Latest / Mechanical Engineering Made Simple / (#40) Topology Optimization to Angstrom Repeatability: Mastering the Full Mechanical Engineering Lifecycle

Transcript

(#40) Topology Optimization to Angstrom Repeatability: Mastering the Full Mechanical Engineering Lifecycle · Mechanical Engineering Made Simple

• 0:00
  Welcome back, design junkies, to Mechanical Engineering Made
• 0:02
  Simple. Today we're doing a deep dive
• 0:04
  into, well, pretty much the whole engineering life cycle.
• 0:07
  We're looking at everything, I mean from the really abstract,
• 0:10
  like theoretical performance modeling, you know, the stuff
• 0:13
  done on supercomputers, all the way down to the the nitty gritty
• 0:18
  specs for fitting parts together, pins, bushings, bolts,
• 0:22
  the works. Yeah, the idea is to give you a
• 0:24
  kind of comprehensive shortcut. How are top tier components
• 0:28
  really designed and built efficiently?
• 0:31
  And if you boil it all down, I think the core theme that
• 0:33
  emerged from everything we looked at is just seamless
• 0:37
  efficiency. Every single decision, whether
• 0:39
  you're talking about maximizing strength or, you know, getting
• 0:41
  motion down to the angstrom level, it's all driven by
• 0:44
  getting the most performance out of the least resources,
• 0:47
  constantly pushing for maximum output from minimum input.
• 0:50
  OK, let's start unpacking that and maybe start at the most
• 0:53
  abstract and topology optimization for listeners who
• 0:56
  maybe haven't used the software. What's this wizardry actually
• 1:00
  trying to do? Right.
• 1:02
  Topology optimization is basically that efficiency theme
• 1:04
  taken to its extreme. You're trying to maximize a
• 1:07
  pert's structural performance, stiffness, low capacity,
• 1:10
  whatever, while using the absolute minimum amount of
• 1:15
  material. And this isn't just academic.
• 1:17
  Less material means, well, direct cost savings in
• 1:20
  manufacturing, obviously, but also a lighter environmental
• 1:23
  footprint, big savings there. It sounds amazing, but the
• 1:26
  shapes that come out of it, they often look really organic,
• 1:29
  almost alien, like a complex honeycomb or something.
• 1:32
  They don't look like something you design traditionally.
• 1:34
  Yeah. How do engineers actually arrive
• 1:36
  at those shapes without just making, you know, a million
• 1:39
  physical prototypes? Yeah, you definitely wouldn't
• 1:41
  want to do it that way. Yeah, it relies heavily on
• 1:44
  advanced computation. Our sources point to
• 1:46
  optimization techniques like genetic algorithms, particle
• 1:50
  swarm optimization, things like that.
• 1:53
  These algorithms essentially run millions of simulated design
• 1:56
  variations inside the computer to kind of evolve the best shape
• 2:00
  for the job. The huge benefit is skipping all
• 2:03
  that costly, time consuming physical trial and error that
• 2:06
  you know traditional design often got bogged down in.
• 2:09
  Really speeds up the concept stage.
• 2:10
  But I wonder, does that super complex optimize shape, while
• 2:14
  it's great for material use, make it incredibly hard or
• 2:16
  expensive to actually make the part?
• 2:18
  Using traditional machining, for instance, seems like a potential
• 2:21
  trade off. Oh, it absolutely can be.
• 2:23
  That's a really important point and it's why topology
• 2:25
  optimization is so often paired with additive manufacturing, you
• 2:29
  know, 3D printing. But you know, something crucial
• 2:32
  there, which leads us to maybe the biggest hurdle for wider
• 2:35
  adoption right now, standardization.
• 2:37
  See different fields, aerospace, medical, automotive, they have
• 2:41
  wildly different needs to extreme temperatures in space
• 2:44
  versus like high volume consumer goods.
• 2:48
  And because those requirements and the material properties
• 2:51
  involved yield, strength, densities, fatigue, life are so
• 2:54
  varied, there's just no single uniform set of optimization
• 2:57
  rules that works everywhere. That lack of common protocols is
• 3:00
  really what holds back broader, more consistent use across all
• 3:04
  industries. That makes total sense.
• 3:05
  OK, let's shift gears then, from optimizing static structures to
• 3:09
  getting really precise, really reliable motion.
• 3:11
  This brings us to flexures. When you need super smooth, fine
• 3:14
  movement, why do flexures often win out over, say, traditional
• 3:17
  bearings? Well, the big advantage with
• 3:19
  flexures is that inherent smoothness you mentioned.
• 3:21
  And also there they're near elastic operation.
• 3:26
  Unlike bearings, there's no rolling or sliding friction,
• 3:29
  right? The movement comes purely from
• 3:31
  the material bending elastically.
• 3:33
  And that high linearity is just absolutely vital for achieving
• 3:37
  the kind of extreme repeatability needed in things
• 3:39
  like semiconductor lithography or nanotech manufacturing.
• 3:44
  Right, And when we start talking formally about how mechanisms
• 3:47
  move, we kind of leave intuition at the door and get into the
• 3:50
  math of constraints. I always need a refresher on the
• 3:53
  rule that governs these kinematics systems.
• 3:55
  Can you walk us through the basics again?
• 3:57
  Sure. We use kinematics to really
• 3:59
  quantify and control motion precisely, and the common
• 4:02
  approach relies on Blanding's rule of complementary patterns.
• 4:06
  It's kind of fundamental equation for designing
• 4:08
  constrained motion. The key idea is this
• 4:11
  relationship. The residual freedom, let's call
• 4:13
  it R, has to equal 6 minus the number of constraints applied C
• 4:17
  So r = 6 -, C. 6 -, C Got it. Yeah, so if you want something
• 4:21
  to move only in one specific direction, like a piston and a
• 4:24
  cylinder, you need to apply exactly 5 constraints.
• 4:29
  That leaves it with just one degree of freedom, r = 1.
• 4:32
  This mathematical approach ensures you get only the motion
• 4:35
  you want and nothing else. No wobble, no unwanted shifts.
• 4:38
  OK. And the sources mentioned
• 4:40
  angstrom level repeatability that's, I mean that's an
• 4:43
  incredibly tiny scale, almost hard to imagine what sort of
• 4:46
  perfect storm of conditions needs to happen for these
• 4:49
  bending components, these flexures to perform that
• 4:52
  precisely without feeling. Angstrom level is amazing.
• 4:56
  To get there, you basically have to eliminate almost all energy
• 4:59
  loss and any potential material degradation.
• 5:02
  So first, the material needs incredibly low hysteresis.
• 5:06
  That means it springs back perfectly after bending without
• 5:09
  losing energy internally as heat.
• 5:10
  OK, low hysteresis. Second, you absolutely cannot
• 5:13
  have any dislocation motion within the materials crystal
• 5:16
  structure. So the stress you apply Sigma
• 5:19
  has to be way way below the materials yield stress Sigma
• 5:22
  yield. Much less than yield makes
• 5:24
  sense. And finally, how you assemble it
• 5:27
  is critical. There can be 0 micro slip or
• 5:30
  friction in any mounting points or joints, and the parts
• 5:32
  themselves can't yield even a tiny bit during assembly.
• 5:35
  It all has to stay purely elastic.
• 5:37
  Wow. OK.
• 5:38
  So when you're picking the material for that kind of
• 5:40
  demanding job, it sounds like you need to look beyond just the
• 5:43
  basic strength number, the yield strength.
• 5:46
  Why the focus on these normalized property ratios?
• 5:49
  That's a great question. For flexures, what we really
• 5:52
  care about is the maximum elastic strain energy the
• 5:54
  material can store before it deforms permanently, and the
• 5:58
  ratio that captures this potential is the yield stress
• 6:01
  divided by the Young's modulus. So signal Y over E.
• 6:04
  Sigma Y over E OK. Yeah.
• 6:05
  That ratio, it tells you how much of material can bend
• 6:07
  elastically relative to how stiff it is.
• 6:10
  So for example, titanium V comes out really well with a ratio
• 6:13
  around 1.00, whereas something like aluminum 7075 is quite a
• 6:17
  bit lower, maybe .70. Using these normalized ratios
• 6:20
  helps designers push the performance like the range of
• 6:23
  motion or stroke, while balancing other factors like
• 6:26
  density or cost. All right, let's move down to
• 6:29
  maybe the most fundamental way we join things the bolted joint.
• 6:33
  This seems like a classic sticking point for engineers
• 6:36
  starting out. When you preload a bolt, clamp
• 6:39
  parts together. How should we think about the
• 6:41
  stiffness? Are the flange and the bolt
• 6:43
  acting in parallel or in series? I yes, the bolted joint puzzle
• 6:48
  during the initial preloading. They act in a serious
• 6:50
  relationship. That's the crucial part to
• 6:52
  remember. Think about springs in series.
• 6:55
  They both see the exact same force, which is the preload
• 6:58
  tension. But because they have different
• 7:00
  stiffnesses, the bolt stretches a certain amount and the flange
• 7:04
  compresses a different. Amount OK, same force, different
• 7:07
  deformation series. Exactly.
• 7:09
  If you incorrectly model them as parallel, your calculation for
• 7:12
  the overall stiffness of the joint will be completely off.
• 7:15
  Right. OK, Now what if that preloaded
• 7:17
  joint also has to resist shear forces?
• 7:20
  You know, forces trying to slide the two parts sideways against
• 7:23
  each other. The gut feeling might be that
• 7:25
  the bolt Shank itself is taking that sideways load.
• 7:27
  Why is that often wrong? That's a really critical piece
• 7:29
  of engineering understanding. In a properly preloaded joint
• 7:33
  under shear, it's overwhelmingly the friction between the clamped
• 7:37
  surfaces that resists the load. That friction is generated by
• 7:40
  the huge clamping force from the preload.
• 7:43
  So the coefficient of friction and getting that preload right
• 7:46
  are absolutely paramount. So the bolt isn't really acting
• 7:49
  like a pin in shear. Not primarily, no.
• 7:52
  If you need direct shear resistance, maybe because of
• 7:55
  high vibration or impact loads where friction might be
• 7:57
  overcome, then you have to add features specifically for that
• 8:00
  like dowel pins or shoulder bolts.
• 8:02
  They are designed to take shear load directly and bearing.
• 8:05
  Got it. And kind of underpinning all
• 8:07
  this assembly work is the actual language of manufacturing, the
• 8:11
  tolerance standards. We have to specify dimensions
• 8:13
  with total clarity. Absolutely.
• 8:15
  The precise communication is key.
• 8:17
  Manufacturing runs on detailed specs.
• 8:19
  It covers everything you know, from the different classes of
• 8:21
  screw threads like unified threads here in the US or metric
• 8:24
  threads elsewhere, all the way to classifying how parts
• 8:27
  physically fit together. It's the difference between
• 8:29
  something that rattles loose and something that's basically a
• 8:32
  permanent bond. And those fits have specific
• 8:35
  names. They do.
• 8:36
  We generally talk about 3 main types.
• 8:38
  Clearance fits where there's always a design gap between the
• 8:41
  parts. Transition fits, where depending
• 8:43
  on the actual tolerances, you might get a small clearance or
• 8:46
  small interference, and then the critical ones are interference
• 8:50
  fits. These are things like force or
• 8:52
  shrink fits, often designated FN 1FN2FN3 and so on.
• 8:56
  FN One. FN 2.
• 8:57
  Yeah, I mean, a parts are intentionally made so one is
• 9:00
  slightly larger than the hole it goes into.
• 9:02
  You have to force them together, maybe by heating one or cooling
• 9:04
  the other. The FN number indicates how much
• 9:07
  interference, how tight that fit is.
• 9:09
  Creates a very strong solid joint.
• 9:11
  OK, bring it back to the preload for a second.
• 9:13
  We need to actually measure that tension to make sure we get the
• 9:15
  clamping force, the friction we need.
• 9:18
  The sources mentioned three main ways to do it, but only 1 is
• 9:21
  really easy in practice. What's the trade off?
• 9:24
  Right, you got three options. Method one, directly measure how
• 9:28
  much the bolt stretches. Very accurate, but usually
• 9:31
  impractical in the field. Method 2.
• 9:35
  Use bolts with embedded strain sensors, also very accurate but
• 9:39
  expensive. Method 3 Measure the applied
• 9:43
  torque needed to tighten the bolt.
• 9:45
  This is by far the easiest and fastest method, which is why
• 9:48
  it's so common. Well, it's not perfect.
• 9:50
  Not for ultra repeatability, no. The big drawback with cork is
• 9:54
  that a huge chunk of the effort you measure, maybe 80 or 90%,
• 9:57
  isn't actually creating tension, it's just overcoming friction.
• 10:00
  Friction in the threads and under the bolt head.
• 10:02
  Exactly. And that friction can vary
• 10:04
  wildly depending on the lubrication, surface, finish
• 10:06
  cleanliness, even temperature. So using torque is fast and
• 10:10
  convenient, but you sacrifice precise control over the actual
• 10:13
  tension achieved. OK, let's switch from static
• 10:16
  joints now to dynamic systems. Belt drives are everywhere.
• 10:20
  When you're trying to optimize these for efficiency, noise
• 10:23
  seems like a really big deal. Are some types of synchronous
• 10:26
  belts just inherently quieter? Oh definitely.
• 10:28
  As a general rule, synchronous belts someones with teeth tend
• 10:31
  to be noisier than say V belts just because of that meshing
• 10:35
  action. But yeah, there are big
• 10:37
  differences. Advanced designs like the Power
• 10:39
  Grip UD3 drives mentioned in the sources are specifically
• 10:42
  engineered to be much quieter than older synchronous designs.
• 10:46
  They have improved tooth profiles and meshing
• 10:48
  characteristics that really reduce that impact noise as the
• 10:51
  teeth engage the sprocket. And if there's 1 guaranteed way
• 10:55
  to make any drive system noisy and fail prematurely, it's
• 10:59
  misalignment, right? What's the kind of critical
• 11:01
  tolerance limit engineers need to stick to their?
• 11:04
  Misalignment is the absolute enemy.
• 11:06
  Whether it's angular, the shafts aren't parallel or parallel
• 11:08
  offset, the sprockets aren't lined up, it causes terrible
• 11:11
  noise, uneven wear, high temperatures, belt tracking
• 11:15
  issues, it's just bad news. Well, what's the magic number?
• 11:18
  The rule of thumb we found is pretty tight.
• 11:21
  Misalignment shouldn't exceed about 1/4 of a degree angularly.
• 11:25
  Or if you think about it linearly, that's roughly 116th
• 11:28
  of an inch offset for every foot of distance between the shafts.
• 11:31
  Wow, that's not much room for error.
• 11:33
  No, it isn't. If you go beyond that, the belt
• 11:36
  will try to climb the flange of the sprocket, generating heat
• 11:39
  and noise and wearing out very quickly.
• 11:42
  OK, shifting over to chains now, they often play that dual role,
• 11:46
  right? Power transmission, but also
• 11:47
  just conveying things. What are the key wear limits we
• 11:51
  look for when deciding it's time to replace a chain?
• 11:54
  Yeah, chain manufacturer are usually quite specific about
• 11:56
  this. It's based on wear elongation,
• 11:58
  basically how much the chain has stretched because the pins and
• 12:01
  bushings have worn down. For chains used in high speed
• 12:04
  power transmission, the typical replacement point is when that
• 12:07
  wear elongation hits 1.5% of the original length. 1 1/2% OK.
• 12:12
  For conveyor chains, which usually see lower speeds and
• 12:15
  stresses, the limit is often a bit looser, maybe around 2%
• 12:19
  elongation. Exceeding those limits risks
• 12:22
  chain jumping or breakage. We also saw some really
• 12:25
  interesting examples of specialized chains, particularly
• 12:28
  ones designed to run without any lubrication.
• 12:30
  Yes, the Lambda roller chains were a fascinating example of
• 12:34
  material science in action. They used bushings made from
• 12:37
  centered metal that's impregnated with oil, and the
• 12:40
  pins often have special coatings.
• 12:42
  This design lets them run without external lubrication and
• 12:45
  at lower speeds. They can apparently outlast
• 12:48
  standard chains needing Lube by maybe up to 30 times.
• 12:52
  Huge maintenance saver, especially in hard to reach
• 12:54
  spots. That's impressive.
• 12:56
  And linking back to noise again, why is it that larger pitch
• 12:59
  chain systems tend to be noisier than smaller pitch ones?
• 13:02
  Our sources confirmed this generally comes down to
• 13:04
  something called chordal action. Chordal action.
• 13:06
  Yet picture the chain links engaging with the sprocket
• 13:09
  teeth. Because the sprocket is a
• 13:11
  Polygon, not a perfect circle, the chain path actually rises
• 13:14
  and falls slightly as each link engages and disengages.
• 13:18
  That up and down oscillation is chordal action.
• 13:21
  With larger pitch chains, the links are longer, the sprocket
• 13:24
  flats are longer, and this chordal rise and fall is just
• 13:26
  more pronounced. It creates more vibration and
• 13:28
  therefore more noise. That also tends to accelerate
• 13:31
  pin and bushing wear, which adds to the noise too.
• 13:34
  OK, wow. We've covered a ton of ground
• 13:36
  here. We went from, you know, abstract
• 13:38
  computer optimization of shapes all the way down to the friction
• 13:41
  in a single bolt thread. When you step back and look at
• 13:43
  all of it. What's the big take away for the
• 13:45
  design junkie listening right now?
• 13:47
  I think the biggest take away is that truly successful high
• 13:50
  performance mechanical engineering demands this
• 13:53
  absolutely seamless integration of everything.
• 13:56
  You have to think about the optimal shape the topology
• 13:58
  algorithm gives you and ensure your flexure hits that angstrom
• 14:01
  level repeatability. Using materials with just the
• 14:04
  right Sigma Y over E ratio and still managed to get the preload
• 14:09
  and friction exactly right on every single fastener holding
• 14:12
  the whole thing together. It's all connected.
• 14:15
  High performance isn't magic. It relies entirely on rigorously
• 14:18
  applying precise standards and understanding the details.
• 14:21
  The bore tolerance on a pulley, the groove profile on a
• 14:24
  sprocket, the class of fit for a pin, everything matters.
• 14:28
  And that leads perfectly into a final thought, because it's easy
• 14:30
  to stop at the spec sheet, right?
• 14:31
  The supplier tells you what the part does.
• 14:33
  But let's think about the manufacturing process itself for
• 14:36
  a second. Consider corrosion resistant
• 14:38
  chains, maybe ones that get chemically plated for
• 14:40
  protection. Plating sounds great, adds
• 14:42
  resistance. But that planning process often
• 14:44
  involves high temperatures. Right.
• 14:47
  And that heat exposure can actually change the physical
• 14:49
  parts in subtle ways. If those chains, for instance,
• 14:52
  are plated after they're assembled, that heat can cause
• 14:55
  the press fit components, the pins forced into the side
• 14:58
  plates, the bushings pressed into the links to actually lose
• 15:01
  some of that critical interference fit they were
• 15:03
  designed with. The heat allows things to relax
• 15:06
  a tiny bit. So here's the trade off to
• 15:08
  ponder. How does a process designed to
• 15:11
  add a benefit like corrosion resistance via plating
• 15:13
  potentially reduce the fundamental strength
• 15:16
  characteristic that interference fit, which in turn might lower
• 15:19
  the chains maximum load capacity or even shorten its wear life
• 15:23
  compared to an unplated, properly assembled chain?
• 15:26
  It forces you, the designer, to look beyond just the final spec
• 15:30
  and consider the entire manufacturing life cycle and its
• 15:32
  potential hidden impacts. That's your deep dive homework
• 15:35
  for this week. Definitely some food for thought
• 15:37
  there. Until next time.