Latest / Mechanical Engineering Made Simple / (#129) From Bias to Blueprint: The Mechanical Engineer's Deep Dive into Strategy, DFM&A, and Power Optimization

Transcript

(#129) From Bias to Blueprint: The Mechanical Engineer's Deep Dive into Strategy, DFM&A, and Power Optimization · Mechanical Engineering Made Simple

• 0:00
  Get ready to torque your nuts hard in your springs and forge
• 0:03
  your mind into a bulletproof study guide.
• 0:06
  We are mechanical engineering made simple.
• 0:09
  Welcome to the deep dive. Today we're really digging into
• 0:14
  something big, the essence of mechanical engineering.
• 0:18
  We're going to try and map out that journey, you know, from
• 0:20
  just an idea, thinking about the whole system.
• 0:22
  All the way to a real physical product.
• 0:24
  Something tangible, efficient, and powered up.
• 0:28
  Exactly. It's a massive field, obviously.
• 0:30
  Oh, huge. But we've tried to structure
• 0:33
  this based on the source materials, breaking it down into
• 0:35
  maybe 3 core areas for you. First up is the strategy, part
• 0:39
  systems and strategy, how you even define the problem.
• 0:42
  Then we get into the making of the thing, design and
• 0:45
  manufacturing. And finally how we powered all.
• 0:47
  Think fluids, power and efficiency.
• 0:49
  All right, let's jump into that first pillar then.
• 0:51
  Yeah, Strategy, systems engineering.
• 0:53
  The sources really hammer home this idea of issue formulation.
• 0:57
  Yeah, and and it's fundamental how you frame or or formulate
• 1:00
  the problem at the very beginning.
• 1:01
  Well, that pretty much decides how you're going to solve it.
• 1:03
  It sets the whole direction. So getting that definition right
• 1:06
  is key. Absolutely key.
• 1:08
  It involves figuring out the client's objectives, what they
• 1:11
  really want, what they value. That's value system design, and
• 1:15
  then brainstorming all the possible ways to get there.
• 1:17
  The alternatives? Well.
• 1:18
  Yeah, there's a catch, right? A human.
• 1:20
  Catch. There always is, isn't there?
• 1:21
  Yeah. Yeah.
• 1:23
  Relying purely on human judgement, even from experts,
• 1:26
  well, it opens the door to cognitive biases.
• 1:29
  We all have them. Right.
• 1:30
  We like to think engineers are purely logical, but these biases
• 1:34
  creep in like selective perception.
• 1:38
  That's a big one. It's basically looking for
• 1:40
  evidence that confirms what you already believe or want to be
• 1:43
  true. You filter out anything that
• 1:45
  contradicts your preferred conclusion.
• 1:47
  It's like wanting a self fulfilling prophecy.
• 1:49
  And expectations is similar. Kinda with expectations you give
• 1:53
  more weight, more validity to information that just confirms
• 1:56
  what you already thought. It feels right because it
• 1:58
  matches your existing beliefs. It's easy to see how it happens
• 2:01
  when you're drowning in data. You grab onto what looks
• 2:04
  familiar, and then there's ease of recall.
• 2:06
  Right. That's where you overestimate
• 2:08
  the likelihood of something, just because examples spring to
• 2:11
  mind easily. Think dramatic failures versus
• 2:15
  quiet successes. The dramatic stuff sticks, so
• 2:17
  you might think it's more common than it really is.
• 2:20
  So if we know about these biases, why are they still such
• 2:23
  a problem? Because, well, knowing isn't
• 2:26
  always enough, is it? Biases aren't just about not
• 2:28
  knowing something. They're like mental shortcuts
• 2:30
  our brains take. So acknowledging this human
• 2:34
  factor, this potential for error right at the start, it means we
• 2:37
  need really solid objective processes later on.
• 2:40
  Which brings us neatly to how we enforce that rigor in the
• 2:44
  physical design stage. DFM and a design for
• 2:47
  manufacturing and assembly. Exactly.
• 2:49
  That's the framework, and it's useful to kind of separate the
• 2:52
  two parts in your mind. DFA, designed for assembly is
• 2:55
  all about the labor part. How can we make putting this
• 2:57
  thing together faster, simpler, with fewer steps?
• 3:00
  Makes sense and D FM. D FM, designed for
• 3:02
  manufacturing, focuses more on the components themselves,
• 3:05
  reducing material waste, making parts easier to fabricate,
• 3:08
  thinking about tooling costs, things like that.
• 3:10
  And using them together, the results can be pretty dramatic.
• 3:15
  Oh, absolutely staggering. Sometimes the sources give this
• 3:19
  fantastic example a Motor Dr. redesign.
• 3:21
  They applied DFM and A principles.
• 3:24
  Then what happened? They slashed the number of parts
• 3:26
  went from 19 components down to just seven.
• 3:29
  Think about that. Wow, 19 down to seven.
• 3:31
  That's huge simplification. And that simplification ripples
• 3:35
  through everything. Assembly time.
• 3:37
  It dropped from over 213 seconds down to about 86 seconds less.
• 3:41
  Than half the time. And the bottom line, the unit
• 3:43
  manufacturing cost, the UMC, it went from $47.22 down to $30.21.
• 3:49
  That's a massive cost reduction per unit, almost $17.00.
• 3:53
  That's not just savings, that's a competitive weapon.
• 3:55
  Precisely. That's the power of thinking
• 3:57
  about manufacturing and assembly right from the design stage.
• 4:00
  It's strategic. So when you have potentially
• 4:03
  thousands of parts, how do you keep track?
• 4:05
  How do you manage that complexity, especially for
• 4:08
  retrieving old designs or planning manufacturing?
• 4:10
  Good question. Engineers use classification
• 4:13
  systems, basically codes for parts.
• 4:15
  There are two main flavors. You'll see.
• 4:16
  Monocode. Yeah, it's hierarchical things
• 4:18
  like a library system. You know, broad category than
• 4:21
  subcategory getting more specific like animals than
• 4:24
  mammals than primates than humans.
• 4:25
  Got it. And the other type.
• 4:27
  Is polycode or sometimes called feature based?
• 4:30
  This is more like using filters. When you shop online you tag a
• 4:33
  part based on specific features. Does it have holes?
• 4:37
  Is a round? What material, regardless of its
• 4:41
  parent category? OK, more flexible in some ways.
• 4:43
  Can be the sources mentioned a really detailed example, an 8
• 4:47
  digit code that captures shape, key features, size, precision
• 4:50
  level, even the material with 8 digits.
• 4:52
  Structured like that, you can represent something like over
• 4:55
  25,000,000 unique part families. 25,000,000 That's incredible
• 4:59
  granularity. It's like a universal language
• 5:01
  for parts within a company. That's the goal.
• 5:04
  It helps you find if you've designed something similar
• 5:06
  before, or helps automate the planning for how to make it.
• 5:09
  Saves reinventing the wheel. OK, so strategy set parts are
• 5:12
  organized, but before we physically make anything, we
• 5:16
  need to predict how it's going to behave.
• 5:18
  That means math, right? Mathematical models.
• 5:21
  Absolutely essential because the things we design, they're
• 5:24
  dynamic, they move, they change over time.
• 5:27
  A robot arm curse suspension airflow over a wing.
• 5:30
  They're not static. Exactly.
• 5:32
  So mathematical models, usually differential equations or
• 5:35
  difference equations, are how we write down the rules.
• 5:37
  Yeah, they say based on where the system is now, here's where
• 5:41
  it's going to be next. It's prediction quantified.
• 5:45
  And there's a kind of unified theory for this.
• 5:47
  Control engineers don't need totally different math for
• 5:50
  electrical versus mechanical systems.
• 5:52
  It's one of the beautiful things, actually.
• 5:54
  We use these idealized system elements.
• 5:56
  Think of them as building blocks.
• 5:58
  Some store energies like a spring or a capacitor.
• 6:01
  Some dissipate energy, like a damper or a resistor.
• 6:04
  Some transform it. And these same blocks apply
• 6:06
  across different physical domains.
• 6:08
  Broadly, yes. You can model mechanical,
• 6:11
  electrical, fluid, even thermal systems using these core energy
• 6:14
  concepts. It gives us a common language.
• 6:16
  And the modern way to express this, especially in control
• 6:19
  theory, is the state variable model.
• 6:22
  Yeah, that's become really central.
• 6:23
  It's a very compact way to package up the system's
• 6:26
  dynamics. You use vectors, a state vector,
• 6:29
  usually 6 * 2, that captures the system's internal condition.
• 6:32
  Like position and velocity. Could be, yeah.
• 6:35
  Then an input vector representing external forces or
• 6:38
  signals acting on it, and an output vector for the things you
• 6:41
  can measure or care about. It's powerful for handling
• 6:44
  complex multi input, multi output systems.
• 6:47
  And this model lets you analyze the system's response, how it
• 6:51
  reacts to inputs. Precisely, we look at things
• 6:54
  like transient performance. When you kick the system, does
• 6:57
  it overshoot the target? How quickly does it settle down
• 7:00
  to a steady state? These are critical for
• 7:02
  performance and safety. How do you measure that?
• 7:04
  We often analyze it in the frequency domain using poles and
• 7:07
  zeros. Their location on the complex
• 7:09
  plane tells you a lot. And for simpler systems like
• 7:12
  standard 2nd order systems, think mass spring damper.
• 7:15
  You can calculate things like overshoot and settling time
• 7:18
  exactly. Based on what?
• 7:20
  Based on two key parameters, the damping ratio, which tells you
• 7:23
  how much oscillation there is, and the natural frequency, which
• 7:26
  tells you how fast it wants to oscillate.
• 7:29
  Knowing the poles and zeros which come from these instantly
• 7:32
  tells an engineer if the system is stable, if it's sluggish, if
• 7:35
  it's too bouncy. It's the mathematical core of
• 7:37
  stability analysis. OK, so math predicts behavior.
• 7:41
  Now let's get back to the physical world, making sure the
• 7:43
  actual parts meet the specs. You mentioned materials earlier.
• 7:47
  They're complicated. Incredibly, even defining a
• 7:51
  property isn't always straightforward.
• 7:53
  Take hardness. Seems simple, right?
• 7:55
  How resistant is it to scratching or indentation?
• 7:59
  Well, the source point out there are at least 15 different common
• 8:02
  tests for penetration hardness and guess what?
• 8:04
  They give different numbers. They all give different
• 8:06
  numerical results, so just saying a material has a certain
• 8:09
  hardness isn't enough to specify which test was used.
• 8:13
  It highlights how critical standards and clear
• 8:15
  communication are. 15 Wow. OK, so assuming we can specify
• 8:19
  the material, what about efficient manufacturing powder
• 8:22
  metallurgy? PM sounds interesting.
• 8:24
  It is. It's a neat process.
• 8:26
  You take very fine metal powders, press them into a dye
• 8:29
  shape and then heat them. Center them so the particles
• 8:32
  bond together. And the advantage?
• 8:33
  Several big ones first, usually very little material waste.
• 8:38
  Unlike machining where you cut away material, often the part
• 8:41
  comes out needing no extra machining it's net shape or
• 8:44
  close to it. Saves steps and material.
• 8:46
  Exactly, plus you can get unique properties.
• 8:49
  You can control the porosity, leave tiny gaps which is great
• 8:52
  for filters or for making self lubricating bearings.
• 8:55
  The pores hold oil. Clever.
• 8:58
  What about processes involving deformation like hot working?
• 9:02
  Right. So hot working means shaping the
• 9:04
  metal plastically, permanently deforming it, but above its re
• 9:07
  crystallization temperature. Now hot is relative.
• 9:11
  How so? Well, for lead that temperature
• 9:13
  is basically room temperature, but for some steels you're
• 9:15
  talking 2000°F or even higher. OK.
• 9:18
  Let's talk about extrusion. Pushing metal through a shaped
• 9:21
  hole, A dye. That's the idea.
• 9:24
  Works great for aluminum, copper, but steel?
• 9:27
  Steel's tough. Its yield strength is very high,
• 9:30
  even when hot under the immense pressure of extrusion.
• 9:34
  It really wants to friction Weld itself to the steel dye.
• 9:36
  Which would be. Bad.
• 9:38
  Catastrophic. Yeah.
• 9:39
  So for extruding steel you need serious lubrication technology,
• 9:43
  often phosphate coatings on the steel billet, or even using
• 9:46
  molten glass as a lubricant layer between the hot steel and
• 9:49
  the dye. Molten glass is lubricant.
• 9:51
  That's wild and piercing, making seamless tubes.
• 9:55
  How does that work without a seam?
• 9:57
  It's counterintuitive. You heat up a solid cylindrical
• 9:59
  billet, then force it over a pointed mandrel wall.
• 10:01
  It's also rotating. The forces cause the metal to
• 10:05
  shear in a specific way, and it actually creates a crack, an
• 10:08
  opening right in the center of the billet ahead of the mandrel.
• 10:10
  The mandrel then just enlarges and smooths out that internally
• 10:13
  generated hole into a tube. It relies on controlled failure
• 10:16
  in a way. Fascinating controlled internal
• 10:19
  crack. OK, parts are made.
• 10:20
  How do we ensure consistent quality?
• 10:23
  Statistical Quality control, SQC.
• 10:25
  Right, you can't inspect every single dimension on every single
• 10:29
  part. Usually it's too slow and
• 10:30
  expensive. So you use statistics Control
• 10:33
  charts like X bar and R charts are common tools.
• 10:36
  What do they? Track.
• 10:37
  The X bar chart tracks the average of samples over time,
• 10:40
  showing if the process center is drifting.
• 10:43
  The R chart tracks the range within samples, showing if the
• 10:46
  variability is changing. And there are rules for
• 10:48
  interpreting these charts. Yes, one key rule of thumb
• 10:52
  mentioned is if you get say, 5 to 7 consecutive points all
• 10:57
  falling on the same side of the center line, even if they're
• 10:59
  still within the upper and lower control limits.
• 11:01
  Something's likely shifted. Exactly, it suggests a non
• 11:05
  random change has occurred in the process and you should
• 11:07
  investigate even if you aren't making bad parts yet.
• 11:10
  It's an early warning system. What about the processes
• 11:12
  inherent ability capability limits?
• 11:15
  That's crucial. Capability limits, often
• 11:18
  calculated as the process average ±3 standard deviations,
• 11:22
  tell you the natural spread of the process, where 99.73% of the
• 11:26
  output is expected to fall if the process is stable.
• 11:29
  Now here's the critical comparison.
• 11:31
  You compare those capability limits to the tolerance
• 11:34
  specified by the design engineer, the required
• 11:37
  precision. And if the tolerance is tighter,
• 11:39
  narrower than what the process is capable of.
• 11:42
  Then you have a problem. That process physically cannot
• 11:45
  reliably produce parts within the required specification.
• 11:49
  You're guaranteed to make scrap or parts needing rework.
• 11:52
  Your only option then is 100% inspection, sorting good from
• 11:56
  bad, which is super inefficient. SQC helps ensure your process is
• 12:00
  actually capable of meeting the requirements in the 1st.
• 12:02
  Place makes sense. OK, let's shift gears from
• 12:05
  making things to powering them. Fluids and energy first
• 12:09
  principles. What is a fluid?
• 12:12
  Fundamentally, it's a substance that continuously deforms or
• 12:15
  flows when you apply a shear stress.
• 12:17
  Doesn't matter how small the stress is.
• 12:18
  Unlike a solid. Right.
• 12:20
  A solid will deform a bit under shear than resist.
• 12:23
  A fluid just keeps moving. Think of water, air, oil versus
• 12:27
  steel or wood. We also sometimes talk about
• 12:29
  ideal fluids, non viscous imaginary versus real fluids
• 12:33
  which have viscosity internal friction.
• 12:35
  Got it. And the most famous equation
• 12:36
  here is probably Bernoulli's. Has to be.
• 12:39
  It's basically a statement of energy conservation applied
• 12:42
  along the streamline in a fluid flow.
• 12:44
  It says the sum of the kinetic energy due to velocity, the
• 12:48
  displacement energy related to pressure, and the potential
• 12:51
  energy due to higher elevation stays constant.
• 12:54
  Foundational stuff. When we need to move fluids, we
• 12:57
  use pumps or fans, turbo machinery.
• 12:59
  How are they selected? I heard something about fan
• 13:02
  pressure. Yeah, the units it's a bit
• 13:04
  quirky. Even though a fan moves gas,
• 13:07
  usually air, the pressure increase recreates.
• 13:09
  The fan head is very often measured and specified in inches
• 13:13
  of water column or in H2O. Inches of water for air.
• 13:17
  Yeah, it's just an old convention that stuck.
• 13:19
  It relates to the height difference in a simple YouTube
• 13:21
  manometer filled with water for selecting pumps though a really
• 13:25
  important parameter is the non dimensional specific speed to N.
• 13:28
  Specific speed? What does it tell you?
• 13:30
  It's a calculated value based on the desired flow rate, the
• 13:33
  required head pressure increase, and the pumps rotational speed.
• 13:37
  What it does is characterize the type of pump impeller geometry,
• 13:40
  Rachel mixed flow axial that will be most efficient for that
• 13:44
  specific combination of flow and head.
• 13:46
  It helps you pick the right kind of pump for the job, ensuring it
• 13:50
  operates near its best efficiency point.
• 13:52
  So it matches the pump design to the system's needs.
• 13:55
  Let's talk power generation gas turbines.
• 13:57
  They seem incredibly complex, especially with emissions
• 14:00
  controls. They are.
• 14:01
  Modern gas turbines, especially large ones for power generation,
• 14:05
  have extremely sophisticated combustion systems to meet
• 14:08
  strict environmental limits, particularly for Nych ox oxides
• 14:11
  of nitrogen. How do they do that?
• 14:13
  One key technology is the dry Lonox combustor, or DLN.
• 14:16
  The dry means they don't inject water or steam, which used to be
• 14:19
  a common way to lower combustion temperatures.
• 14:21
  And that's not Sox. DLN systems use clever fuel
• 14:25
  nozzle designs and premixing chambers.
• 14:27
  Premixing. Yeah, They mix the fuel and air
• 14:29
  very thoroughly before they burn.
• 14:31
  The goal is to maintain a very consistent, lean excess air,
• 14:34
  fuel air ratio everywhere in the combustion zone.
• 14:38
  This avoids localized hotspots where most Nos forms, allowing
• 14:41
  low emissions over a wide range of operating loads.
• 14:44
  It's tricky control and. When we talk about efficiency
• 14:47
  overall, there's a concept beyond simple percentages,
• 14:50
  exergy, and EGM. Right.
• 14:53
  Exergy or availability is arguably the true measure of
• 14:56
  energy quality or work potential.
• 14:58
  Standard efficiency often just compares energy in versus energy
• 15:01
  out. But exergy accounts for the
• 15:03
  irreversibilities in any real process.
• 15:05
  Friction, heat transfer across a temperature, different things
• 15:08
  that destroy the potential. They do useful work.
• 15:10
  So minimizing lost work potential.
• 15:12
  Exactly, and that leads to EGM entropy generation minimization.
• 15:16
  Entropy generation is directly linked to exergy destruction, so
• 15:20
  EGM is a powerful modern optimization technique.
• 15:24
  It tells engineers how to choose design parameters like the
• 15:27
  diameter of a pipe, the length of heat exchanger tubes,
• 15:30
  specifically to minimize the destruction of power while the
• 15:34
  device does its job. It's a more fundamental approach
• 15:37
  to optimization than just maximizing simple efficiency.
• 15:40
  That sounds powerful, but applying it must run into real
• 15:44
  world messiness like in heat exchangers.
• 15:46
  Oh definitely. Heat exchangers look simple on a
• 15:48
  diagram. Fluid A flows pass, fluid B heat
• 15:52
  transfers. But in reality getting the fluid
• 15:54
  to flow uniformly across all the heat transfer surfaces is a huge
• 15:57
  challenge. The big issue is flow Mal
• 16:00
  distribution. Some of the fluid finds
• 16:01
  shortcuts bypassing the main heat transfer area.
• 16:04
  Some might leak between passes. The sources mentioned that in
• 16:07
  some situations, especially with viscous fluids and laminar flow,
• 16:10
  these ineffective streams, bypass and leakage can actually
• 16:13
  account for as much as 60% of the total flow. 60% that's
• 16:17
  crippling the performance. It is so real.
• 16:21
  Heat exchanger design involves lots of clever mechanical
• 16:24
  details. Things like adding bypass
• 16:26
  ceiling strips inside the shell just to force the fluid to go
• 16:30
  where it's supposed to. It's not just about the
• 16:32
  thermodynamics. The fluid mechanics inside are
• 16:34
  critical and complex. Thinking bigger picture, system
• 16:36
  choices have global impacts. Refrigeration, for example.
• 16:40
  We hear about ODP and HDWP, right?
• 16:43
  Two key environmental metrics for refrigerants ODP is ozone
• 16:47
  depletion potential. It measures how much a substance
• 16:50
  damages the stratospheric ozone layer relative to a baseline
• 16:53
  refrigerant R11 OK. Ozone layer impact and HDWP.
• 16:58
  That's Halo carbon global warming potential.
• 17:00
  It measures how much refrigerant contributes to global warming,
• 17:03
  again relative to R 11 / a certain time horizon.
• 17:06
  So we moved away from ozone depleting CFCS to HFCS.
• 17:10
  Correct HFCS. Hydrofluorocarbons generally
• 17:12
  have zero ODP because they don't contain chlorine.
• 17:15
  That was a success for the ozone layer, but many HFCS turned out
• 17:18
  to have very high HGWP values. They're potent greenhouse gases.
• 17:22
  So solving one problem created another.
• 17:24
  To some extent, yes. Now the push is towards
• 17:26
  refrigerants with both 0 ODT and very low or 0 HTWP.
• 17:32
  Things like hydrocarbons, propane, isobutane.
• 17:35
  The challenge there is that many of these are flammable, which
• 17:37
  introduces safety considerations that the old inert refrigerants
• 17:41
  didn't have. It's complex trade-offs.
• 17:43
  These efficiency and environmental concerns tied
• 17:45
  directly into managing energy use in buildings and industry
• 17:49
  right energy audits. Absolutely.
• 17:51
  An energy audit is basically a systematic investigation of how
• 17:55
  energy is being used and where it's being wasted.
• 17:58
  What does an auditor look for? It's usually 2 main things.
• 18:01
  First, eliminating unnecessary energy use altogether.
• 18:05
  Are lights or HVAC systems running when nobody's there?
• 18:08
  Can we use time clocks or occupancy sensors?
• 18:11
  Are there drafts? Core insulation?
• 18:13
  What they call leak Plugging? Simple stuff often yields big
• 18:15
  savings. And the.
• 18:16
  Second thing, minimizing the energy required for the tasks
• 18:19
  that are necessary. Can we use more efficient
• 18:20
  motors? Can we optimize process heating?
• 18:23
  Can we recover waste heat? Like with industrial furnaces
• 18:26
  you mentioned, they're more sophisticated now.
• 18:27
  Yeah, exactly. Modern continuous furnaces, say
• 18:30
  for heat treating metals, aren't just big hot boxes anymore.
• 18:34
  They often have carefully designed temperature zones and
• 18:37
  maybe a Preheat zone using waste heat from hotter zones in the
• 18:39
  main firing zone, then a soak zone to ensure uniform
• 18:42
  temperature. Why the zones this zoned
• 18:45
  approach? This temperature profile allows
• 18:47
  them to process more material faster, improve capacity, but do
• 18:51
  it without just burning way more fuel.
• 18:53
  It increases throughput and maintains fuel efficiency.
• 18:56
  Smart design. So looking back, we've traced
• 18:58
  this path from grappling with human biases in the initial
• 19:03
  strategy. Yeah, that messy human element
• 19:05
  to the. Rigor of mathematical modeling
• 19:07
  for predicting system stability. Quantifying reality.
• 19:10
  Then driving down costs and complexity with DFM and a the.
• 19:13
  Practicalities of making things efficiently.
• 19:16
  And finally, chasing ultimate thermodynamic efficiency with
• 19:19
  concepts like exergy and EGM and power and fluid systems.
• 19:23
  It really covers a huge range doesn't it?
• 19:25
  From psychology to to advance thermodynamics.
• 19:27
  It does. So what's the big take away for
• 19:30
  someone trying to grasp mechanical engineering?
• 19:32
  What's the core tension here? I think if you boil it down,
• 19:35
  it's this constant balancing act.
• 19:37
  On one hand, you have the inherent variability and let's
• 19:40
  face it, the flaws in human judgement, those cognitive
• 19:44
  biases we started with. You have to manage that, right?
• 19:47
  But on the other hand, you're striving for absolute
• 19:49
  mathematical precision and physical optimization.
• 19:52
  EGM state variable control, tight tolerances.
• 19:56
  You need designs that are rigorously engineered, cost
• 19:59
  effective, safe, durable and now increasingly environmentally
• 20:03
  sound. Managing the subjective while
• 20:05
  pursuing the objective. That's a good way to put it,
• 20:08
  navigating that tension using structured processes and deep
• 20:11
  technical knowledge to bridge that gap.
• 20:13
  That, I think, is the real heart of the engineering mission we've
• 20:15
  been talking about. Keep your nuts torqued and
• 20:18
  spring hard design junkies.