Latest / Mechanical Engineering Made Simple / (#22) Vector Calculus to Heatsinks: Bridging Math and Design of Thermal Fluids

Transcript

(#22) Vector Calculus to Heatsinks: Bridging Math and Design of Thermal Fluids · Mechanical Engineering Made Simple

• 0:00
  Welcome back, design junkies, to mechanical engineering made
• 0:02
  simple. Great to be back.
• 0:04
  Today we are diving straight into, well, the really core
• 0:08
  stuff, the foundational mathematics that underpins
• 0:11
  thermal and fluid dynamics. That's right, vector calculus
• 0:14
  and conservation laws. Yeah, and you know, one of our
• 0:17
  listeners shared some sources and they highlighted something
• 0:19
  we see a lot, don't we? Textbooks, papers.
• 0:23
  They often jump straight from these core differential
• 0:25
  equations like Boom into really complex applications.
• 0:29
  Oh yeah, things like thermal spreading analysis or optimizing
• 0:32
  flow channels. Exactly.
• 0:34
  And there's this, this gap, this Canyon almost between the pure
• 0:38
  math and the practical problem. It really is.
• 0:40
  You can't fully appreciate the solutions, the clever designs,
• 0:44
  if you don't speak the language the problem is written in, you
• 0:46
  know? Absolutely.
• 0:47
  So our mission today is to try and bridge that gap, connect
• 0:51
  these abstract mathematical tools, your gradients, your
• 0:54
  divergences, to things we can actually measure and design
• 0:57
  around, like heat flow paths or how stable a fluid flow is.
• 1:01
  Precisely. We need to get comfortable with
• 1:03
  the physics of fields before we can really manage the fluxes,
• 1:07
  the energy transfer, the momentum transfer.
• 1:09
  It's about understanding that language first.
• 1:11
  OK, so let's unpack this a bit. When we're analyzing heat
• 1:14
  transfer or fluid flow, we're fundamentally dealing with
• 1:17
  fields. That's the key concept fields.
• 1:20
  Like temperature, that's a scalar field, right?
• 1:23
  Just has a magnitude at every point.
• 1:24
  Just a number. How hot is it here?
• 1:26
  But. Velocity.
• 1:28
  That's different. It's a vector field.
• 1:29
  It has magnitude and direction everywhere.
• 1:31
  How fast and which way the fluids moving at that point?
• 1:34
  So to describe how these fields change across space, we need
• 1:39
  these tools, these vector operators.
• 1:42
  Exactly, and maybe the most intuitive one to start with is
• 1:46
  the gradient, usually written as enable of T for temperature.
• 1:49
  The gradient of temperature. Conceptually, what's it telling
• 1:52
  us? It's giving you 2 key pieces of
• 1:54
  information. One, the direction of the
• 1:56
  steepest change in temperature. Think of it like the steepest
• 2:00
  uphill direction on a temperature map, and two, the
• 2:04
  rate of that change how steep is that hill?
• 2:06
  So if you put a hot ship on a cold heat sink, the gradient
• 2:10
  vector points directly along the fastest path heat wants to take.
• 2:13
  Right, and in simple rectangular coordinates your basic XYZ
• 2:17
  system, the formula looks pretty clean.
• 2:20
  It does. The gradient of T is just.
• 2:22
  Well, it's a vector sum. You take the partial derivative
• 2:24
  of T with respect to X times the X unit vector plus the partial
• 2:28
  of T with respect to Y times the Y unit vector plus the partial
• 2:31
  of T with respect to Z times Z unit vector.
• 2:34
  Equation 1.31 in the source. Straightforward enough, but.
• 2:38
  And this is a big, but the math changes quite a bit depending on
• 2:41
  your coordinate system. Right cylindrical or spherical
• 2:44
  coordinates always tricky. Exactly.
• 2:46
  Imagine analyzing flow in a pipe or heat spreading from a
• 2:50
  cylindrical heater. You switch to cylindrical
• 2:52
  coordinates or Theta Z and suddenly that gradient formula
• 2:56
  gets more complex. Specifically, look at the
• 2:59
  angular component, how temperature changes as you move
• 3:01
  around the cylinder, the Theta direction.
• 3:03
  What happens there? You get this extra factor.
• 3:05
  It's E sub Theta times 1 / r times partial T partial Theta.
• 3:09
  See equation 1.36. One over, so the radius matters
• 3:13
  directly in the gradient calculation.
• 3:15
  Immensely, that 1 / R term tells you that the same angular change
• 3:18
  in temperature contributes less to the overall gradient if
• 3:21
  you're far away from the center, large R because it's spread over
• 3:24
  a wider arc. OK, so close to the center a
• 3:27
  small angular change looks steep.
• 3:29
  Far away, it's shallower. Precisely, geometry dictates the
• 3:33
  math. It's not just a change of
• 3:35
  variables, it changes the structure of the operator.
• 3:37
  OK. So that's the gradient focused
• 3:39
  on slope and direction of change.
• 3:41
  What about the other big operators for fields, especially
• 3:44
  vector fields like velocity? Well, we absolutely need to talk
• 3:47
  about divergent nabla dot V and curl nabla cross.
• 3:51
  Divergent. What's the physical intuition
• 3:53
  there? Think of divergent divergent of
• 3:56
  velocity as measuring expansion or contraction.
• 3:59
  If you draw a small imaginary box in, the fluid is more fluid
• 4:03
  flowing out-of-the-box than throwing it.
• 4:05
  If the divergent is positive, the fluid is expanding like it's
• 4:08
  coming from a source within the box.
• 4:11
  If it's it's negative, it's contracting, heading towards a
• 4:13
  sink. If it's zero, well, we'll get to
• 4:16
  that. It's crucial for mass
• 4:18
  conservation. Got it.
• 4:19
  Source or sink and curl nubble across V?
• 4:22
  Curl measures rotation, local rotation, or vorticity.
• 4:27
  Imagine placing a tiny tiny paddle wheel in the flow.
• 4:30
  OK, would it spin? The curl vector tells you the
• 4:33
  axis it would spin around and how fast.
• 4:35
  If the curl is 0 everywhere, the flow is called irritational.
• 4:38
  Which I assume makes the math easy.
• 4:40
  It's significantly easier in many cases, yeah.
• 4:43
  And sometimes these operators get combined, right?
• 4:45
  Yeah, I see the Laplacian mentioned Nabila Squared T.
• 4:48
  Yes, the Laplacian is basically the divergent of the gradient.
• 4:51
  So for temperature, Nabil squared T involves second
• 4:53
  derivatives like partial squared T, partial X ^2 plus partial
• 4:57
  square T, partial y ^2 and so on.
• 4:59
  Equation 1.35 shows the rectangular form.
• 5:02
  And that's important because. Because setting the Laplacian of
• 5:06
  temperature equal to 0 nebula $2.00 and four describes steady
• 5:11
  state heat conduction with no internal heat generation, it's
• 5:15
  often the starting point for solving basic heat transfer
• 5:17
  problems in solids. All right, so we have this
• 5:19
  mathematical language gradient divergent curl Laplacian.
• 5:24
  Now let's apply to the fundamental rules governing how
• 5:27
  things, especially fluids, move conservation laws.
• 5:30
  Starting with the big one, conservation of mass, this gives
• 5:34
  us the continuity equation, it's most general form.
• 5:36
  For a compressible fluid, the equation looks like this partial
• 5:39
  rho partial T plus partial with respect to X of rho V sub X plus
• 5:44
  partial with respect to Y of rho V sub Y plus partial with
• 5:47
  respect to Z of rho V sub Z = 0. That's equation 1.58.
• 5:52
  Whoa, OK, but break that down. Partial rho, Partial T.
• 5:54
  That's the rate of change of density rho over time.
• 5:57
  At a fixed point, it's how much mass is accumulating or
• 6:00
  depleting in a tiny volume got. It and the other terms.
• 6:03
  Those are the divergent terms. Essentially, they represent the
• 6:06
  net rate at which mass is flowing out of that tiny volume
• 6:09
  in each direction XYZ. So accumulation rate plus net
• 6:12
  outflow rate equals 0. Mass isn't created or destroyed
• 6:16
  in that volume. Exactly.
• 6:17
  It's just a mathematical statement of mass conservation.
• 6:20
  But you mentioned compressible flow.
• 6:22
  What happens if the fluid is incompressible like water or
• 6:25
  oil, where the density swirl is basically constant?
• 6:29
  That's where the magic happens. If rho is constant, its
• 6:32
  derivatives are 0, so that first term partial rho partial T
• 6:36
  vanishes, and you can pull the constant rho out of the other
• 6:40
  derivatives. Since rho isn't 0, you can
• 6:42
  divide the whole equation by it. What you're left with is
• 6:45
  beautiful, which is partial V sub X, partial X plus partial V
• 6:49
  sub Y, partial Y plus partial V sub Z partial Z = 0, the
• 6:53
  equation 1.61. Wait, that's just.
• 6:56
  The divergent of the velocity vector nabla VO not V equals.
• 7:00
  Wow, so for incompressible flow the velocity field has zero
• 7:03
  divergent everywhere. Exactly.
• 7:05
  It mathematically states that the fluid isn't expanding or
• 7:08
  contracting locally. What flows in must flow out.
• 7:12
  It's incredibly elegant and simplifies fluid dynamics
• 7:15
  immensely. OK.
• 7:16
  That's a huge insight. Now when we're tracking fluids,
• 7:18
  there's this other concept mentioned, the substantial
• 7:20
  derivative or material derivative.
• 7:21
  Why do we need that? Right, the substantial
• 7:23
  derivative, often written as capital D by capital DT.
• 7:26
  We need it because usually we don't care just about how things
• 7:30
  change at a fixed point in space.
• 7:32
  That's the partial derivative with respect to time partial
• 7:34
  partial T. We care about what happens to a
• 7:36
  specific chunk of fluid as it moves along.
• 7:38
  Precisely, we want to know the rate of change experienced by a
• 7:42
  fluid particle as it follows its path.
• 7:44
  Think about tracking a floating particle and measuring its
• 7:47
  temperature change. OK, so how does the substantial
• 7:49
  derivative capture that? It combines 2 effects.
• 7:51
  Look at equation 1.55 for density.
• 7:54
  DRO by DT equals partial rho partial T.
• 7:58
  That's the local change at a point, right?
• 8:00
  Plus V sub X times partial rho partial X + v sub Y times
• 8:04
  partial rho partial y + v sub Z times partial rho partial Z.
• 8:08
  So those extra terms involve the velocity V sub XV, sub YV sub Z
• 8:13
  and the spatial gradients partial row, partial X,
• 8:15
  etcetera. Exactly.
• 8:17
  That second group of terms is the convective derivative.
• 8:20
  It accounts for the change the particle experiences simply
• 8:23
  because it's moving to a new location where the density or
• 8:26
  temperature or whatever property is different.
• 8:28
  Ah, I see. So DDT local change plus
• 8:32
  convective change. It follows the particle.
• 8:34
  You got it. It's essential for applying
• 8:36
  Newton's laws to fluids correctly.
• 8:39
  Which brings us to conservation of momentum, the next big
• 8:44
  conservation law. This is basically Newton's
• 8:46
  second law. Faber applied to a fluid element
• 8:49
  or a control volume. The source is stated as a
• 8:52
  theorem. The net force F on a control
• 8:55
  volume equals the time rate of increase of momentum inside plus
• 8:58
  the net rate at which momentum flows out across the surface.
• 9:01
  Equation 1.66. That's the integral form, yeah.
• 9:04
  It accounts for all the forces. Pressure pushing on the
• 9:06
  surfaces. Viscous shear stress is acting
• 9:08
  like friction and body forces like gravity and when you.
• 9:11
  Take that principle, apply it using the substantial
• 9:13
  derivative, and specialize it for, say, incompressible laminar
• 9:17
  flow with constant viscosity. You end up with a set of complex
• 9:21
  nonlinear partial differential equations, the famous or maybe
• 9:25
  infamous Naver Stokes equations like equations 1.69 and 1.70
• 9:31
  shown for cylindrical coordinates.
• 9:32
  They look incredibly complicated, especially in
• 9:35
  cylindrical or spherical coordinates.
• 9:36
  They are, but the core idea, the engineering insight, is that
• 9:40
  they are fundamentally just $5 for a fluid.
• 9:43
  They balance inertia, pressure gradients, viscous forces, and
• 9:46
  external forces to determine the fluids acceleration and
• 9:49
  therefore its velocity field. They describe how pretty much
• 9:52
  every liquid in most gases move. OK, wow.
• 9:55
  So we've built up the language and the fundamental laws.
• 9:58
  Let's shift gears now and connect this back to more
• 10:00
  tangible, practical things a designer actually grapples with,
• 10:04
  starting with heat generation inside a solid.
• 10:06
  Good idea. Let's take a simple case.
• 10:08
  A slim conductor, maybe a wire or flat plate, with heat being
• 10:12
  generated uniformly throughout its volume.
• 10:14
  Maybe it's resistive heating $2.00.
• 10:16
  OK, uniform volumetric heat generation and let's say both
• 10:19
  ends or surfaces are kept cool at a constant temperature
• 10:22
  subzero. Right, So heat is generated
• 10:25
  everywhere inside and it has to flow out to the cooler
• 10:28
  boundaries. Where will it be hottest?
• 10:30
  Intuitively right middle, farthest from the cool edges.
• 10:33
  Exactly. And there's a simple formula for
• 10:36
  that peak temperature T Max. It's T Max to 0 plus the QG
• 10:39
  Frankie L2. That's equation 1.4 here.
• 10:43
  L is the thickness or total length and K is the thermal
• 10:46
  conductivity. T Max equals T 0 + Q sub G * l
• 10:51
  ^2 over AK. OK, now here's a really useful
• 10:55
  practical insight buried in that equation, highlighted as
• 10:58
  equation 1.5 in the source material.
• 11:01
  Compare this case distributed generation to a case where the
• 11:04
  same total amount of heat is generated, but all at a single
• 11:08
  point or plane right in the center.
• 11:10
  The thermal resistance for this distributed case where heat is
• 11:13
  generated everywhere is only 1/4 of the resistance you get if all
• 11:16
  the heat originated from the center 1/4.
• 11:18
  Just by spreading the heat generation out.
• 11:20
  Yeah, it makes a huge difference in keeping that peak temperature
• 11:23
  down, spreading the load. Essentially, it fundamentally
• 11:26
  changes how you might design something to avoid hotspots.
• 11:28
  That's a fantastic nugget, but often the biggest thermal
• 11:32
  challenge isn't the resistance within a material, right?
• 11:35
  It's at the joins, the interfaces.
• 11:37
  Oh. Absolutely.
• 11:39
  What engineers call thermal contact resistance one or out
• 11:42
  Fowler. This is often the dominant
• 11:44
  resistance in an assembly. Why does it happen?
• 11:47
  I mean if you bolt 2 flat metal plates together, shouldn't heat
• 11:51
  flow easily? You'd think so, but flat at our
• 11:53
  scale isn't flat at the microscopic scale.
• 11:56
  Every surface, even polished ones, has tiny peaks and
• 11:59
  valleys. Asperities.
• 12:00
  Like microscopic mountains. Exactly.
• 12:03
  So when you press 2 surfaces together, they only touch at the
• 12:06
  tips of the highest peaks. The actual solid to solid
• 12:09
  contact area might only be like 1 or 2% of the apparent area you
• 12:13
  see. Whoa, really?
• 12:14
  That. Little Yep.
• 12:15
  So heat has two main paths, direct conduction through those
• 12:19
  few tiny solid spots, which is usually efficient but limited in
• 12:22
  area, and conduction across the gas, usually air trapped in the
• 12:26
  gaps between the peaks and gases are generally poor conductors.
• 12:29
  So the gap is the main barrier. It's a huge barrier and modeling
• 12:33
  the overall heat transfer across that interface.
• 12:35
  The contact conductance P is really complex.
• 12:38
  Equation 1.9 gives a sense of it.
• 12:40
  Let's see, it depends on the material conductivity of CIGAL,
• 12:43
  but also surface properties like micro hardness centers, the
• 12:46
  clamping pressure. Right, more pressure squash
• 12:48
  increase in contact area, but it also depends on the gas and the
• 12:51
  gap, its conductivity at calalars and even its mean free
• 12:54
  path Lambda the mean. Free path.
• 12:57
  That's a molecular gas property. It is it.
• 12:59
  Shows how microscopic physics affects the macroscopic heat
• 13:01
  transfer, especially if the gap is very small or the pressure
• 13:04
  very low. It's a fascinating mix of
• 13:06
  mechanics, material science, and heat transfer.
• 13:08
  OK, so. Interfaces are critical.
• 13:11
  Now shifting back to fluids in motion, convection.
• 13:13
  How do we quantify heat transfer to a moving fluid, say inside a
• 13:17
  pipe or channel? The workhorse.
• 13:19
  Equation there is Newton's law of cooling QQ equals HSTSTS on
• 13:23
  LA equation 1. Dollar 12 Q equals.
• 13:26
  HS Delta T, where Q is the heat rate, S is the surface area,
• 13:30
  delta T is the temperature difference between the surface
• 13:32
  and the bulk fluid and and lower dollars is the convective heat
• 13:36
  transfer coefficient. And finding a dollars, well
• 13:40
  that's often the main goal in convective heat transfer
• 13:42
  analysis. It wraps up all the complexity
• 13:45
  of the fluid flow and thermal conditions into one number but
• 13:49
  dollars. Isn't just a constant, is it?
• 13:51
  Especially inside a duct? No.
• 13:53
  Definitely not, particularly in the entrance region.
• 13:56
  When fluid first enters a pipe, say from a large reservoir, its
• 14:00
  velocity profiles usually uniform.
• 14:02
  As it flows down the pipe, viscosity acts, the layers near
• 14:06
  the wall slow down, and a boundary layer grows until the
• 14:08
  velocity profile becomes stable or fully developed.
• 14:11
  OK, the velocity. Profile develops represented by
• 14:14
  $6. And the same thing happens with
• 14:16
  temperature. If the pipe wall is heated, a
• 14:19
  thermal boundary layer grows from the wall into the fluid.
• 14:22
  The temperature profile also develops over a thermal entrance
• 14:25
  length $6.00. So both.
• 14:26
  Velocity and temperature profiles are changing near the
• 14:29
  entrance, correct? And because the dollar depends
• 14:32
  critically on the shape of those profiles, especially the
• 14:34
  gradients right at the wall, a dollar changes significantly
• 14:37
  along the entrance length. It's usually very high right at
• 14:39
  the start and decreases as the flow develops until until.
• 14:42
  Both profiles become fully developed, meaning their shapes
• 14:45
  stop changing far downstream. The dollars often settles down
• 14:49
  to a constant asymptotic value for a given flow regime and
• 14:52
  boundary condition like the. Example given for a laminar flow
• 14:56
  in a round tube with uniform wall heat flux exactly.
• 14:59
  That's a classic case. Once the flow is fully
• 15:02
  developed, the dimensionless form of Lawler is called the
• 15:04
  Nusselt number. No, a lie becomes constant for
• 15:07
  that specific case. Nu equals 4.364.
• 15:11
  Nu equals 4. .364 A fixed number.
• 15:14
  A fixed number. Doesn't matter if it's water or
• 15:17
  oil or air, as long as it's laminar.
• 15:19
  Doesn't matter if the pipe is tiny or huge.
• 15:22
  That dimensionless group captures the physics.
• 15:24
  It's incredibly powerful. For designers, that is powerful.
• 15:26
  OK, one last practical application area mentioned
• 15:29
  geometric optimization. How do we find the best shape or
• 15:33
  size for heat transfer components right specifically?
• 15:36
  Looking at forced convection, like blowing air through a heat
• 15:38
  sink with parallel fins or plates, often you have a fixed
• 15:41
  pressure drop available from your fan Delta P1 and maybe a
• 15:45
  fixed total volume for the heat sink.
• 15:46
  OK, constrained. Design.
• 15:48
  The question is, what's the optimal spacing D effect between
• 15:51
  those plates or fins to maximize heat removal?
• 15:54
  How do you? Figure that out the insight.
• 15:56
  Is that the best performance happens when you size the
• 15:58
  channels just right, so the thermal boundary layers growing
• 16:01
  from opposite walls just meet or merge right as the fluid is
• 16:05
  about to exit the channel length L?
• 16:08
  So you use the full channel volume for heat transfer without
• 16:11
  choking the flow too much? Exactly.
• 16:13
  Make the channels too wide and the core flow doesn't get heated
• 16:16
  effectively. Make them too narrow and the
• 16:18
  friction skyrockets for the same flow rate, or your flow rate
• 16:21
  drops for the fixed pressure drop.
• 16:23
  So there's an. Optimum and the sources link
• 16:26
  this to something called the Bijan number.
• 16:28
  Yes, the. Bijan number Bijan dollar is a
• 16:30
  dimensionless group representing the imposed pressure difference.
• 16:34
  It's defined as Bijan elbow Delta PL-2 MU Alpha, where MU is
• 16:38
  viscosity and alpha is thermal diffusivity delta.
• 16:41
  PL. squared over MU alpha OK. And there are correlations
• 16:44
  linking this Bijan number to the optimal geometry for parallel
• 16:48
  plates. Equation 5.47 gives V John Froff
• 16:50
  at approx 2.7 Tex 14D dot. Over L is about 2.7 * B to the
• 16:56
  power of -1 fourth. So the available pressure drop
• 16:59
  directly tells you the best aspect ratio for your channels.
• 17:02
  That's the power. Of it, it connects an
• 17:04
  operational constraint delta P2 directly to a physical design
• 17:08
  dimension down dimensional analysis and optimization
• 17:12
  theory. Wow.
• 17:14
  OK, what a journey we've taken. Then we started way back with
• 17:17
  the abstract math of vector fields, gradients, divergent
• 17:21
  curl, the basic. Language use that.
• 17:24
  Language to write down the fundamental laws of conservation
• 17:26
  of mass and momentum, leading to things like the continuity and
• 17:29
  Navier Stokes equations, the core physics.
• 17:31
  And differential form and then we.
• 17:33
  Saw how these underlying principles manifest in really
• 17:36
  practical engineering problems. Calculating peak temperatures
• 17:40
  with internal generation. The nightmare of thermal contact
• 17:42
  resistance, finding heat transfer coefficients and even
• 17:45
  optimizing the physical shape of channels using the Bijan number.
• 17:49
  It really shows. The connection, doesn't it?
• 17:51
  From the most abstract math to the most concrete design
• 17:54
  decisions, it absolutely. Does and I highlight something
• 17:57
  crucial you touched on when those full differential
• 18:00
  equations like Navier Stokes are just too beastly to solve
• 18:04
  analytically, which they often. Are engineers.
• 18:06
  Have another powerful tool, Dimensional Analysis.
• 18:09
  Exactly. The sources mentioned the
• 18:12
  Buckingham Π theorem. It's a systematic way to take
• 18:15
  all the variables in a complex problem and group them into a
• 18:18
  smaller set of dimensionless numbers like the Nusselt.
• 18:21
  Number we saw or Reynolds number, Pranth, Nell number
• 18:23
  precisely. These dimensionless groups
• 18:25
  capture the essential physics and ratios governing the system.
• 18:29
  By focusing on these groups, we can design smarter experiments,
• 18:32
  develop empirical correlations like that one new 4.3644 or the
• 18:36
  Bajan number correlation, and get reliable engineering answers
• 18:40
  even when a full analytical solution is out of reach.
• 18:43
  It's a cornerstone of practical fluid mechanics and heat
• 18:45
  transfer, a pathway to. Solutions when the pure math
• 18:48
  hits a wall. A fantastic take away indeed.
• 18:51
  Until the next. Equation designed junkies.
• 18:53
  We are mechanical engineering made simple.