Latest / Mechanical Engineering Made Simple / (#111) DFM&A to UMC: The Core Playbook for Engineering Profit, Part Reduction, and Factory Flow Mathematics

Transcript

(#111) DFM&A to UMC: The Core Playbook for Engineering Profit, Part Reduction, and Factory Flow Mathematics · Mechanical Engineering Made Simple

• 0:00
  Welcome back, design junkies, to Mechanical Engineering Made
• 0:02
  Simple. Today we are undertaking a deep
• 0:06
  dive into what really makes or breaks a product.
• 0:08
  We're going past the, you know, the glossy surface design.
• 0:12
  We're drilling right down into the manufacturing ecosystem.
• 0:15
  That's really where the profit gets locked in.
• 0:16
  That's absolutely right. Yeah.
• 0:18
  What we want to do today is layout the core playbook, really
• 0:21
  the fundamental tools engineers use day in, day out to wrangle,
• 0:25
  well, huge complexity and hit some really tough cost targets.
• 0:28
  We're focusing on three pillars designed for manufacturing and
• 0:32
  assembly, DFM and A, then the classification systems for parts
• 0:35
  and materials, and finally the planning math that keeps the
• 0:38
  factory running smoothly. So basically, we're uncovering
• 0:40
  the tricks of the trade, the methods that turn a neat idea
• 0:43
  into something you can actually build affordably at scale.
• 0:46
  And for anyone wondering how finance really steers
• 0:49
  engineering, we'll show you exactly how those cost targets
• 0:51
  are set. We'll even reveal how sometimes
• 0:54
  just, you know, merging 2 tiny parts can unleash massive
• 0:57
  savings all the way down the line.
• 0:59
  It's pretty amazing. Really is those small changes
• 1:02
  have huge ripple effects. OK, let's start at the
• 1:04
  beginning. The path to profit really kicks
• 1:07
  off right at the start of design with something called Design for
• 1:10
  Manufacturing and Assembly. DFM and A get this wrong and
• 1:15
  well, the whole product line can suffer basically forever.
• 1:18
  Yeah, Defeminae is all about designing cost out from the very
• 1:22
  beginning by simplifying the product.
• 1:25
  It's an upfront strategy. You're trying to slash the total
• 1:28
  product cost, and the main way you do that is by cutting down
• 1:30
  the number of individual parts. Fewer parts just means simpler
• 1:33
  assembly, less overhead to manage, and honestly fewer
• 1:37
  chances for things to go wrong. Fewer defects.
• 1:39
  Right. And this is often where the
• 1:42
  organizational stuff gets messy, isn't it?
• 1:44
  We saw that source, that internal memo, classic example,
• 1:46
  engineering basically saying they will not have the resources
• 1:49
  to apply the design for manufacturing assembly process
• 1:52
  and they felt manufacturing should just sort of concur with
• 1:54
  whatever they came up with. Ouch.
• 1:56
  Oh, that memo. Yeah, that's the textbook
• 1:59
  definition of failed concurrent engineering.
• 2:02
  It screams silo mentality. Right.
• 2:03
  Design gets tossed over the wall and manufacturing is just
• 2:06
  expected to magically figure out how to build it affordably.
• 2:10
  It doesn't work. DFM and a needs a multi
• 2:12
  functional team involved from day one.
• 2:15
  Engineering, manufacturing, purchasing, quality, Everyone
• 2:18
  has to be pushing for standardization and
• 2:20
  simplification together. If you wait until that design is
• 2:22
  frozen, forget it. Your best cost saving
• 2:25
  opportunities are already gone. Poof.
• 2:27
  OK. So to make sure that cross
• 2:28
  functional thing actually happens, there's a key rule rule
• 2:31
  #1 the product team must get a specific cost target at the
• 2:35
  start. That's the unit manufacturing
• 2:36
  cost or UMC. That's the number they
• 2:39
  absolutely have to hit. Exactly that UMC target is non
• 2:42
  negotiable and it's calculated with a pretty fundamental
• 2:44
  equation. UMC equals BL plus MC plus TA
• 2:47
  all right. Let's break that down.
• 2:48
  UMC equals BL plus MC plus TA. Let's start with BL.
• 2:52
  What's burdened labor? OK, BL burdened labor.
• 2:56
  It's not just the wage you pay the person on the line, it's the
• 2:59
  direct labor cost plus benefits plus all the relevant overhead.
• 3:03
  So for instance, if the base wage is say 18.75 per hour, you
• 3:08
  might add 138% for overhead benefits.
• 3:12
  That brings the actual cost, the burden cost up to 44.06 per hour
• 3:18
  and that overhead covers everything you know, factory
• 3:21
  heating, insurance, supervision. Wow, OK, so assembly time gets
• 3:25
  really expensive really fast, which loops right back to why
• 3:29
  cutting parts is so vital for DFM and A.
• 3:31
  But let's move to MC material cost.
• 3:33
  That sounds simple, but I bet it isn't just the cost of the raw
• 3:36
  steel or plastic. Not even close.
• 3:38
  MC material cost is actually pretty complex.
• 3:41
  You got the base cost of the part itself, sure, but then you
• 3:43
  add packaging and then there's crucial adders, things like DIP,
• 3:47
  that's duty insurance and freight.
• 3:49
  That alone can be 4 percent, 12% extra.
• 3:51
  And on top of that you have material acquisition cost.
• 3:54
  Maddock Q which covers, you know, the purchasing departments
• 3:56
  work handling inventory and so on.
• 3:58
  That can add another 6 to 16%. Whoa.
• 4:01
  OK. So if I'm in purchasing, I might
• 4:03
  actually spend more energy negotiating that mat AQ
• 4:07
  percentage down or optimizing the freight routes then haggling
• 4:10
  over the raw material price itself.
• 4:12
  You absolutely might. It's a huge lever and it matters
• 4:15
  a lot because MC is almost always the biggest chunk of the
• 4:18
  total UMC. You can only squeeze labor so
• 4:20
  much. Right?
• 4:21
  But materials sourcing, logistics, that's where the
• 4:25
  biggest savings of the biggest cost overruns usually happen.
• 4:28
  Makes sense. And the last piece was TA
• 4:31
  tooling amortization. Right.
• 4:33
  TA is simply the cost of the specialized tools you need.
• 4:36
  Think molds for plastic parts, dies for stamping metal, maybe
• 4:39
  custom fixtures for assembly. You take the total cost of that
• 4:42
  tooling and divide it by the total number of units you expect
• 4:44
  to make over the products entire life.
• 4:47
  So if you have a let's say 1,000,000 cost mold and you plan
• 4:50
  to make 100,000 parts with it, each part carries 10 units of
• 4:53
  cost just for that tool and you have to account for that in the
• 4:56
  UMC up front. OK, the big take away here seems
• 4:58
  crystal clear. Cutting parts is king.
• 5:02
  It reduces the material cost MC, it cuts assembly time BL, and it
• 5:06
  reduces the chances of defects. Every extra part is another
• 5:10
  opportunity to mess something up.
• 5:11
  So part reduction really is the golden rule in DFM and A.
• 5:14
  Precisely. It hits cost, time and quality
• 5:17
  all at once. All right, so we've simplified
• 5:19
  the product, but engineers still face this, well, universe of
• 5:23
  components, thousands of different screws, brackets,
• 5:26
  materials, tools. How do they possibly keep track
• 5:29
  of it all? How do they bring order to that
• 5:31
  chaos? This leads us to classification
• 5:34
  and coding systems, right? Exactly.
• 5:36
  These systems are absolutely essential for managing
• 5:38
  complexity. Classification does a few key
• 5:40
  things. It enables generative process
• 5:43
  planning, basically telling the system automatically how a part
• 5:46
  should be manufactured. It helps with design retrieval,
• 5:48
  stopping engineers from constantly reinventing the
• 5:50
  wheel, you know, finding similar parts that already exist.
• 5:53
  And it helps standardized tool design.
• 5:55
  The real value is driving efficiency by enforcing
• 5:58
  standardization across the board.
• 6:00
  And the way this works is often through a part family code.
• 6:03
  The source mentioned a standard 8 digit alphanumeric code.
• 6:07
  It captures things like basic shape, any special features,
• 6:11
  size, precision, level and the material used.
• 6:14
  Yeah, think of it like a unique fingerprint or DNA profile for
• 6:18
  every single component. And that system is incredibly
• 6:21
  powerful. When you calculate the sheer
• 6:23
  number of unique combinations possible with that kind of code,
• 6:26
  considering all the variations in shape, features, size,
• 6:29
  precision, material, you end up with something like 2.5 * 10 to
• 6:33
  the power of 71 possible unique part identifiers.
• 6:37
  It's staggering. You need structure to handle
• 6:39
  that. That number is mind boggling and
• 6:41
  the structure itself uses these hierarchical trees.
• 6:44
  We saw two main types mentioned, E trees and N trees.
• 6:47
  Right. First you have the E tree which
• 6:49
  stands for exclusive tree. This is for characteristics that
• 6:52
  are mutually exclusive, like the most basic shape definition the
• 6:55
  system makes you choose. Is this part fundamentally
• 6:58
  rotational like a shaft or is it non rotational like a bracket?
• 7:01
  You have to pick one path. It defines the core geometry
• 7:04
  type. OK, one or the other.
• 7:06
  And then there's the entry non. Exactly.
• 7:09
  The entry handles attributes or features that can exist at the
• 7:12
  same time. Think common things like does it
• 7:15
  have holes, threads, maybe slots or notches?
• 7:18
  Or has it had a specific heat treatment since a single part,
• 7:22
  say a rotational shaft, could easily have both a hole and a
• 7:26
  thread? The entry lets you select
• 7:28
  multiple paths, multiple features simultaneously.
• 7:31
  Got it. And this classification idea
• 7:33
  extends beyond just the part geometry, doesn't it?
• 7:36
  There are also detailed systems for materials, a material
• 7:39
  taxonomy covering, I think it said, 79 different material
• 7:42
  families, metals, non metals, compounds.
• 7:45
  Yeah, and what's really interesting now is how those
• 7:47
  material taxonomies are evolving.
• 7:48
  It's not just about strength or melting point anymore.
• 7:51
  Modern systems heavily factor in sustainability criteria, things
• 7:55
  like recyclability and the energy needed to produce the
• 7:58
  material in the first place. For example, the sources pointed
• 8:01
  out that making steel from recycled material uses about 75%
• 8:05
  less energy than making it from raw iron ore.
• 8:07
  That's huge. So that material classification
• 8:10
  isn't just an inventory list, it feeds directly back into the UMC
• 8:14
  calculation we talked about earlier, right, It impacts the
• 8:16
  energy cost, the resource efficiency, even the end of life
• 8:20
  considerations for the product. Absolutely connects everything
• 8:23
  together. OK, let's shift gears.
• 8:25
  We've simplified the design. We've organized all the parts
• 8:27
  and materials. Now how do we plan the actual
• 8:30
  production? How do you predict how many
• 8:33
  you'll need to make and schedule that?
• 8:34
  Well, sometimes chaotic factory floor.
• 8:37
  Right now we move into the mathematics of planning.
• 8:40
  Factories need a decent glimpse into the future to order
• 8:42
  materials and schedule work effectively.
• 8:45
  One really common and critical tool here is quantitative
• 8:48
  forecasting, specifically a time series method called exponential
• 8:52
  smoothing. Exponential smoothing.
• 8:55
  OK. Can you translate that math
• 8:56
  concept into simpler terms for us?
• 8:59
  Sure, conceptually think of exponential smoothing as a kind
• 9:02
  of weighted moving average, but the key difference is that it
• 9:06
  heavily weights the most recent demand data when predicting the
• 9:09
  next period's demand. It's great for situations where
• 9:12
  demand isn't perfectly stable, where it fluctuates.
• 9:15
  The formula looks like this. The new forecast, let's call it
• 9:18
  F sub T equals alpha times the actual demand from the last
• 9:22
  period D sub t -, 1 plus ( 1 minus alpha.
• 9:26
  Close parenthesis times the forecast we made for that last
• 9:29
  period. F sub T -, 1.
• 9:31
  OK. F sub T equals alpha times D sub
• 9:34
  t -, 1 + 1 minus alpha times F sub T - 1.
• 9:37
  So that alpha, the smoothing constant, that's like how much
• 9:41
  we trust the very latest sales number.
• 9:42
  If alpha is high, we're basically saying whatever
• 9:45
  happened last week is probably what's going to happen.
• 9:47
  Next week, exactly. A high alpha makes the forecast
• 9:49
  very responsive, maybe even jumpy if sales suddenly spiked.
• 9:54
  A high alpha makes the next forecast jump right up to a low
• 9:57
  alpha. Smooth things out more, relying
• 10:00
  more on the historical average. So once you have that high level
• 10:03
  forecast for the finished product, you need to translate
• 10:06
  it into orders for all the bits and pieces.
• 10:08
  That's where Materials Requirements Planning, or MRP
• 10:11
  comes in. MRP takes the master production
• 10:13
  to schedule to plan for finished goods and explodes it down into
• 10:16
  detailed schedules for every single component, screw, nut and
• 10:20
  bolt needed. And the key thing about MRP, the
• 10:22
  source stressed, is that it's designed for dependent demand.
• 10:25
  What's the difference between dependent and independent demand
• 10:28
  here? Good question.
• 10:29
  Independent demand is the demand for the final product that the
• 10:33
  customer buys, say the number of complete bicycles sold each
• 10:36
  month that's driven by the market.
• 10:38
  Dependent demand is the demand for the components needed to
• 10:41
  build that final product, the tires, chains, pedals, frames
• 10:45
  for those bicycles. That demand is directly
• 10:47
  dependent on how many bikes we decide to build, and it's
• 10:50
  usually very lumpy, very irregular.
• 10:52
  You might need parts for a batch of 1000 bikes.
• 10:55
  One more and 0 the next. I see.
• 10:58
  So older inventory models often assume demand is smooth and
• 11:01
  constant. But MRP is built for that lumpy
• 11:04
  dependent demand reality, which lets companies order parts much
• 11:08
  closer to when they're actually needed.
• 11:10
  Precisely. That's the huge win with MRP.
• 11:12
  It helps you minimize your average inventory levels because
• 11:15
  you're not holding mountains of parts just in case you order
• 11:17
  them based on the actual production plan.
• 11:19
  Less cash tied up in stock. Makes perfect sense.
• 11:22
  OK, so we have the materials planned.
• 11:24
  Now sequencing on the factory floor, you often have multiple
• 11:27
  different jobs waiting to be processed on the same machine.
• 11:30
  How do you decide which one goes first?
• 11:31
  Is it does first first come, first served sometimes, but
• 11:35
  often not the most efficient way.
• 11:37
  A very powerful rule for sequencing is the shortest
• 11:40
  processing time. Rule or SPTSPT says that to
• 11:43
  minimize the average time job spend waiting in the system,
• 11:46
  that's called a mean flow time, and also to minimize the average
• 11:49
  lateness of jobs, you should sequence the jobs in order of
• 11:53
  their processing time, starting with the quickest job first.
• 11:57
  Wait. Run the shortest jobs first?
• 11:58
  That sounds counterintuitive maybe?
• 12:00
  What if I have a really big high priority, high profit job
• 12:04
  waiting? Shouldn't that jump the queue?
• 12:05
  From a pure immediate profit perspective on that one job,
• 12:08
  maybe. But SPT optimizes the overall
• 12:11
  system flow by getting the quick jobs out of the way fast.
• 12:14
  You clear the decks, reduce congestion, and lower the
• 12:17
  average wait time for all jobs waiting behind them.
• 12:20
  It increases the overall throughput and reduces the total
• 12:22
  work in progress inventory. It's about system efficiency,
• 12:26
  not just one jobs priority. OK.
• 12:27
  That's a fascinating trade off, optimizing the whole system's
• 12:31
  flow versus maybe maximizing immediate return on one specific
• 12:34
  order. It's a constant balancing act,
• 12:36
  and that sequencing challenge gets magnified massively on
• 12:40
  something like a mixed model assembly line.
• 12:42
  Oh yeah, like a car plant, building sedans, SUV's,
• 12:46
  different trim levels, all on the same line.
• 12:49
  Must be a nightmare. It's incredibly complex.
• 12:52
  The big challenge there is called production leveling.
• 12:54
  You're trying to smooth out the demand for components coming
• 12:57
  from all the upstream workshops feeding that final line.
• 13:01
  If you bunch too many complex option heavy cars together, you
• 13:04
  might starve the line of certain parts.
• 13:06
  If you run only simple models, you might build up excess
• 13:08
  inventory of other parts. It requires very sophisticated
• 13:11
  sequencing. OK, so we've designed it, costed
• 13:14
  it, planned it, scheduled it. Let's look at actually making
• 13:17
  it, the tools, the materials, and crucially, the quality.
• 13:21
  Starting with some basic challenges like just machining
• 13:23
  plastics. Seems simple, but maybe not.
• 13:26
  Yet plastics can be tricky. They're generally poor
• 13:29
  conductors of heat, so when you cut them, the heat builds up
• 13:32
  right at the tool tip. This makes the tool run really
• 13:35
  hot, and for thermoplastics especially, that heat can soften
• 13:39
  the plastic, making it gummy, causing it to bind or clog the
• 13:42
  tool. Which is why, particularly for
• 13:44
  faster speeds or longer cuts, carbide tools often end up being
• 13:47
  more economical than standard high speed steel.
• 13:50
  They just handle the heat better.
• 13:51
  Right. They stay harder at higher
• 13:52
  temperatures. OK, so that limitation, heat or
• 13:55
  sometimes extreme hardness of the material that leads us to
• 13:58
  non traditional machining or NTM.
• 14:00
  Why do engineers need these more exotic methods?
• 14:03
  We turn to NTM processes basically when conventional
• 14:06
  cutting tools just won't work well or at all.
• 14:09
  There are three main drivers. First, if the shape you need is
• 14:12
  really complex or maybe incredibly small.
• 14:15
  Second, if the material itself is just too hard, typically
• 14:18
  something over about 400 Brunel hardness.
• 14:21
  And 3rd, when you need extremely tight tolerances and you
• 14:24
  absolutely need to avoid introducing residual stresses
• 14:27
  into the part, which conventional cutting can
• 14:29
  sometimes do. NTM uses energy sources like
• 14:32
  light, electricity, chemicals instead of physical force.
• 14:35
  OK, give us an example. How powerful are we talking?
• 14:38
  Laser beam machining? LBML.
• 14:40
  BM is pretty incredible. It uses a super focused beam of
• 14:42
  light to melt and vaporize material away locally.
• 14:45
  The energy density is so intense it can basically sublime turn
• 14:49
  directly from solid to gas almost any material.
• 14:52
  We're talking temperatures potentially reaching 41,000°C.
• 14:56
  You can etch or cut incredibly fine features into very hard
• 14:59
  materials with it. 41,000° C and then something quite different.
• 15:03
  Electrochemical grinding ECG. Yeah, ECG is cool because it's
• 15:07
  mostly not mechanical grinding. It combines an abrasive wheel,
• 15:11
  yes, but the main material removal mechanism is a nautic
• 15:15
  dissolution, basically electrochemical depleting.
• 15:18
  About 95 to 98% of the material comes off through this
• 15:21
  electrical action. Only a tiny fraction, maybe 2 to
• 15:24
  5%, is actual mechanical abrasion by the wheel.
• 15:27
  This means very little force, very little heat, and minimal
• 15:31
  tool wear, even on hard materials.
• 15:32
  Fascinating. So we have tools for almost any
• 15:35
  job, but how do we ensure quality when you're turning out
• 15:38
  thousands, maybe millions of parts?
• 15:40
  Checking every single 100% inspection sounds ideal, but
• 15:43
  often isn't practical, right? It's usually worse than
• 15:46
  impractical, it's often ineffective, it's incredibly
• 15:49
  costly, it slows everything down, and honestly human
• 15:52
  inspectors get fatigue and miss things anyway.
• 15:54
  You still don't catch all the defects.
• 15:56
  That's why industry relies heavily on Statistical quality
• 15:59
  control, or SQC. We use carefully designed
• 16:03
  sampling plans. We inspect a small,
• 16:05
  statistically representative sample of parts from a batch or
• 16:08
  production run, and based on the results from that sample, we
• 16:11
  make a decision about the quality of the entire batch.
• 16:13
  OK, using statistics to infer the quality of the whole lot.
• 16:17
  And this brings us to a really crucial conflict point, the
• 16:20
  machines capability versus the designers specification, the
• 16:24
  idea of capability limits versus specified tolerance.
• 16:27
  This feels like where the rubber really meets the road.
• 16:30
  It absolutely is a machines capability.
• 16:32
  Limits tell you what that machine can actually produce
• 16:34
  consistently. They're typically calculated as
• 16:37
  X bar ±3 Sigma. X bar is the average or mean
• 16:41
  dimension the machine produces, and Sigma is the standard
• 16:44
  deviation, a measure of its variability.
• 16:47
  So X bar ±3 Sigma defines the range where about 99.73% of all
• 16:53
  parts made on that machine will naturally fall, assuming the
• 16:56
  process is stable. That's its inherent capability.
• 16:59
  OK, that's what the machine does, and the specified
• 17:01
  tolerance is what the designer wants it to do, the range they
• 17:04
  put on the drawing. Exactly, and here's the critical
• 17:06
  check. If the tolerance range specified
• 17:09
  by the designer on the drawing is tighter, narrower than the
• 17:12
  machines calculated capability limits that ±3 Sigma range, and
• 17:17
  that machine is fundamentally incapable of reliably meeting
• 17:19
  the specification, it will naturally produce some parts
• 17:22
  outside the required tolerance. In that situation, you have no
• 17:25
  choice. You must implement 100%
• 17:28
  inspection despite all those drawbacks, because you know the
• 17:30
  process itself will generate defects.
• 17:33
  Or you have to buy a much more precise and likely much more
• 17:35
  expensive machine. Or maybe even go back and argue
• 17:39
  with the designer to relax the tolerance of possible.
• 17:42
  Which brings us right back around to DFM and A and cost.
• 17:45
  Precisely. It's all interconnected design
• 17:47
  choices, process capability, cost and quality.
• 17:51
  So looking back, we've really covered the whole engineering
• 17:53
  arc today, haven't we? We started with setting the cost
• 17:55
  discipline through the unit manufacturing cost target and
• 17:58
  simplifying the product itself with DFM and A.
• 18:00
  Then we move it into managing the inherent complexity using
• 18:03
  classification systems for parts and materials and planning the
• 18:06
  factory flow with forecasting and sequencing rules like SPT.
• 18:10
  And finally, we look at the tools, the quality checks like
• 18:13
  SQC and that critical link between process capability and
• 18:16
  design tolerance. It's these engineering tools
• 18:19
  that keep complexity manageable and ensure efficient quality
• 18:22
  production. It's a fascinating system and
• 18:24
  you know, we really emphasized how reducing park count is the
• 18:27
  cornerstone of cost reduction in traditional DFM and a But
• 18:31
  thinking ahead, I want to leave our listeners with a question,
• 18:34
  something maybe not directly in today's sources, but a logical
• 18:37
  extension. Consider the rise of additive
• 18:40
  manufacturing, 3D printing. If you can now potentially print
• 18:44
  a highly complex assembly, maybe something that used to be 10
• 18:47
  parts as a single integrated piece, does that core DFMNA
• 18:51
  principle of fewer parts is always better still hold in
• 18:55
  quite the same way? Or has the complexity and
• 18:57
  therefore the focus for cost reduction shifted?
• 19:00
  Has it moved away from assembly labor and part count and more
• 19:03
  towards the material science itself, the printing parameters,
• 19:06
  the post processing? It changes the equation,
• 19:08
  doesn't. It it absolutely does.
• 19:10
  A very interesting thought. Until the next equation design
• 19:12
  and junkies, we are mechanical engineering made simple.