Latest / Mechanical Engineering Made Simple / (#42) Micro-Precision to Catastrophic Failure: Essential Standards for Component Design, Tolerance Stack-Up, and Thermal Stress

Transcript

(#42) Micro-Precision to Catastrophic Failure: Essential Standards for Component Design, Tolerance Stack-Up, and Thermal Stress · Mechanical Engineering Made Simple

• 0:00
  Welcome back design junkies to Mechanical Engineering Made
• 0:02
  Simple, our deep dive today. Well, it tackles the really
• 0:06
  foundational stuff and non negotiable principles of
• 0:09
  engineering components. We're talking everything from
• 0:12
  how we specify a fit that's, you know, absolutely perfect down to
• 0:15
  the Micron almost to how we handle massive torque transfer
• 0:19
  and crucially, those kind of devastating effects of thermal
• 0:22
  stress. We are going deep today,
• 0:25
  starting high level with manufacturing standards, then
• 0:27
  drilling right down to the nitty gritty mechanical constraints
• 0:30
  that make or break a machine. That's right.
• 0:33
  And if you want the shortcut, I guess, to real mechanical
• 0:36
  literacy, the mission today is pretty simple.
• 0:38
  We're aiming to give you the blueprints for understanding why
• 0:41
  parts fail sometimes, how power gets moved optimally through a
• 0:44
  system, and what specific engineering standards actually
• 0:48
  exist to stop those expensive and sometimes really dangerous
• 0:52
  assembly errors. Yeah, our sources today, they're
• 0:55
  quite a stack, covering everything from, you know, the
• 0:57
  complexities of tolerance stack up equations all the way to the
• 1:00
  specific advantages of newer synchronous built profiles.
• 1:03
  OK, let's unpack this then, starting right at the basics.
• 1:06
  Dimensional control and the critical importance of having
• 1:12
  standards in the 1st place. We really rely on them now, but
• 1:15
  the history, especially in mass production, shows just how vital
• 1:18
  these decisions were. Think back maybe to World War 2.
• 1:21
  the US military was churning out equipment, spare parts, just
• 1:25
  huge scale, often different contractors all over the place.
• 1:29
  And the sources note this really serious problem.
• 1:32
  Parts made exactly to the drawings to the specs they were
• 1:34
  given. They often just wouldn't
• 1:36
  assemble when they arrived overseas.
• 1:37
  Yeah, that inefficiency was, well, it was catastrophic back
• 1:40
  then. It really proved that just
• 1:42
  specifying a size that's not enough.
• 1:44
  You absolutely have to specify the variation allowed in that
• 1:47
  size, the tolerance. That whole crisis LED pretty
• 1:50
  directly to the tolerance standardization we take for
• 1:52
  granted today. So when engineers look at
• 1:54
  dimensional pollers now, they usually think along 2 main
• 1:57
  lines, 2 core philosophies. The first one, maybe the most
• 2:00
  straightforward, is called functional limits.
• 2:02
  Which sounds pretty much like what it is, I guess.
• 2:05
  Exactly. These are just the tolerances
• 2:07
  strictly needed for a part to physically assemble and do its
• 2:10
  job correctly. That's it.
• 2:12
  If it needs to hold a shaft, the whole tolerance is purely about
• 2:15
  holding that shaft, nothing more.
• 2:17
  OK. And the second philosophy, it
• 2:19
  goes beyond just function, right?
• 2:20
  It does. That's inspection driven design.
• 2:22
  Now this one's interesting. Here, the dimensional tolerance
• 2:26
  has come not just from performance needs, but also from
• 2:28
  how you plan to measure the part.
• 2:30
  It considers the measurement technique itself and
• 2:33
  importantly, the inherent measurement uncertainty of the
• 2:36
  gear you're using to inspect the finished part.
• 2:39
  It basically acknowledges that look, if you can't measure it
• 2:42
  accurately out in the field or on the shop floor, then that
• 2:45
  super tight tolerance on the drawing is kind of meaningless,
• 2:48
  isn't it? That sounds like designing
• 2:50
  almost for compromise though, like prioritizing ease of
• 2:53
  measurement. Doesn't that risk maybe
• 2:55
  substandard performance if the actual functional need is
• 2:58
  tighter than what you can reliably measure?
• 3:00
  That's the balancing act. It absolutely requires careful
• 3:03
  thought. The goal isn't really to
• 3:04
  compromise performance, but to make sure manufacturability and
• 3:08
  inspectability actually line up with the function.
• 3:11
  I mean, if your inspection uncertainty is bigger than your
• 3:13
  functional tolerance band, your design is essentially untestable
• 3:17
  in practice and that introduces massive risk down the line.
• 3:20
  Right O whichever philosophy you lean towards these tolerances,
• 3:24
  they add U in a real machine and that gets us to the the dreaded
• 3:28
  tolerance stack up for you listening.
• 3:30
  This is where engineers have to trace what's called a derived
• 3:33
  stack path. It's conceptual really,
• 3:35
  following how dimensions link up to your mating surfaces.
• 3:38
  These surfaces often get internal labels.
• 3:40
  You might see things like no dollar $2.00 and $2.00 on
• 3:42
  drawings. Just markers to follow the
• 3:44
  chain. Yeah, those markers just help
• 3:46
  track the potential error accumulation.
• 3:48
  You analyze every single dimension that contributes to
• 3:51
  the final Torrance or interference, whatever it is
• 3:53
  you're looking at. This leads you to an overall
• 3:56
  maximum and a minimum possible value that clearance.
• 4:01
  So if we, you know, crunch the numbers for a typical assembly,
• 4:04
  we might find a final stack result like X = 25 point o ±
• 4:10
  0.66mm that .66 that represents the absolute worst case
• 4:16
  accumulation of all the little variations from every single
• 4:20
  part in that chain. And if that worst case still
• 4:22
  allows assembly and function, then OK, the design passes that
• 4:25
  check. And this is exactly why
• 4:27
  standardized fits are so crucial, especially when we look
• 4:29
  at something ubiquitous like screw threads.
• 4:31
  You've got ANSI and ISO standards defining thread
• 4:34
  classes. They exist purely to guarantee
• 4:37
  that a bolt from manufacturer A will fit a nut from manufacturer
• 4:40
  B. Interchangeability.
• 4:41
  Right, and the most common classes you'll bump into for
• 4:43
  just general applications are class 2A for external threads.
• 4:46
  So think bolt screws and class 2B for internal threads like
• 4:49
  nuts or tapped holes. OK. 2A external. 2B Internal.
• 4:53
  And what's really particle here is the subtle difference between
• 4:55
  that A&B, the class 2A, the external thread, includes the
• 4:59
  necessary allowance. This allowance means the maximum
• 5:02
  possible diameters of the external threads are
• 5:04
  deliberately made a little bit less than the basic nominal
• 5:07
  diameter. Why?
• 5:10
  Why build in that that deliberate looseness, I guess
• 5:13
  for the? Bolt.
• 5:14
  Well, it's there's a couple of really powerful purposes, things
• 5:17
  that help assembly and stop failures later.
• 5:20
  First, it really minimizes galling and seizing.
• 5:23
  Galling is, well, it's when tiny bits of metal tear off the
• 5:26
  thread surfaces and it basically Weld themselves together under
• 5:29
  pressure. Yeah, it's a common disaster,
• 5:32
  especially with power tools during assembly.
• 5:34
  That allowance gives a bit of breathing room.
• 5:36
  Second, it accommodates things like plated finishes or
• 5:38
  coatings. You can plate the bolt and it
• 5:40
  still fits because you had that initial clearance.
• 5:43
  It won't exceed the tolerance limits after plating.
• 5:46
  OK, so the external part, the bolt is intentionally a bit
• 5:49
  undersized, prevents that friction welding, and leaves
• 5:52
  room for coatings. Makes sense?
• 5:54
  Exactly. Now contrast that with class 2B,
• 5:56
  the internal threat, the nut. It's minimum diameters are set
• 6:00
  as basic. What that means is there's no
• 6:03
  initial allowance provided, no clearance built in.
• 6:06
  When both parts are made right at their maximum material
• 6:09
  limits, they're tight as possible condition.
• 6:12
  So the nut at its biggest, heaviest metal condition is
• 6:15
  essentially right on the theoretical size.
• 6:17
  No wiggle room added there. Got it.
• 6:19
  So that gives you really robust, tight fit when that's what you
• 6:22
  actually need. OK, let's shift gears a bit.
• 6:24
  We've kind of locked down the perfect static fit, but the next
• 6:27
  big challenge is making sure that perfect fit moves exactly
• 6:30
  how we want it to, and just as importantly, how it doesn't move
• 6:33
  in ways we don't want. We're moving from statics to
• 6:35
  dynamics component constraints. Yeah, this is that classic 6° of
• 6:40
  freedom problem, isn't it? For many machines you need to
• 6:43
  rigidly constrain motion in five axes, say up, down Y and out Z,
• 6:48
  and rotation around all three axes XYZ, while allowing smooth
• 6:52
  free movement in only one axis. Maybe a long axe.
• 6:56
  Think of a high precision machine carriage, like on a
• 6:58
  lathe or a milling machine. You want it to glide
• 7:01
  effortlessly along the ways X, but be incredibly stiff against
• 7:05
  any cutting forces pushing it sideways Y or vertically Z.
• 7:08
  And in the real world, trying to get that perfect constraint, it
• 7:11
  runs into all sorts of practical issues like slight misalignment
• 7:15
  during assembly, maybe some errors and how it's put
• 7:18
  together, and crucially, thermal effects heat.
• 7:21
  Yes, a really critical practical problem our sources highlighted
• 7:25
  deals with this thermally induced growth.
• 7:28
  It's particularly nasty and high speed, high load situations like
• 7:31
  bearing configurations. Thermal growth can be this sort
• 7:33
  of silent killer, a systemic failure mechanism just waiting
• 7:36
  to happen in a high performance system.
• 7:38
  Can you maybe elaborate on how severe this can be?
• 7:40
  There was a pretty stark warning in the source material, wasn't
• 7:43
  there? There really was.
• 7:44
  Let's consider one example set up, maybe example three from the
• 7:48
  notes. It's a configuration constrained
• 7:51
  radially and also what's called half axially.
• 7:55
  Now to manage the inevitable thermal expansion and
• 7:57
  contraction as things heat up and cool down, you might decide
• 8:00
  to add a spring preloader. But OK, adding that spring,
• 8:03
  well, it increases your part count, adds cost, and maybe most
• 8:06
  critically, it often introduces asymmetry into the system.
• 8:10
  You end up with different axial stiffness values, pushing left
• 8:13
  versus pushing right. That's a huge engineering trade
• 8:16
  off you've just introduced. And what about the really
• 8:18
  extreme cases? The warning seemed pretty dire.
• 8:21
  It was the sources explicitly warned that at high speeds and
• 8:25
  high loads, thermal growth, especially if you have rigid
• 8:27
  bearing configurations like a double face to face arrangement.
• 8:30
  That was example 5. It may literally kill the
• 8:33
  bearing sets kill. The bearing sets.
• 8:34
  Wow. Yeah, it's because as the
• 8:37
  temperature climbs, the shafts expand, the housings expand.
• 8:41
  If the bearings are rigidly locked in place axially, that
• 8:44
  expansion has nowhere to go except into putting immense
• 8:47
  axial preload onto the bearings themselves.
• 8:50
  That preload pushes the balls or rollers into the races way, way
• 8:54
  harder than they were ever designed for.
• 8:56
  That leads directly to rapid fatigue, overheating, and pretty
• 8:59
  quickly catastrophic failure. So the system basically squeezes
• 9:03
  itself to death just by getting hot during normal operations.
• 9:06
  That's a good way to put. It OK so to try and get around
• 9:08
  these issues with rigid constraints, especially in very
• 9:11
  high precision environments. The source has also mentioned
• 9:14
  alternatives like flexures. Flexures.
• 9:17
  Yes. They're quite elegant solutions,
• 9:19
  really. A flexure is typically
• 9:20
  monolithic, meaning it's machined from a single piece of
• 9:23
  material. It's cleverly designed with thin
• 9:25
  sections that allow it to be very stiff in some directions,
• 9:28
  say Y&Z, but compliant, able to bend elastically in another
• 9:32
  direction like X. So it replaces complex sliding
• 9:35
  mechanisms which always have issues with friction, where
• 9:39
  backlash diction replaces all that with pure elastic bending.
• 9:43
  No sliding, no backlash, very precise.
• 9:46
  OK, now let's accelerate into power transmission.
• 9:50
  How do we actually move torque cleanly from one place to
• 9:52
  another? Let's start with maybe the
• 9:54
  oldest method chains. Chains, yeah, they're the
• 9:57
  workhorse, aren't they? Very economical, especially for
• 9:59
  lower speeds and really large loads.
• 10:01
  They boast very high strength. Typically the ultimate tensile
• 10:04
  strength is something like 70% higher than the fatigue limit
• 10:07
  that's often showed as point A on a stress life curve.
• 10:10
  So huge safety margin there for unexpected overloads.
• 10:13
  Plus you generally need smaller sprockets compared to belt
• 10:15
  bullies to handle the same amount of torque.
• 10:17
  Space saving but. There's always a, but isn't
• 10:20
  there. Their big drawback relates
• 10:22
  directly to their shape, their geometry.
• 10:24
  It's called chordal action. That's the one.
• 10:26
  Chordal action basically means speed variation.
• 10:29
  It's caused by the fact that a chain engages with the sprocket
• 10:32
  not as a smooth circle, but it's a series of straight lines, like
• 10:35
  the size of a Polygon. So imagine a bicycle chain going
• 10:39
  around the sprocket. As the sprocket rotates, the
• 10:42
  effective radius where the chain is pulling changes slightly.
• 10:45
  It rises and falls as each link engages and disengages.
• 10:49
  This causes these tiny but cyclical speed ups and slowdowns
• 10:52
  in the output speed. It's inherent to the chains
• 10:55
  bracket geometry, and for precision motion control that
• 10:58
  variation is often completely unacceptable.
• 11:01
  Right, which is where synchronous belts really shine.
• 11:03
  They're specifically designed to get rid of that speed variation,
• 11:06
  or at least minimize it dramatically.
• 11:08
  Let's look at some of the newer technologies like a power grip
• 11:11
  GT3 was mentioned. Yeah, the Power Grip GT3 is a
• 11:13
  good example of the advancements made entirely through tweaking
• 11:17
  the tooth profile geometry. See traditional synchronous
• 11:21
  belts. The older ones often had
• 11:22
  trapezoidal shaped teeth. The problem there is that the
• 11:25
  tooth only really contacts the sprocket pulley at certain
• 11:28
  points, usually near the corners.
• 11:30
  This means the main stress line where the force is concentrated
• 11:33
  ends up right at the base of the tooth in the root.
• 11:36
  That stress concentration is what limits the belt's life.
• 11:38
  It initiates cracks there. OK, so how does an advanced
• 11:41
  profile like the GT3 fix that that structural weak point?
• 11:45
  The Power Group GT3 and similar advanced profiles use a more
• 11:49
  curved, optimized tooth shape, specifically engineered to allow
• 11:54
  full flank contact. The entire side face of the belt
• 11:57
  tooth engages smoothly with the pulley groove.
• 11:59
  By spreading that load across the whole flank, you eliminate
• 12:02
  that critical stress concentration line down at the
• 12:04
  tooth root. This drastically increases the
• 12:07
  belt's life is power capacity. And because you get this
• 12:10
  continuous full contact, it also significantly reduces that
• 12:13
  chordal effect we talked about with chains.
• 12:15
  Much smoother power delivery, ideal for precision indexing or
• 12:18
  robotics. Now when you're actually sizing
• 12:21
  one of these belt drives, especially if it's being driven
• 12:23
  by something dynamic like ADC servo motor or a stepper motor
• 12:27
  which can have really High Peak torques, how do you handle
• 12:30
  those? Do you size for the peak?
• 12:32
  That's a good question for those intermittent high loads, like
• 12:35
  the maximum torque the motor can put out, or it's stall torque,
• 12:37
  or maybe the holding torque For a stepper, it's generally
• 12:40
  considered reasonable to size the belt drive for that peak
• 12:42
  value using a design service factor of just 1.0.
• 12:46
  The rationale is that these peaks are usually brief.
• 12:49
  OK, so size for the peak at 1.0. But then crucially, you must
• 12:54
  check that the normal everyday running loads fall within a
• 12:57
  higher design service factor, typically somewhere between 1.5
• 13:01
  and 2.0. This ensures the built isn't
• 13:04
  being constantly hammered near its limit during routine
• 13:06
  operation. Gives you that safety monitoring
• 13:09
  for a long life. Makes sense?
• 13:10
  Oh, and just a quick practical design tip from the sources,
• 13:14
  they generally don't recommend using sprockets or pulleys that
• 13:17
  have a diameter smaller than the width of the belt itself.
• 13:20
  It tends to cause excessive belt tracking forces trying to pull
• 13:23
  the belt sideways off the pulley.
• 13:26
  Good tip. These are the devices that
• 13:32
  connect two shafts together end to end to transmit torque and
• 13:36
  there seems to be quite a spectrum here.
• 13:37
  The sources mentioned things like torsionly rigid couplings
• 13:40
  like a Reynold flex maybe, then torsionly flexible ones like pin
• 13:44
  flex or gear flex, and then something quite different fluid
• 13:47
  couplings with hydrostart given as an example.
• 13:49
  Yeah, that's a good overview. Fluid couplings like the
• 13:52
  Hydrostart system. They're really brilliant for
• 13:54
  managing startup transients, especially with big motors
• 13:57
  driving high inertia loads. Think about starting a large
• 14:00
  conveyor belt or a big industrial fan.
• 14:02
  When that big motor kicks on direct online, it draws this
• 14:05
  massive spike of current from the mains and it puts a huge
• 14:09
  mechanical shock through the whole drivetrain.
• 14:11
  With a fluid coupling like Hydra Start, the motor actually starts
• 14:15
  up under very low load. The coupling uses hydraulic
• 14:18
  fluid and an impeller turbine mechanism to provide a very
• 14:21
  soft, delayed engagement of the driven.
• 14:23
  Machine. So instead of that sudden jolt,
• 14:26
  that bang, it's more like a smooth ramp up of speed for the
• 14:29
  load. Exactly.
• 14:30
  The water spins up quickly close to its operating speed, almost
• 14:33
  unloaded, drawing only a brief manageable current peak just to
• 14:36
  switch on. Then the fluid coupling
• 14:38
  gradually picks up the load, accelerating it smoothly from 0
• 14:41
  RPM right up to speed. This drastically reduces wear
• 14:45
  and tear on gearboxes, belts, chains, everything downstream.
• 14:48
  And it's much kinder to the electrical supply infrastructure
• 14:51
  too. No massive voltage dips.
• 14:53
  Very clever. OK, let's start wrapping things
• 14:56
  up by looking at the language of the engineer, the specific tools
• 15:00
  and codes they use to actually verify and execute these designs
• 15:03
  we've been talking about. We need tools for checking
• 15:05
  things, geometry verification. The sign bar came up as an
• 15:08
  example. The sign bar, yes, a fundamental
• 15:12
  tool in any decent machine shop or inspection lab.
• 15:16
  It's used for setting up and checking very precise angles or
• 15:19
  tapers. So the example given was
• 15:21
  checking apart with a specific taper, 6 1/8 inches per foot.
• 15:25
  To check that, you'd use a standard, say, 5 inch sign bar.
• 15:29
  You place the part in the sign bar, and then you elevate one
• 15:31
  end of the sign bar using precision gauge blocks until the
• 15:33
  tapered surface of the part is exactly parallel to your
• 15:36
  reference surface, usually a granite surface plate.
• 15:39
  The required height H of the gauge block stack needed to
• 15:42
  achieve that parallelism is calculated for that specific
• 15:45
  taper. In a 5 inch bar, the height H
• 15:47
  works out to be 2.396 inches. Did you get that reading?
• 15:50
  You know your taper is correct. It's a very precise numerical
• 15:53
  check of the geometry. Simple, but effective.
• 15:57
  OK, next, the language of quantification.
• 15:59
  If you're designing a drive system, you absolutely have to
• 16:01
  size it based on the torque it needs to handle.
• 16:03
  Right. And there's a crucial, very
• 16:05
  common formula for sizing rotating Dr. systems based on
• 16:08
  torque, usually measured in pound inches.
• 16:10
  In the US customary system, it's torque Q in pound inches equals
• 16:15
  63,025 times the shaft horsepower divided by the shaft
• 16:20
  RPM. So Q = 6325 times HP over RPM.
• 16:25
  Exactly that constant 63, 2025. It just bundles up all the
• 16:29
  conversion factors need to get from horsepower, which is power,
• 16:32
  a rate of energy and RPM, rotational speed directly into
• 16:36
  torque, a quisting force in those specific units, pound
• 16:39
  inches. It's the immediate link between
• 16:41
  the power you're putting in and the actual mechanical twisting
• 16:43
  capacity of the Shafter drive. Absolutely fundamental
• 16:46
  calculation. Got it.
• 16:47
  And finally making the design A reality.
• 16:50
  We need to talk to the machines themselves.
• 16:51
  The language is CNC code right? G codes and M codes.
• 16:54
  Yep, computer numerical control. G codes generally define the
• 16:58
  geometry of the motion, how the tool moves.
• 17:00
  M codes, control machine functions, coolant on, off,
• 17:03
  skindle, start, stop, things like that.
• 17:05
  So G01G01 is just basic linear interpolation, telling the
• 17:10
  machine to move the tool in a straight line from where it is
• 17:12
  to the next specified point. The workhorse of most machining
• 17:15
  and for machine actions. MO5M05 is the standard command
• 17:18
  to stop the main spindle rotation in a normal controlled
• 17:22
  way, and MO6M06 is pretty much universally the command to
• 17:25
  initiate an automatic tool change sequence.
• 17:27
  OK, G01 move straight MO five stop spindle MO 6 change tool.
• 17:32
  The machine's basic vocabulary and these tools, these machines,
• 17:35
  they also dictate how we have to process the materials
• 17:37
  themselves. What about grinding?
• 17:39
  How fast should we be grinding these often very hard tool
• 17:42
  materials? Speed is critical in grinding.
• 17:45
  When you're grinding things like high speed steel or maybe some
• 17:48
  non ferrous tool materials, the surface speed of the grinding
• 17:51
  wheel, how fast the abrasive particles are moving past the
• 17:55
  workpiece has to be kept in a fairly tight range.
• 17:58
  The recommendation is typically between 4500 and 6500 surface
• 18:01
  feet per minute. If you go too fast you risk
• 18:04
  generating too much heat. That can cause two major
• 18:06
  problems, tempering, which actually softens the hardened
• 18:10
  steel, ruining the tool, and heat checking, which is where
• 18:13
  tiny little cracks form on the surface due to the intense
• 18:16
  localized thermal stress. Both are bad news.
• 18:18
  So speed control is vital. Absolutely, and this need for
• 18:22
  precision, it's even reflected right there on the grinding
• 18:25
  wheel itself. In the standard wheel marking
• 18:26
  system you might see a code like A-306-L5V2351.
• 18:32
  That whole string tells an experienced machinist a lot.
• 18:35
  For instance, the 36 refers to the abrasive grain size, 36 is
• 18:38
  quite course, and the V tells you the bond type holding the
• 18:41
  abrasive grains together is vitrified, which is a hard
• 18:44
  glassy ceramic bond very common for general purpose grinding.
• 18:47
  Wow, a whole language just in that cloud.
• 18:50
  OK, I think we've covered a massive amount of ground here.
• 18:53
  Everything from making sure a Class 2A thread has that tiny
• 18:56
  bit of essential clearance to prevent galling, to picking a
• 19:00
  built profile that won't fail prematurely at the tooth root,
• 19:04
  all while constantly battling against catastrophic thermal
• 19:07
  expansion and bearings. Absolutely.
• 19:09
  And I think the big take away for you, the listener is that
• 19:12
  every single component selection, every design choice,
• 19:16
  whether it's picking a clever flexure to get precise motion or
• 19:19
  just calculating the required torque capacity using that
• 19:22
  simple horsepower formula we mentioned, it's all dictated by
• 19:25
  this interplay of precise numerical analysis,
• 19:28
  understanding the established trade-offs and adhering to
• 19:30
  critical industry standards. It's all deeply interconnected.
• 19:33
  So the next time you look at some complex piece of machinery,
• 19:36
  maybe you'll see it a bit differently.
• 19:38
  You'll know that it's reliability, it's very function
• 19:41
  is rooted deep down in things like preemptive thermal checks,
• 19:44
  careful calculation of tolerance stack ups, and even the elegant
• 19:48
  load distributing geometry of something as seemingly simple as
• 19:51
  a synchronous pulley groove. And if you enjoyed thinking
• 19:54
  about quantifying those systems and preventing failures, just
• 19:58
  remember that the cutting edge is always pushing further.
• 20:01
  Design optimization now often involves engineering at the
• 20:03
  micro level, like deliberately using highly localized buckling
• 20:07
  bodes in these complex periodic cellular microstructures.
• 20:11
  Basically tiny, intricate internal frameworks.
• 20:15
  To maximize strength while minimizing weight.
• 20:17
  That's a whole other level where failure analysis is almost built
• 20:20
  right into the materials design. Fascinating stuff.
• 20:22
  Until the next equation design junkies, we are mechanical
• 20:25
  engineering made simple.