tend to cancel each other out.
• 15:16
  Quite often, a part that's a bit too big might be next to one
• 15:20
  that's a bit too small. So they don't all line up in the
• 15:22
  worst way. Not usually, no.
• 15:25
  So the assembly tolerance instead of growing linearly with
• 15:28
  N parts, it grows much slower. It's proportional to the square
• 15:32
  root of the number of parts. Square root of N versus N That's
• 15:35
  a big difference. Huge difference.
• 15:37
  This allows for significantly looser and therefore much
• 15:40
  cheaper detail tolerances on your components.
• 15:43
  But there's always a but, isn't there?
• 15:44
  That benefit must come with a major caveat.
• 15:47
  RSS doesn't give you that 100% certainty, right?
• 15:49
  Correct. You trade certainty for cost.
• 15:52
  The basic RSS method is, well, inherently optimistic.
• 15:56
  It relies on some pretty stringent assumptions being
• 15:58
  true. Like what?
• 16:00
  Like the variations must be statistically independent, and
• 16:02
  crucially, the manufacturing process for each part must be
• 16:05
  perfectly centered, meaning the average size produced is exactly
• 16:09
  at the midpoint of the tolerance zone.
• 16:11
  And if those are true? Even if those conditions are
• 16:13
  perfectly met, basic RSS typically only guarantees
• 16:17
  something like 99.73% assembly fit, assuming a normal
• 16:21
  distribution. That's the six Sigma concept ±3
• 16:25
  standard deviations, but it's not absolute 100% certainty like
• 16:30
  worst case. 99.73% sounds pretty good, but not 100%.
• 16:34
  What happens in the real world when those assumptions get
• 16:37
  violated, particularly when the manufacturing process isn't
• 16:40
  perfectly centered? When it shifts, we hear about
• 16:42
  mean shifts. Mean shifts, yes, that's a
• 16:45
  critical real world factor. Mean shifts are systematic
• 16:48
  biases. This happens when the average
• 16:50
  size of the machine is actually producing starts to drift away
• 16:53
  from that perfect specified target size you put on the
• 16:56
  drawing. Maybe tool wear?
• 16:57
  Maybe temperature changes? And this isn't random.
• 16:59
  No, that's the key. This bias is often systematic
• 17:02
  and persistent across hundreds, maybe thousands of assemblies.
• 17:05
  It's not like random part to part variation that cancels out.
• 17:08
  So a more robust kind of hybrid analysis acknowledges this
• 17:12
  reality. How does that work?
• 17:13
  You stack the systematic mean shifts arithmetically.
• 17:16
  Treat them as worst case because they are constant for a batch
• 17:20
  and then you only stack the remaining random variation
• 17:23
  statistically using RSS. So worst case for the known
• 17:26
  drift, RSS for the random noise around it.
• 17:29
  Exactly. That gives you a much more
• 17:31
  realistic error boundary than pure RSS, while still being less
• 17:34
  demanding than pure worst case. It's a common practical
• 17:38
  approach. So what does this all really
• 17:40
  mean for you, the designer listening in?
• 17:42
  Well, designing robust parts. It requires deep fluency in the
• 17:46
  standards, whether you're using ASME or ISO and making
• 17:49
  calculated choices. You have to constantly balance
• 17:51
  that absolute certainty you get from worst case tolerance
• 17:54
  stacking against the very real cost benefits of statistical
• 17:58
  methods. Yeah, and remember, tolerances
• 18:00
  are never just arbitrary numbers you pull out of thin air.
• 18:03
  They form the enforceable legal contract between the design
• 18:06
  team, the customer and the supplier regarding the products
• 18:08
  performance of fit. And it's worth remembering too
• 18:11
  that implementing complex geometric requirements, things
• 18:14
  like true position run out profile tolerances, often
• 18:18
  requires advanced inspection equipment.
• 18:20
  We're talking coordinate measuring machines CMMS to
• 18:23
  actually verify them properly. Absolutely, verification is key.
• 18:26
  So maybe the most important question for you as the design
• 18:29
  engineer isn't just what tolerance is required.
• 18:32
  Maybe it's more like, what is the most robust design
• 18:35
  alternative I can think of? One that maximizes the
• 18:37
  manufacturing tolerance, keeps costs down without sacrificing
• 18:42
  the reliability of the final product.
• 18:43
  That's the ultimate goal. Keep those tolerances as loose
• 18:46
  as you possibly can, and focus your recision your tightest
• 18:49
  controls only where the actual physics of function absolutely
• 18:52
  demand it.