Latest / Mechanical Engineering Made Simple / (#133) Systems Thinking From Bias to Physics

Transcript

(#133) Systems Thinking From Bias to Physics · Mechanical Engineering Made Simple

• 0:00
  Welcome back to Zion Junkies, to mechanical engineering made
• 0:02
  simple. Today we are attempting what
• 0:06
  feels like a small miracle. We're trying to give you a high
• 0:09
  speed shortcut through, well, pretty much the entire field of
• 0:13
  modern mechanical and systems engineering.
• 0:15
  We've synthesized a massive collection of material covering
• 0:18
  everything from, you know, the high level methodologies down to
• 0:21
  the really hands on thermodynamic models.
• 0:24
  It really is the ultimate convergence deep dive.
• 0:27
  Our mission here is to show you how those big strategic
• 0:30
  decisions, what we call systems thinking, is totally tied to the
• 0:34
  most detailed, you know, granular technical processes
• 0:38
  happening on the shop floor. Things like calculating fluid
• 0:41
  resistance or optimizing a heat exchanger.
• 0:43
  Exactly. We often think of them as two
• 0:45
  separate worlds, don't we? The big picture strategy over
• 0:48
  here and the detailed technical stuff over there.
• 0:51
  But they're not. They're completely fused.
• 0:53
  You just can't have brilliant execution without a sound
• 0:56
  strategy. And that strategy has to be
• 0:58
  grounded in the hard physics of what's actually possible to
• 1:01
  build. Right.
• 1:02
  So we've structured this deep dive into three major areas to
• 1:05
  really highlight that connection.
• 1:07
  First, we're going to tackle the messy part, the human element,
• 1:11
  how cognitive bias messes with complex systems analysis.
• 1:15
  Second, we pivot to the really practical tools for efficiency
• 1:19
  specifically designed for manufacturing and assembly and
• 1:23
  classification codes. And finally, we ground ourselves
• 1:26
  in the fundamental physics you need to model dynamic systems.
• 1:30
  We're talking fluids, heat, and energy optimization.
• 1:33
  All right, let's. Unpack that starting at the top
• 1:35
  right with the big picture Ystems engineering methodology.
• 1:40
  When a designer gets a challenge, the first step isn't
• 1:42
  sketching, it's something called issue formulation.
• 1:45
  That is king. Absolutely everything starts
• 1:47
  there. The way you define a problem,
• 1:49
  the client objectives, the context, all the possible
• 1:52
  alternative that critically influences the final outcome.
• 1:55
  It's an iterative interactive process.
• 1:57
  Because look if you define the wrong problem from the start.
• 2:01
  You're guaranteed to get the wrong solution, doesn't matter
• 2:03
  how brilliant your final machine is.
• 2:05
  Exactly. And when you're tackling these
• 2:07
  big messy problems, the analytical methods usually fall
• 2:10
  into two big buckets. The first is what we call expert
• 2:13
  opinion methods. OK, so that's when you don't
• 2:15
  have hard data. Precisely.
• 2:17
  It's essential when you don't have an established theory or
• 2:19
  reliable data. Think of, you know, large scale
• 2:22
  surveys or a Delphi method where you build consensus among
• 2:25
  professionals. They're useful, but, and this is
• 2:29
  a huge but, they introduce the unavoidable risk of human
• 2:32
  judgement. Which means cognitive bias and
• 2:36
  something called value incoherence, yes.
• 2:39
  That's a great way to put it. People, even with the best
• 2:41
  intentions, can be really inconsistent or even contradict
• 2:44
  themselves, especially if they're already invested in a
• 2:47
  certain outcome. I can see that.
• 2:49
  So what are the big biases we should be watching out for?
• 2:51
  Well, one that comes up constantly is selective
• 2:54
  perception. This often leads to this desire
• 2:57
  for like a self fulfilling prophecy where a decision maker
• 3:00
  only looks for information that supports the conclusion they
• 3:03
  already want to reach. Right, they just filter out all
• 3:06
  the contradictory evidence. It's the classic confirmation
• 3:09
  bias loop. If you believe your new material
• 3:11
  is the best, you only run the tests that you know it will
• 3:14
  pass. Absolutely.
• 3:16
  Then there's another one called ease of recall.
• 3:19
  If a piece of data is easy to remember, maybe it was a
• 3:21
  dramatic failure or a recent success.
• 3:24
  We subconsciously give it much more weight than data that's,
• 3:27
  you know, buried in some report from five years ago.
• 3:29
  So that completely distorts your sense of risk and probability.
• 3:32
  Completely. And the last one is
• 3:33
  expectations. This is where we just
• 3:36
  automatically attach more validity to information that
• 3:39
  confirms what we already believe.
• 3:41
  It makes it dangerously easy to ignore the one piece of data
• 3:45
  that could have saved you from a huge, costly mistake.
• 3:48
  So the whole point of systems engineering isn't just managing
• 3:51
  data, it's about managing our own flawed thinking.
• 3:54
  That's it. The goal is to build a system
• 3:56
  for information acquisition and processing that improves our
• 3:59
  judgment and our choices. We should probably say, I mean,
• 4:02
  mental shortcuts exist for a reason, right?
• 4:05
  We can't perfectly analyze every single tiny decision.
• 4:08
  Oh, of course not. Cognitive heuristics like
• 4:10
  satisficing are necessary. Satisficing.
• 4:13
  That's looking for a good enough solution, not the perfect one.
• 4:16
  Exactly. If you're just picking a
• 4:18
  standard bolt from a catalog, satisficing is perfectly
• 4:21
  rational. You don't have time to optimize
• 4:23
  that. But if satisficing is OK for a
• 4:25
  fastener, it's absolutely fatal when you're designing a complex
• 4:28
  industrial system. And that brings us to Part 2,
• 4:31
  engineering efficiency, where we build discipline systems to
• 4:34
  force that optimal outcome. Let's talk DFM and a design for
• 4:38
  manufacturing and assembly. DFM and A is where the rubber
• 4:41
  really meets the road on cost and quality.
• 4:44
  It breaks down into two main parts.
• 4:46
  You have DFA design for assembly, which is all about the
• 4:49
  labor side, reducing the time and complexity of actually
• 4:52
  putting the thing together. And then DFM design for
• 4:55
  manufacturing which focuses on the material and tooling costs,
• 4:59
  making sure the part is cheap to make before you cut any steel.
• 5:02
  A key insight from all the sources is that the single
• 5:04
  biggest win from a good DFM and a process is just a radical
• 5:08
  reduction in part count. That's the cascade effect,
• 5:10
  isn't? It it is.
• 5:11
  You reduce the part count, and that automatically simplifies
• 5:14
  assembly. It lowers your overhead, it cuts
• 5:16
  assembly time, and it increases quality because there are just
• 5:20
  fewer opportunities for things to go wrong.
• 5:21
  We have a brilliant, really tangible example of this.
• 5:25
  It's a Motor Dr. redesign. The original assembly was pretty
• 5:29
  complex. I had 19 separate parts. 19 OK.
• 5:33
  After they went through the DFM and a process, they slashed that
• 5:36
  down to just seven parts. Wow, from 19 to 7.
• 5:39
  That's a huge reduction in complexity.
• 5:42
  And look at the gains that came with it.
• 5:43
  This is where it hits the bottom line.
• 5:45
  Assembly time fell from 213.4 seconds down to just 86.34
• 5:52
  seconds. Incredible.
• 5:53
  And even more important, the total unit manufacturing costs.
• 5:57
  The UMC went from 47.22 to 30.21.
• 6:02
  That's a reduction of almost 17 per unit.
• 6:05
  I mean, at scale that's millions of dollars in savings and that.
• 6:07
  Only happens when you get people out of their silos, which brings
• 6:10
  up the power of multidisciplinary teams.
• 6:12
  Right, you need engineering, manufacturing, purchasing,
• 6:16
  marketing, even field service all in the same room.
• 6:19
  Because that's how you get a complete picture of the problem.
• 6:22
  What is field service actually seeing fail in the real world?
• 6:25
  What are customers telling market about the price?
• 6:28
  When everyone helps build the solution, Everyone is committed
• 6:32
  to it. O To manage all this complexity,
• 6:35
  even in these leaner designs, engineers need organizing tools.
• 6:39
  This is where classification and coding systems come in.
• 6:42
  These are basically Productivity Tools.
• 6:43
  They reduce unnecessary variety and help with standardization.
• 6:47
  You see 2 main types. The 1st is monocode.
• 6:50
  Think of it like a library's Dewey Decimal system.
• 6:52
  It's hierarchical. You go from general to specific.
• 6:55
  Yeah, there's Polycode. Polycode is more like using tags
• 6:57
  on a website. It's feature based.
• 6:59
  A specific digit means a specific feature.
• 7:01
  So an engineer could instantly find all parts made of aluminum
• 7:04
  that also have a 5mm hole. So you can just combine the tags
• 7:07
  you need. Exactly, and the power of these
• 7:10
  systems is just staggering. An 8 digit alphanumeric code for
• 7:14
  example, covering shape, size, material that can represent 2.5
• 7:19
  * 10 to the 7th unique combinations.
• 7:22
  Two and a half million combinations from one code.
• 7:25
  Oh 25,000,000. 25,000,000 Wow. That's how you manage that level
• 7:29
  of variety. And this code then drives things
• 7:31
  downstream like process planning ACAPP system.
• 7:35
  A computer aided process planning system can use that
• 7:39
  code to instantly pull up the right manufacturing steps.
• 7:42
  Which means the planners have to deeply understand the
• 7:44
  capabilities of all the different fabrication processes,
• 7:46
  machining, casting, forming to choose the best one.
• 7:50
  For that part, the code literally directs the production
• 7:52
  path. OK, so we've gone from strategic
• 7:54
  thinking to design efficiency. Now let's make the final jump
• 7:58
  into the Core physics Part 3, Decoding Dynamic systems.
• 8:02
  This is all about the mathematical models that
• 8:04
  underpin everything we design. Right, because a modern engineer
• 8:07
  deals with systems that are part mechanical, part electrical,
• 8:11
  part fluid all at once. So you need a unified way to
• 8:14
  model them. These mathematical models are
• 8:16
  usually differential equations based on physical laws.
• 8:20
  And the key to unifying them is this idea of idealized system
• 8:23
  elements. That's the Rosetta Stone, Yes.
• 8:25
  Every system can be broken down into elements that either store
• 8:28
  energy like a spring, dissipate energy like a damper, or
• 8:32
  transform energy like a motor. And that concept works for
• 8:35
  mechanical, electrical, fluid, thermal.
• 8:38
  All of them. It applies universally.
• 8:40
  It gives you a consistent way to model a really complex reality.
• 8:43
  So let's get specific. With fluid mechanics, we have to
• 8:46
  distinguish between real fluids and ideal fluids.
• 8:49
  Yes, real fluids are viscous. They deform under stress.
• 8:52
  You know, the actual water in a pipe.
• 8:54
  Ideal fluids are non viscous. They don't really exist, but
• 8:57
  they're incredibly useful for simplifying the first round of
• 9:00
  calculations. And Bernoulli's equation is
• 9:02
  foundational for that. It is.
• 9:04
  It's this beautiful, elegant statement that the sum of
• 9:08
  velocity, pressure and potential energy is constant along a
• 9:12
  streamline. It's all about conservation of
• 9:14
  energy. But things get weird and really
• 9:16
  counterintuitive when you add compressibility in high speeds.
• 9:19
  This is where the physics gets fun.
• 9:21
  So in normal slow incompressible flow, if you make a pipe wider,
• 9:26
  the fluid slows down. We all know that right?
• 9:28
  However, in supersonic flow, so Mach number greater than one, if
• 9:33
  you have an expanding area, it actually leads to an increasing
• 9:36
  velocity. Wait, hold on, that's completely
• 9:37
  backward. Why does the physics flip when
• 9:39
  you break the sound barrier? Because the flow is moving
• 9:42
  faster than the pressure waves can travel upstream to signal
• 9:46
  that there's a change happening. The flow can't feel the
• 9:49
  expansion coming, so instead of slowing down, the expanding area
• 9:52
  just provides the space for the flow to accelerate.
• 9:54
  It's critical for designing things like rocket nozzles.
• 9:57
  That is fascinating. OK, let's shift to energy
• 10:00
  systems, thermodynamics, and the first law.
• 10:03
  The first law is the conservation of energy law.
• 10:05
  You can't create or destroy energy, you can only change its
• 10:08
  form. But for optimization, just
• 10:11
  conserving the amount of energy isn't enough.
• 10:13
  We have to look at the quality of the energy, its usefulness,
• 10:16
  and that's where exergy comes in.
• 10:18
  So exergy is the maximum useful work you can get out of a
• 10:22
  system. Correct.
• 10:23
  And that idea pairs with a method called entropy Generation
• 10:26
  minimization, or EGM. The whole goal of EGM is to
• 10:30
  minimize the destruction of power, that is, the irreversible
• 10:33
  loss of energy's usefulness while a device is doing its job.
• 10:38
  It's critical for optimizing cooling systems, heat
• 10:40
  exchangers, cryogenics. Speaking of heat exchangers,
• 10:43
  let's finish with that. Why is designing something like
• 10:46
  a big shell and tube heat exchanger so complex that it
• 10:49
  always needs computer modeling? Because the calculations are
• 10:52
  highly iterative, you're solving these complex equations that
• 10:55
  depend on their own answers. So you have to run through these
• 10:57
  convergence loops over and over. You just can't do it by hand.
• 11:00
  And the quality of the answer depends on the quality of your
• 11:03
  input data. Critically, especially your
• 11:05
  temperature profiles in the real world with complex fluids, they
• 11:09
  are not simple straight lines like in a textbook.
• 11:11
  And on top of all that math, there are hard physical limits
• 11:15
  you can't ignore. For example, the restriction on
• 11:18
  PV squared AH. Yes, the product of velocity
• 11:21
  squared times density. That limit represents this
• 11:25
  constant battle between fluid dynamics and just mechanical
• 11:29
  destruction. You want the fluid to move
• 11:31
  faster for better heat transfer. But the mechanical physics of
• 11:34
  erosion and two vibration fights back that PV squared limit.
• 11:38
  It's usually kept below 4000 lbs mass per foot second squared.
• 11:42
  That's the practical line in the sand.
• 11:44
  Go past it and the fluid starts physically destroying the
• 11:47
  equipment. It often limits your liquid
• 11:49
  velocity to only 7 to 10 feet per second.
• 11:51
  It's the perfect example of how the theoretical optimum gets
• 11:53
  slammed by mechanical reality. So if you pull all this
• 11:56
  together, we started with mitigating our own cognitive
• 11:59
  biases in that first issue formulation phase.
• 12:02
  Then we move to streamlining production with DFM and a
• 12:05
  slashing a part count from 19 down to seven and using coding
• 12:08
  systems that can define 25,000,000 unique combinations.
• 12:12
  And we ended by applying fundamental physics to optimize
• 12:15
  flow and heat transfer, all while respecting hard
• 12:19
  constraints like that PV squared limit.
• 12:21
  So being a great engineer means understanding the methodology of
• 12:24
  design just as much as you understand the math of the
• 12:27
  physics. The process is what guarantees
• 12:29
  the quality. And this brings us right back to
• 12:31
  Part 1. And it raises a really important
• 12:33
  question for you, the designer. Since we know cognitive biases
• 12:36
  like selective perception are just part of being human, and
• 12:40
  since systems like DFM and A rely on humans to set the
• 12:43
  targets and interpret the codes, how can we design processes that
• 12:46
  actively prevent that human tendency to just prove what we
• 12:49
  already believe? How do you design a system
• 12:52
  that's so robust it forces a selection of a truly optimal
• 12:55
  solution, even if that solution is counterintuitive or
• 12:58
  inconvenient for the team? A system that prevents us from
• 13:01
  just settling for the comfortable, good enough
• 13:03
  solution to satisficing that question really defines the next
• 13:07
  frontier of resilient systems design.