Latest / Mechanical Engineering Made Simple / (#141) Why Flawless Engineering Drawings Fail in Reality

Transcript

(#141) Why Flawless Engineering Drawings Fail in Reality · Mechanical Engineering Made Simple

0:00
Welcome back to Mechanical Engineering Made Simple.
0:03
Wait, no, let me read the intro correctly.
0:05
Welcome back to Mechanical Engineering Made Simple.
0:07
I'm your host, and today we're exposing the gap between what
0:10
school drilled into your head like gospel and what actually
0:14
happens when real parts, real dirt, and real deadlines show
0:18
up. Stick around.
0:20
This could save you from learning the hard way on the
0:22
shop floor. If you haven't already, hit that
0:24
like and follow and drop a comment if you enjoyed it.
0:27
We love hearing from the folks who keep this podcast running.
0:30
Absolutely. O imagine this you.
0:32
You just sent six months designing this high pressure
0:36
hydraulic valve. Right.
0:37
Oh boy, I know where this is going.
0:39
Yeah, you ran the finite element analysis.
0:41
Yeah, your CAD model is pristine.
0:43
Every tolerance is just, you know, mathematically perfect.
0:45
Flawless on the screen. Exactly.
0:47
You send the drawings out, the parts get machined, they pass
0:51
every single quality control inspection in the lab, and they
0:54
get installed on a customer's heavy machinery in the field.
0:56
And everyone is happy. Until they aren't, because three
1:00
weeks later the valve completely shatters in the field, just
1:03
sprays hydraulic fluid everywhere, shuts down a
1:07
massively expensive production line.
1:09
Your boss is furious, the customer is threatening
1:12
lawsuits, and you were just sitting there staring at this
1:15
completely flawless engineering drawing, asking yourself, you
1:20
know, what on earth went wrong? It is.
1:23
I mean, it's the ultimate engineering nightmare, right?
1:25
And yet it happens every single day.
1:27
Every day, because we are taught, especially in
1:30
university, to view mechanical engineering as this binary
1:34
perfect world. You design A bracket, the CAD
1:37
software shows those crisp infinite lines and you think,
1:40
well, I've solved the problem. Right, because the math works.
1:42
Yeah, but the real world doesn't care about your CAD model or
1:45
your math. The real world introduces
1:47
unpredictable human behavior. Thermal expansion, microscopic
1:51
tool wear, and, you know, global supply chain miscommunications.
1:54
Which is exactly our mission for this deep dive.
1:57
Today we're talking to you, the listener, the engineer, the
1:59
designer, the machinist, who is just tired of that disconnect.
2:02
We're tearing it all down today. We are taking a hard, unfiltered
2:06
look at the messy, chaotic reality of mechanical
2:09
engineering. We're going to breakdown the
2:11
biggest pain points you face in the wild.
2:13
Call out these massive misconceptions that are honestly
2:17
still somehow being taught in textbooks.
2:19
All the textbook myths are the the worst.
2:20
They really are, and we'll show you how to avoid these traps and
2:24
answer the most common questions you have about bridging that gap
2:27
between theoretical design and the actual dirty shop floor.
2:32
And we aren't just giving you our, you know, our personal
2:35
gripes or opinions here. We are grounding this and hard
2:38
data, which I love. Yeah, we're looking at a massive
2:40
global survey from the now naming the Pain and Requirements
2:44
Engineering Initiative. It's known as Napier.
2:47
We're also tearing apart the actual text of the ASME and
2:50
isodimensioning and tolerancing standards.
2:53
Plus, we'll be looking at real world metrology audits, the
2:56
harsh reality of tolerance stack up models and failure analysis
3:00
frameworks from the Mechanical Engineers Handbook.
3:02
It is a comprehensive breakdown of why parts fail and why
3:06
projects go off the rails. So I want to start right at the
3:10
genesis of every project, the requirements phase.
3:13
The infamous requirements phase. Right, because think back to
3:17
your engineering exams or your senior design project.
3:19
You sit down and you were handed a perfectly defined problem.
3:23
Here are the Givens. Here is what you need to find.
3:26
Exactly. The constraints are clearly
3:28
listed, the performance goals are bulleted out, and all you
3:31
have to do is just crank through the math to find the one right
3:34
answer. But the data shows that this
3:36
this locked blueprint concept is a complete illusion in the real
3:40
world. It's a total fantasy.
3:42
According to that neighbor E data we pulled, requirements are
3:45
actually the number one cause of project pain across the board.
3:48
It really is a staggering reality check.
3:50
And, you know, we need to emphasize that this wasn't some
3:52
small localized survey of a few disgruntled engineers.
3:56
Right. It was huge.
3:57
This covered over 200 companies across 10 different countries,
4:01
spanning multiple industries. The data totally shatters the
4:05
myth that you can just get a spec sheet from a client, lock
4:08
it in a drawer, and start designing, right?
4:11
The top pain points they reported weren't about complex
4:14
physics or material shortages. They were incomplete or hidden
4:18
requirements, massive communication flaws between the
4:22
project team and the customer, and what the researchers call
4:26
moving targets. Moving targets, so just
4:28
constantly, changing goals constantly.
4:31
But wait, I'm going to push back on the customer a little bit
4:33
here. I mean, isn't it fundamentally
4:35
the client's job to know what they are paying us to build?
4:39
You would certainly think so. If they had me a specification,
4:42
why can't I just treat that like a boundary condition and
4:44
thermodynamics and just get to work?
4:46
I mean, that's where engineers get tripped up, right?
4:48
The data points to this fascinating mix of human
4:51
psychology and systemic corporate errors.
4:54
The primary causes for these problems weren't technical at
4:57
all. So what were they?
4:58
Well, first, the survey highlighted a severe lack of
5:00
experience within the requirements engineering teams
5:03
themselves. But more importantly, the root
5:05
cause almost always traces back to what they politely call
5:09
severe time boxing. Severe time boxing, which is
5:13
corporate, speak for unrealistic punishing deadlines.
5:16
Precisely picture this scenario because I guarantee it happens
5:19
constantly to our listeners. It's Friday afternoon.
5:22
Always Friday afternoon. Always the sales and marketing
5:25
team is trying to close a massive contract to hit their
5:28
quarterly quota to win the bid. They promised the client a
5:32
custom, fully integrated system with a ridiculously aggressive
5:36
delivery date, and then they just throw these wishful
5:39
promises over the wall to the engineering department.
5:42
They create a hard deadline before the engineering team has
5:44
even had a single chance to evaluate the scope of work.
5:47
Or even ask the client basic functional.
5:50
Questions. Exactly.
5:51
So the engineers start with a metaphorical gun to their head,
5:54
rushing to design based on an incredibly abstract under
5:57
specified wish list. And because the client
6:00
themselves often doesn't actually know what they need.
6:03
I mean, they only know the symptom they want to fix, right?
6:05
Yes, exactly. They know their current machine
6:08
is too slow or too loud. They don't know the engineering
6:11
solution. So the initial requirements are
6:13
wildly vague. They allow for just massive
6:16
variations in interpretation. And this completely reframes how
6:20
I view my education, honestly. We spend four years in
6:23
university treating requirements as a strict math problem, but
6:28
based on this Napier data, it's actually a human communication
6:31
problem. Or even the corporate politics
6:33
problem. It's entirely a communication
6:35
problem, and if you as the engineer assume a specification
6:38
is locked just because a customer wrote it down in a
6:42
preliminary e-mail, you are setting the entire project up
6:45
for a late stage catastrophic failure.
6:48
OK. But what about the whole agile
6:49
versus waterfall debate? You hear companies say, oh we
6:52
use a plan driven waterfall approach, so our requirements
6:55
are perfectly locked before we draw a single line of cash.
6:58
Yeah, people love to say that. Does the data back that up at
7:00
all? Not even slightly.
7:02
The Napier E study actually broke the data down by company
7:05
size and process model, explicitly comparing agile
7:09
environments to plan driven waterfall environments.
7:12
And what do they find? The results were incredibly
7:15
revealing. Even in companies where the
7:18
entire corporate philosophy is predicated on planning
7:21
everything perfectly upfront before doing any actual work,
7:25
moving targets were still listed as a top issue.
7:27
Wow, the idea that you can freeze a design in the real
7:30
world is just a fantasy. It reminds me of a chef trying
7:34
to execute a Michelin star recipe.
7:36
You know, engineering driven design where you just blindly
7:39
follow the initial spec. It's like the chef writing an
7:42
incredibly complex 10 course menu without ever checking what
7:45
ingredients are actually in the pantry and completely forgetting
7:49
to ask the diner if they have any severe food allergies.
7:51
And then being absolutely shocked when the diner sends the
7:53
meal back. It's a great analogy.
7:55
And to take it further, the diner, in this case the
7:58
customer, is constantly changing their order while the food is
8:01
already on the stove. Like actually I want chicken
8:03
instead of beef while the beef is already searing.
8:06
Exactly. Exactly, and the study showed
8:08
that in both large organizations and small agile companies, these
8:12
changing goals lead directly to blown budgets, severe time
8:17
delays, and massive post implementation rework.
8:21
Oh, the rework. The rework has to be the most
8:23
demoralizing part for an engineer.
8:24
It destroys morale. The respondents in the study
8:27
specifically cited poor product quality as a direct result of
8:30
hidden requirements. Because when the target moves in
8:32
the 11th hour, you aren't engineering an elegant solution
8:36
anymore. No.
8:37
You're just duct taping a rushed patch onto an existing design to
8:41
meet the deadline. Yeah, and in small agile
8:43
companies it's often worse. Really.
8:45
Why is that? Because they rely on continuous
8:48
customer feedback, right? Yeah, but the customers are
8:50
frequently too busy to actually attend the Sprint meetings.
8:53
Oh, that's so true. They just ghost the engineering
8:55
team. Yeah.
8:57
So the engineers are left sitting in a room guessing what
8:59
the customer functionally needs, and when requirements remain
9:03
abstract, you invariably build something that might function
9:07
perfectly on paper, but fails completely in the user specific
9:11
real world context. OK, so let's play out a best
9:13
case scenario. Let's pretend you miraculously
9:17
survive the requirements phase. A rare miracle, but OK.
9:20
You have incredible communication, You dig out all
9:23
the hidden requirements, you lock down the functional needs,
9:25
and you draft the perfect geometric drawing.
9:28
You set it off to quality control lab or the shop floor.
9:31
You feel great. Great time to celebrate, right?
9:33
Right. But the moment that drawing hits
9:35
the installment room, here's everything falls apart again.
9:38
Oh, it really does. Let's talk about the absolute
9:40
reality check of geometric dimensioning and tolerancing, or
9:43
GD and T, and the story of the inspector who essentially went
9:48
rogue. This is one of my absolute
9:50
favorite topics because the gap between theory and reality here
9:54
is just startling. If you walk onto a shop floor
9:57
and ask a junior engineer what an engineering growing does,
10:00
they will tell you it tells the manufacturer exactly how to make
10:03
and measure apart, but it absolutely does not.
10:06
Hang on, if I'm looking at a blueprint loaded with GD and T,
10:10
you know, positional tolerances, flatness, perpendicularity, It
10:15
looks like a highly rigorous, strict set of instructions.
10:17
It is rigorous, yes, but it describes A theoretical
10:20
boundary. It does not describe a physical
10:22
action. The ASMEY 14.5 standard, which
10:26
is the Bible for dimensioning and tolerancing in North
10:29
America, explicitly defines theoretical tolerance zones.
10:33
In fact, if you read the mathematical companion to that
10:35
standard ASMEY 14.5 point one, it explicitly States, and I
10:40
quote this document is not intended it as a gauging
10:43
standard. Wait, what?
10:44
So it tells you what the Final Shape needs to fit inside of,
10:46
but it completely ignores the tools used to verify it.
10:49
Yes. It does not dictate whether the
10:51
quality tech should use a pair of hand calipers, a hard
10:55
functional gauge, an optical comparator, or a half million
10:58
super expensive coordinate measuring machine.
11:01
You know CMM. And that leads to this
11:03
phenomenal concept in the sources called Methods
11:05
Divergent. I really want you to break this
11:07
down for the listener because the implications of this are
11:09
huge. Methods divergent is the
11:11
terrifying reality where 2 completely valid industry
11:15
accepted measurement methods give you 2 totally different
11:19
results on the exact same physical part.
11:21
That is terrifying. To illustrate this, let's look
11:24
at a radical real world case study from an audit of a
11:26
contract inspection lab. Yes, the So what inspector set
11:30
the scene for us. OK, so an auditor goes into this
11:32
lab to verify their processes. They are looking at a drawing
11:36
for a large cylindrical component.
11:38
The drawing has a GDNT call out for the perpendicularity of the
11:41
cylinder's axis, and the drawing explicitly, unequivocally states
11:46
that this perpendicularity must be measured relative to datum A.
11:50
That's the primary mounting phase.
11:52
So the instructions are clear. Datum A is the anchor.
11:55
There's no ambiguity there. Crystal clear, but the auditor
11:58
watches the inspector set the part up on the surface plate,
12:01
and the inspector completely ignores datum A just.
12:04
Ignores it. Completely.
12:05
Instead, they indicate the part and measure the perpendicularity
12:09
relative to a completely different surface.
12:11
Let's just call it Surface B. OK, as an engineer, my blood
12:14
pressure is spiking just hearing that they just ignored the
12:18
drawing. I assume the auditor immediately
12:20
flagged this as a critical failure.
12:21
Well, the auditor stopped them and asked why are you measuring
12:25
off surface B when the print explicitly calls out datame and
12:28
the inspector looked at the auditor and essentially said So
12:31
what? So what the sheer audacity of.
12:34
That I know it sounds incredibly audacious, but you have to look
12:37
at the context of the physical part itself.
12:39
The perpendicularity tolerance allowed on the drawing was huge.
12:43
It was 0.4mm. Which is pretty massive for
12:47
machining. That is a massive tolerance zone
12:49
for a precision machined part. The inspector knew based on
12:53
their experience with the machining process that the
12:56
microscopic difference in orientation between datum A and
12:59
surface B was practically non existent compared to that
13:02
massive 0.4mm zone. So they basically made a
13:06
judgment call. They decided the error
13:08
introduced by using the wrong surface was so tiny it wouldn't
13:11
affect the final pass or fail result.
13:14
Exactly. They optimized their setup time
13:16
for the business because the measurement uncertainty easily
13:19
met the functional requirements of the part.
13:21
But it highlights the core issue here.
13:23
The drawing is not an absolute law.
13:26
It requires interpretation and judgement by the person holding
13:29
the tools. It really is like the legal
13:30
system. Yeah, the 2D drawing is the
13:33
written law passed by the legislature.
13:35
It's supposed to be black and white, but the inspector or the
13:38
software they are using is the judge in the courtroom.
13:41
That is a perfect way to look at it.
13:42
And we all know that two different judges can interpret
13:44
the exact same written law in completely opposite ways based
13:48
on precedent, the context of the trial, or even just their mood
13:51
that day. That is a brilliant analogy.
13:54
And if we want to talk about judges, the most powerful judge
13:57
in modern manufacturing isn't a human at all.
14:00
Let me guess, it's the software. It is the CMM software.
14:05
Modern coordinate measuring machines are absolute marvels.
14:09
They use a Ruby tipped probe to touch a physical surface
14:12
thousands of times, generating a massive cloud of data points in
14:16
3D space. Right, but a cloud of dots isn't
14:19
a diameter. How does the machine turn a
14:22
million microscopic dots into a single number on an expection
14:26
report? Right, it has to run some kind
14:27
of mathematical fitting algorithm.
14:29
Exactly, and the most common default algorithm in almost all
14:33
CMM software is called least squares.
14:36
OK, let's pause and explain least squares simply without
14:39
getting bogged down in the heavy math.
14:41
Sure. Imagine you have a bunch of push
14:43
pins stuck into a cork board in a rough circle.
14:45
OK, I'm picturing it. These squares is like taking a
14:48
flexible wire ring and trying to place it among the pins so that
14:51
the total distance from every single pin to the wire is
14:54
minimized. So it's finding the mathematical
14:56
center. Yes, it averages out the highs
14:59
and the lows to find the best possible mathematical center.
15:03
It is incredibly mathematically stable and highly repeatable.
15:06
Which sounds great. Engineers love repeatability.
15:10
But what happens when reality gets messy?
15:13
What if the machinist left a tiny Burr on the part, or a
15:17
literal speck of dirt falls onto the cylinder before the probe
15:20
touches it? This is exactly where least
15:23
squares becomes dangerous. If your part has a speck of dirt
15:26
on it, or if the lathe was chattering and created a three
15:28
lobe shape instead of a perfect circle.
15:31
The least scores algorithm just blindly averages that dirt and
15:34
those lobes right into the diameter calculation.
15:38
It pulls the theoretical circle outward to accommodate the
15:40
average. But the part isn't going to
15:41
average out the dirt when you try to shove it into a mating
15:44
hole on the assembly line. The physical hole doesn't care
15:47
about the mathematical average, it only cares about the physical
15:49
extremes. The dirt is going to crash into
15:52
the wall of the hole. Precisely.
15:54
The real world doesn't care about averages, it cares about
15:57
absolute maximum material conditions.
16:00
This is why a different algorithm, called the minimum
16:02
circumscribed algorithm is functionally much more accurate
16:06
for an external feature like a shaft.
16:08
Let's use an analogy for minimum circumscribed.
16:11
If least squares is averaging the pins, what is minimum
16:15
circumscribed doing? Imagine taking a rigid steel
16:18
ring and trying to slide it over those pins on the cork board.
16:22
The ring will stop as soon as it hits the two or three outermost
16:24
pins, the absolute highest points.
16:27
It simulates a perfect theoretical cylinder collapsing
16:29
around the absolute outermost extremes of the physical part,
16:32
dirt burrs and all. OK, so if I'm visualizing this,
16:36
I have a shaft. It has a tiny speck of metal
16:39
stuck to the side. The quality tech measures it
16:41
with the CMM. If the software drop down menu
16:44
is set to Lee squares, it averages the spec into the rest
16:47
of the data. The calculated diameter stays
16:49
relatively small and the software says the part passes
16:52
inspection. Yes, it passes, but if the tech
16:55
goes to lunch, a different tech comes back.
16:57
Measures the exact same physical shaft with the exact same
17:01
machine, but changes the software drop down menu to
17:04
minimum circumscribed. The software suddenly simulates
17:07
the real mating hole. It catches that high point, the
17:10
calculated diameter balloons outward, and the part fails.
17:13
Exactly the exact same physical part passes or fails entirely
17:18
based on an invisible software algorithm that the quality
17:21
technician clicked on. That is wild.
17:24
And what terrifies me is that many design engineers sitting in
17:27
their cubicles have absolutely no idea these algorithms even
17:30
exist, let alone that they are dictating the fate of their
17:33
tolerances. That is deeply unsettling.
17:35
I mean, if you are listening to this and relying on least
17:37
squares algorithms right now for tight tolerance mating parts,
17:40
you might be passing bad parts and sending them to the field
17:43
without ever knowing it. And industry has known about
17:46
this danger for decades. Back in 1988, there was a major
17:50
government industry data exchange program alert, a GID AB
17:53
alert, that actually mandated stopping the use of CMMS for
17:57
certain geometric measurements in the aerospace and defense
18:00
sectors. 1988. Yes, the core issue is that CMMS
18:05
were verifying parallelism without accounting for the
18:07
physical form error of the surface, completely violating
18:11
the functional intent of the design.
18:13
And yet we are still fighting these exact same methods,
18:16
divergent battles today. So just to recap our nightmare
18:19
so far, your requirements changed at the last minute, the
18:22
drawing was interpreted loosely by a rogue inspector, and the
18:25
CMM software might be hiding burs using math.
18:28
Sound like a standard Tuesday? Now let's turn the dial up on
18:31
the chaos. What happens when you take that
18:33
drawing and send it to an overseas manufacturer?
18:36
Because this is where the global clash between ASME and ISO
18:38
standards creates absolute carnage.
18:41
This is a trap that catches even the most seasoned veteran
18:44
engineers. You see, engineering symbols
18:46
look like a universal language. They really do.
18:49
Whether you are in Detroit, Stuttgart or Shenzhen, a circle
18:53
with a cross through it means position, and two parallel lines
18:57
mean parallelism. So engineers naturally assume
19:01
the definitions behind the symbols are identical globally.
19:04
It's a fair assumption, right? Yeah.
19:06
I mean, math is math. It is the most expensive
19:08
assumption you can make. While the symbology is nearly
19:11
identical, metrology experts estimate that up to 65% of
19:16
geometric tolerances are either mathematically defined or
19:19
interpreted differently between the North American standard
19:22
ASMEY 14.5 and the international standard ISO eleven O 1.
19:28
Wait, stop right there. 65% yes, more than half of the
19:31
engineering drying language is essentially a false cognate.
19:34
Yes, it's staggering. Let's look at one of the most
19:36
dangerous examples, concentricity.
19:38
Imagine you draw 2 circles inside each other and you put a
19:41
concentricity symbol pointing to them.
19:43
If you are an American engineer designing under ASME rules,
19:46
concentricity is an incredibly strict, complex control.
19:50
It controls the opposed median points of a feature.
19:52
Meaning what? Practically speaking.
19:54
Meaning you literally have to analyze the cross sections of
19:58
the part to ensure the mass is perfectly evenly distributed
20:01
around center axis. It's often used for high speed
20:05
spinning parts to prevent vibration.
20:07
It's incredibly difficult to measure.
20:09
You essentially need an optical comparator or complex CMM
20:13
routines to find the median points of every single cross
20:16
section. OK, but what if I send that
20:18
exact same drawing with that exact same symbol to a shop in
20:21
Europe operating under ISO rules?
20:24
Under ISO, that identical symbol means something entirely
20:27
different. It simply controls the axis of
20:30
the actual mating envelope. Really, Yes.
20:32
It does not care about opposed median points or mass
20:35
distribution. It just means the theoretical
20:37
center line of the cylinder has to be within a certain zone.
20:41
It's a completely different mathematical definition,
20:43
requiring entirely different, much simpler inspection
20:46
routines. So sending an ASME print to an
20:48
ISO shop is basically like using a Spanish dictionary to read an
20:51
Italian book. Exactly.
20:53
And to take that further, you might recognize the word Burrow
20:56
and think you understand the sentence, but in Spanish you
20:59
were talking about a donkey, and in Italian you just said butter.
21:03
And on the shop floor, that butter is a high speed shaft
21:06
that passes ISO inspection, gets installed in your machine, and
21:09
immediately vibrates itself to pieces because the European shop
21:13
didn't balance the opposed median points like the American
21:16
engineer thought they would. Exactly.
21:18
And you, the designer, are the one paying for the wreckage.
21:21
You are paying for it because you failed to specify the
21:24
language. But the divergent goes even
21:27
deeper than complex GD and T symbols.
21:29
It affects the most basic foundation of mechanical design,
21:33
a simple size tolerance, like a basic diameter call out.
21:37
How can you mess up a diameter? It's literally just the distance
21:40
across a circle. It comes down to a philosophical
21:43
difference in manufacturing history.
21:45
Under North American ASME rules, we have something called rule
21:48
#1, also known as the envelope principle.
21:51
Number one. This rule dictates that a size
21:53
tolerance doesn't just control the local size, it automatically
21:56
controls the form of the part as well.
21:58
Give me a visual of that. Imagine you have a shaft with a
22:02
maximum allowable diameter of 1 inch.
22:05
Rule #1 says that if the shaft is machined right at that one
22:08
inch maximum limit, what we call maximum material condition, it
22:12
must be perfectly straight and perfectly round.
22:15
It must not violate A theoretical perfectly
22:18
cylindrical envelope of 1 inch. Which makes perfect sense
22:21
because you are guaranteeing that the chef will actually
22:25
slide into a one inch hole. In the real world, you are
22:28
protecting the physical assembly.
22:30
Exactly. It's rooted in the American
22:32
history of mass production and interchangeable parts, but
22:35
internationally, ISO approaches this entirely differently.
22:38
ISO operates on the default principle of independency.
22:41
Independency meaning what? This means a size tolerance only
22:44
controls the local 2 point size, like measuring across the shaft
22:48
with a micrometer at various spots.
22:49
It does not automatically control straightness or
22:51
roundness. So under ISO, I could have a
22:54
shaft that measures exactly 1 inch thick at every single point
22:57
along its length, but the entire shaft is bent like a banana.
23:00
Yes, understand it ISO rules. Unless you explicitly add a
23:05
specific modifier symbol to the drawing, like a letter E inside
23:09
a circle to invoke the envelope requirement, that banana shaped
23:12
shaft is a perfectly legal acceptable part.
23:16
You've got to be kidding me. Nope, it meets the print.
23:18
The fact that it will never, ever slide into the mating
23:21
assembly is not the machinist's problem, it's yours.
23:24
So an American engineer sends a basic drawing to Asia, assumes
23:28
rule #1 is universally understood, forgets to add the
23:31
little E symbol because they don't even know ISO rules exist,
23:34
and receives a shipping container full of banana shaped
23:36
shafts that measure perfectly but can't be assembled.
23:39
And the manufacturer will demand to be paid because they made the
23:42
parts exactly to the ISO standard.
23:44
If the title block of your drawing doesn't explicitly state
23:47
whether ASME or ISO governs the interpretation, you have handed
23:51
them a blank check for ambiguity.
23:53
What about something insanely simple like, you know, rounding
23:56
off a sharp corner or radius tolerance?
23:58
Let's say I call out a corner to have a radius of 0.2 inches
24:02
±0.01. There is no way to mistranslate
24:06
a curve. You'd be shocked.
24:08
Even a simple radius is a complete nightmare on the shop
24:11
floor. Let's look at how ASME Y 14.5
24:15
defines a radius tolerance. It doesn't just create a simple
24:18
plus or minus boundary, it creates a Crescent shaped non
24:22
uniform tolerance zone that actually tapers down to 0 width
24:26
at the tangent points where the curve meets the flat edges.
24:30
Wait, it tapers to zero? Yeah, but if the tolerance zone
24:33
is 0 at the edges, that means you have absolutely 0 margin for
24:36
error during machining. Yes, that is physically
24:39
impossible to manufacture. It is functionally impossible to
24:42
manufacture perfectly, and it's incredibly subjective to
24:45
measure. The standard tries to help by
24:47
stating the physical curve must be a fair curve without
24:49
reversals. Fair curve?
24:51
What does that even mean? Exactly what defines a fair
24:54
curve? It's highly subjective.
24:56
Because of this bizarre geometry, measurement
24:58
repeatability for radius tolerances and quality labs is
25:01
notoriously awful. So how do we fix it?
25:03
Do we just stop using radius tolerances?
25:06
Many modern metrology experts strongly recommend exactly that.
25:09
They advocate abandoning plus or minus radius tolerances for
25:13
complex or critical curves entirely.
25:15
And use what instead? Instead, they suggest using a
25:18
profile of a surface geometric tolerance, which creates a
25:22
uniform, even boundary along the entire curve, making it easily
25:26
verifiable for the CMM and completely understandable for
25:30
the machinist. OK, we need to take a breath.
25:32
The print is a linguistic minefield.
25:35
Yeah. It really is.
25:35
But let's say we get past all that.
25:37
The German shop understands you want ASME rules.
25:41
The CMM is programmed right with the minimum circumscribed
25:44
algorithm. Now someone actually has to cut
25:46
metal. Here comes the physics.
25:48
Let's step onto the fabrication floor and talk about why physics
25:51
hates perfection, specifically looking at fabrication
25:54
nightmares and tolerance Stack us.
25:57
This is where we confront the physical violence of machining.
25:59
Textbooks teach a highly idealized, almost sterile
26:02
version of manufacturing. Right, like it's just math
26:05
happening to metal. They teach you that if you need
26:07
1/2 inch hole in a block of steel, you load 1/2 inch drill
26:10
bit into the drill press. You pull the lever, the metal
26:14
shavings fly and you get a perfect half inch hole.
26:18
Any machinist listening to this is laughing right now.
26:21
Because a drilled hole is practically never the exact size
26:24
of the drill bit used to make it.
26:26
It's almost always oversized. Why is that?
26:29
What is physically happening at the microscopic level of the
26:32
cutting edge? You have multiple physical
26:34
forces fighting against your perfectly straight line.
26:37
First, look at the machine itself.
26:40
The drill press spindle is rotating on bearings.
26:42
Those bearings have microscopic play, meaning the spindle weaves
26:46
or wobble slightly as it spins. This is called run out.
26:49
So the drill bit isn't just spinning, it's sweeping in a
26:52
tiny microscopic orbit. Exactly.
26:54
Second, look at the drill bit. A twist drill has cutting edges
26:57
called lips. If those licks are ground at
27:00
even slightly different angles or lengths, which happens
27:02
constantly as tools wear and are resharpened, the cutting forces
27:06
become unbalanced. The drill is constantly being
27:09
pushed off sender as it cuts. 3rd, as the metal heats up, the
27:13
chip can bind on one side of the flute, pushing the drill
27:16
laterally. All of these physical kinematics
27:19
guarantee that your hole will be slightly oversized, potentially
27:22
lobed like a rounded triangle, or tapered from top to bottom.
27:25
So you have a hole that is slightly oversized and weirdly
27:29
shaped, and let's say a shaft has to go into that hole, but
27:32
the shaft was turned on a lathe that was vibrating slightly, so
27:34
the shaft is slightly undersized and WAVY.
27:37
Now you're seeing it. When you try to assemble these
27:39
two imperfect parts, you run into tolerance stack up.
27:42
Yes, and this brings us to how engineers are taught to handle
27:46
that variation. Engineering schools heavily
27:49
emphasize worst case stack up analysis.
27:51
Which means what conceptually? It means designing the assembly
27:54
under the incredibly pessimistic assumption that every single
27:57
part in the assembly will roll off the manufacturing line at
28:00
its absolute worst allowable extreme.
28:04
It assumes the hole is as small as legally allowed and the shaft
28:07
is as fat as legally allowed simultaneously.
28:10
And if you do the math for that, worst case scenario, you
28:13
guarantee that the parts will always sit together 100% of the
28:17
time, no matter what. That sounds like good
28:19
engineering. It guarantees a fit, yes, but it
28:21
is a financially ruinous way to design high volume products.
28:25
If you assume the worst case scenario for 10 different parts
28:28
stacking together, the only way to make the final assembly
28:32
function is to tighten the individual tolerances on every
28:35
single part to microscopic, incredibly expensive levels.
28:39
I see if a worst case stack up is like packing for a weekend
28:42
trip, assuming you'll face a Blizzard, a heat wave, and a
28:46
category 5 hurricane all at the exact same time.
28:49
Aren't we just wasting the client's money buying massive
28:52
suitcases and gear we will never use?
28:54
Exactly, you are paying for perfection you don't need.
28:57
So how do you confidently move away from that without losing
29:01
sleep over the one in a million part that won't fit?
29:03
You move to statistical approaches.
29:05
The mathematical reality is that the odds of every single part in
29:08
an assembly hitting its worst case extreme at the exact same
29:11
time are astronomically low. So what's the alternative to
29:14
worst case? For high volume manufacturing
29:16
engineers use models like the root sum square or RSS method.
29:21
This model relies on normal distribution.
29:23
You know the bell curve, right? It assumes most parts will
29:26
naturally fall near the nominal center dimension.
29:29
By using the statistical sum of the variances, you can
29:32
mathematically prove that you can open up the individual part
29:36
tolerances, making them much cheaper and faster to machine,
29:39
while still guaranteeing that 99.73% of your assemblies will
29:43
go together perfectly. That sounds like a massive cost
29:46
savings, but is the bell curve always accurate?
29:49
I mean, a drill bit doesn't stay perfectly sharp forever.
29:52
You've hit on the exact flaw of basic RSS.
29:55
It assumes the manufacturing process is perfectly centered
29:58
and stays there, but in reality tools wear down.
30:01
Right. Friction is real.
30:02
A lathe cutting tool gets dull, and as it dulls, the diameter of
30:06
the parts it produces slowly drifts larger and larger.
30:08
Over a shift, the bell curve shifts.
30:11
So how do the pros handle tool drift?
30:13
They use an advanced model called the estimated mean shift.
30:17
I won't bore you with the dense algebra, but conceptually you
30:20
can think of the mean shift formula as a built in.
30:24
Pessimism dial. A pessimism dial?
30:26
I like that. Explain.
30:27
It allows the engineer to mathematically adjust the
30:30
tolerant stack based on how much they expect the specific machine
30:34
tool to drift off center over time.
30:37
Combined with the standard bell curve of the parts, you are
30:40
literally modeling the physical degradation of the shop floor
30:43
into your CAD math. Which requires the designer
30:46
sitting in an air conditioned office to actually understand
30:49
the physical reality of tool wear on a lathe.
30:52
This directly ties into the three design philosophies
30:55
mentioned in the Mechanical Engineer's Handbook.
30:56
Let's break those down. The handbook categorizes how
30:59
companies approach design. First is engineering driven
31:03
design. This is the classic arrogant
31:05
over the wall approach. Well, we all know this.
31:07
One, the engineer specifies incredibly tight tolerances
31:10
because the math says so, throws the drawing to manufacturing and
31:13
says figure it out. It usually results in
31:15
unmanufacturable parts and massive scrap rates.
31:18
Then you have process driven design.
31:20
Process driven is the opposite. The tolerances are chosen purely
31:24
based on what the machines on the floor can easily and cheaply
31:27
achieve. The risk here is that the shop
31:30
makes a perfectly cheap part that fails functionally because
31:33
the tolerances were too loose for the real world application.
31:37
And the third is inspection driven design where you only
31:40
dimension things that your quality lab has the tools to
31:43
easily measure, which sounds like the tail wagging the dog.
31:46
It is if all three of these in isolation are terrible.
31:50
What is the solution? Concurrent engineering.
31:53
Right, getting everyone together.
31:54
The designer, the machinist and the quality inspector must be in
31:57
the same room during the requirements phase, balancing
32:00
functional needs with machining capabilities and measurement
32:03
realities. OK, we have survived the
32:05
abstract requirements, negotiated the subjective CMM
32:08
algorithms, and mathematically accounted for the drill wobble.
32:12
Let's say you do everything right.
32:14
Sounds impossible, but let's say we do.
32:16
The part is machined perfectly to the drawing, it passes
32:19
inspection and it ships. The real world has one final
32:23
boss waiting for you, the physical environment.
32:27
Let's talk about material realities, failures and the
32:29
insidious size effect. This is where perfectly designed
32:32
parts mysteriously tear themselves apart in the field.
32:36
When engineers are designing a system, they need to know how
32:38
strong the metal is. Right, you have to know what
32:41
you're building with. So they pull up a database or a
32:43
standard material handbook. They might look up a standard
32:46
carbon steel forging like ASTMA One O 5 and the book says it has
32:51
a minimum yield strength of £36,000 per square inch.
32:56
Or they look at a 182 stainless steel and it says 40,000 PSI.
33:00
Those numbers look rock solid. They're published in
33:02
international standards. Why wouldn't I trust them?
33:04
Because you have to ask how those numbers were generated.
33:06
They are entirely valid for the highly controlled, pristine room
33:10
temperature laboratory conditions in which the tiny
33:13
material test coupons were pulled apart.
33:15
Keyword being tiny. Exactly, but material properties
33:18
are highly sensitive to how they are tested, and critically, the
33:22
physical volume of the final part you were designing.
33:25
This introduces the size effect. I want to focus on this size
33:29
effect because the source is highlighted specifically for
33:32
brittle materials like ceramics. Why does size matter so much
33:36
more for a ceramic ray Dome on a missile than it does for a steel
33:39
beam? It comes down to the mechanics
33:42
of failure. Ductile materials like steel
33:44
yield. If they get overstressed, they
33:46
stretch into form before they break, which redistributes the
33:49
load. They bend before they break.
33:51
Right. But brittle materials like
33:53
ceramics or glass do not not stretch.
33:56
They fracture suddenly, catastrophically.
33:58
And that fracture almost always originates from a microscopic
34:01
defect, a tiny pore, a micro crack, or an impurity in the
34:04
grain structure. So if I test a tiny one inch
34:07
ceramic coupon in a lab. The probability of a major
34:11
strength limiting flaw existing inside that tiny volume of
34:14
material is statistically very low.
34:16
The coupon shows incredible tensile strength, but let's say
34:21
you use that data to build a massive ceramic component that
34:24
is 3 feet wide. The volume of the material just
34:27
increased exonentially. And as the volume increases, the
34:30
mathematical probability that a microscopic catastrophic defect
34:34
exists somewhere inside that part skyrockets.
34:37
Precisely the overall strength of the bulk material drops
34:40
dramatically compared to the tiny lab sample.
34:42
You simply cannot extrapolate small scale test data directly
34:46
to large scale structures without accounting for a Weibull
34:49
statistics which mathematically model the probability of flaw
34:52
distribution across larger volumes.
34:54
If I can try an analogy here, right, relying solely on that
34:57
tiny test coupon is like test driving a sports car in an empty
35:00
the perfectly paved parking lot at 10 miles an hour.
35:03
And that assuming that same car is guaranteed to survive the
35:05
Indy 500. That is exactly what we do.
35:07
We rely heavily on digital simulations like finite elements
35:11
analysis, which often assume materials are perfectly
35:14
homogeneous continuous blocks of matter, but real materials have
35:18
grain boundaries, inclusions, and voids.
35:21
The Mechanical Engineer's Handbook breaks failures down
35:24
into four categories. Material, Design, Fabrication,
35:28
and Service. We've talked about design and
35:30
fabrication, but service failures?
35:33
Those seem like the wildest wild card of all.
35:35
Service conditions are the ultimate test because the user
35:38
takes control and users are beautifully unpredictable.
35:41
That's a nice way of saying they ruin everything.
35:43
Pretty much they change the operating conditions.
35:46
They might introduce high frequency cyclic vibration that
35:49
wasn't in your spec sheet. They subject the part to massive
35:52
thermal shock by spraying cold water on a hot engine block.
35:56
Or they expose the part to harsh chemical environments.
35:59
The chemistry aspect is terrifying.
36:01
You can design A beautifully strong part and a seemingly
36:04
harmless environment destroys it.
36:05
Yes, think about hydrogen embrittlement where hydrogen
36:08
atoms actually diffuse into the metal lattice and make it
36:11
brittle as glass or stress corrosion cracking where
36:15
exposing a part to ammonia or even just caustic chlorides like
36:19
sea salt in the air causes microscopic cracks to propagate
36:23
rapidly under normal operating stress.
36:25
And what about physical wear? The sources mentioned a really
36:28
counterintuitive fact about erosion and impact based based
36:32
on material type. This is a classic example of why
36:34
you can't rely on intuition. Imagine you are designing A
36:37
pipeline to carry a highly abrasive slurry like sand mixed
36:41
with water. OK, if you make that pipe out of
36:44
a ductile metal, the metal erodes most severely when the
36:47
sand particles impact the surface at a direct 90° angle.
36:51
The kinetic energy gouges the soft metal like a spoon scooping
36:55
ice cream. So a 90° elbow joint in a metal
36:57
pipe is the most vulnerable spot.
36:59
Makes sense. What if I would make it last
37:01
longer so I line the pipe with an incredibly hard ceramic?
37:04
If you line it with ceramic, the failure mechanics completely
37:07
float. Ceramics are incredibly hard, so
37:09
a 90° direct impact often just bounces off.
37:13
However, ceramics fail most severely under low angle impacts
37:16
around 20 to 30°. Wait, really?
37:19
Shallow angles? Yes.
37:22
At shallow angles, the abrasive particles act like microscopic
37:25
chisels, catching the microscopic grain boundaries and
37:28
flaking off the brittle surface over time.
37:30
That is fascinating. The exact geometry that saves a
37:33
metal pipe will destroy a ceramic pipe.
37:36
If you don't know the specific failure mechanics of your
37:38
material, you're perfectly dimension part is doomed.
37:42
Which brings up the question of catching these flaws.
37:44
We use non destructive testing or NDT like X-rays or
37:48
ultrasound. Can't we just mandate that the
37:50
part has 0 flaws? You can mandate it, but you'll
37:52
never build anything. NDT standards acknowledge
37:55
physical reality. Look at the ASME Boiler and
37:58
Pressure Vessel code, which governs some of the most
38:00
critical dangerous equipment on Earth.
38:02
It doesn't demand perfection, it outlines very specific
38:06
acceptance criteria for flaws. For example, what's acceptable?
38:10
In a metal casting, linear indications which look like tiny
38:13
cracks on an X-ray are entirely acceptable and legal to leave in
38:17
the part if they are smaller than 0.19 inches long in a
38:21
section that is half an inch thick.
38:23
Really, just leave the crack in there.
38:25
Yes, if the section is thicker up to an inch, the acceptable
38:28
flaw size actually grows to 0.38 inches.
38:32
So the safety code acknowledges that flaws will exist, and it is
38:36
mathematically OK with that. It just demands that the flaws
38:39
be small enough not to propagate into a catastrophic failure
38:43
during the designed lifespan. Exactly.
38:45
This is why we use safety factors in our math.
38:48
A safety factor isn't just a random multiplier, it is a
38:51
mathematical buffer. For our own ignorance.
38:53
We over dimension parts precisely because we cannot
38:56
perfectly predict the service environment, the exact size of
38:59
the hidden flaws or the unpredictability of the user.
39:02
OK, this has been an incredible journey through the gauntlet to
39:05
survive all this. The abstract requirements from
39:08
the sales team, the subjective CMM software passing bad parts,
39:12
the drill bit wobbling off center, the size effect breaking
39:16
our ceramics and the sea salt corroding our steel.
39:19
An engineer needs a very specific mental toolkit.
39:22
They absolutely do. Knowing calculus is not enough,
39:26
how do you actually survive this chaos?
39:28
Let's talk about the engineering habits of mind.
39:31
This is a crucial pivot. The Royal Academy of Engineering
39:34
recognize that technical skills alone do not prevent these real
39:37
world failures. They commissioned a major study
39:40
to figure out how great veteran engineers actually think and
39:44
operate differently than novices.
39:46
They identified 6 core engineering habits of mind or E
39:49
hom. Let's walk through these because
39:51
I want the listeners to evaluate their own mindset against this
39:53
list. What is the first habit?
39:55
The first is systems thinking. This is the ability to zoom out
39:59
and see how every isolated component connects.
40:02
It's the realization that the microscopic wobble of the drill
40:05
press affects the tolerance stack up.
40:08
Which dictates the gap in the assembly, which limits the
40:11
allowable wear in the field, which ultimately dictates the
40:14
maintenance schedule the customer has to follow.
40:16
You cannot design a component in a vacuum.
40:19
Everything is a domino effect. You have to see the forest and
40:22
the tree simultaneously. Yeah.
40:25
What is the second habit? Problem finding.
40:28
This is arguably the most important habit related to our
40:31
earlier discussion on moving targets and requirements.
40:34
Problem finding isn't just taking the client's spec sheet
40:37
and solving the math, it's digging deeper to figure out
40:40
what they actually need, which is often completely different
40:43
from what they asked for. Give me an example of problem
40:45
finding in action. A customer comes to you and says
40:48
our mounting brackets keep snapping.
40:50
We need you to design A bracket that is twice as thick and made
40:53
of titanium. So you just build a thick
40:55
titanium bracket. A novice engineer does, but the
40:58
problem finding engineer goes to the field, looks at the machine
41:02
and asks why it's snapping. They realized the true issue is
41:06
a high frequency harmonic vibration originating from the
41:09
motor. The solution isn't a stiffer
41:11
bracket, the solution is a rubber dampener to isolate the
41:14
vibration I. Love that.
41:16
Don't just build a faster horse. Invent the car.
41:19
Look past the stated symptom habit #3.
41:22
Visualizing this is the ability to move fluidly from the
41:25
abstract math to concrete physical space.
41:28
It's the mental rehearsal of reality.
41:31
This makes me think of the classic mechanics complaint.
41:33
This was designed by an engineer who has never held a wrench.
41:36
Exactly. The visualizing engineer rotates
41:39
the CAD model in their head, imagines the human assembly
41:42
worker standing next to it, and realizes, wait, there's only an
41:45
inch of clearance here, you can't fit a standard socket
41:48
wrench in there. I need to move this bolt.
41:50
It prevents downstream nightmares.
41:52
Habit 4. Improving this is the
41:55
relentless, almost obsessive drive to make things better by
41:58
tinkering, prototyping and conjecturing.
42:01
It's the absolute refusal to accept the first workable
42:04
solution as the final solution. And habit 5.
42:07
Creative problem solving. The report heavily stresses that
42:11
engineering is a team sport. It requires bringing in people
42:14
from different disciplines, electrical manufacturing
42:17
quality, generating ideas together and engaging in
42:21
generous but rigorous critiquing of each other's work without
42:24
taking it personally. Which is hard for a lot of us.
42:27
And the final habit. Adapting this means testing,
42:30
analyzing the failure, reflecting and changing course,
42:34
and the study makes a vital distinction here between
42:36
technical change, where you just apply known skills in a new way,
42:39
and adaptive change. What's the difference?
42:42
Adaptive change requires emotional intelligence.
42:44
It requires the humility to abandon a failing path that
42:48
you've poured hundreds of hours into, admit you were wrong, and
42:51
try a completely new route. I have to point out a massive
42:54
contradiction here between how we are trained and how we are
42:57
expected to work. School actively penalizes
43:00
failure. If you design A bridge on a
43:03
midterm and it fails, you get an F.
43:05
But these improving and adapting habits of mind require failing
43:09
fast prototyping, and making mistakes to learn.
43:13
Are we accidentally training engineers to be terrified of the
43:16
exact trial and error process that actually leads to
43:19
innovation? Many industry leaders argue that
43:22
we absolutely are. The report notes that
43:24
traditional university curriculums rely heavily on
43:27
analytical theoretical math, with clear right and wrong
43:30
answers. It encourages A rigid, fearful
43:33
approach. So that's the fix.
43:35
To cultivate these real world habits, we need more project
43:38
based learning where students are forced to deal with messy
43:41
open-ended problems, brutal budget constraints, and the
43:44
friction of dynamic teamwork. Because the shop floor requires
43:48
you to find the least wrong answer that fits the budget and
43:51
timeline, not the perfect answer at the back of the book.
43:54
That is the essence of good engineering practices.
43:56
Look at how international safety standards treat this.
43:59
ISO 13849-1 is a standard for designing safety related control
44:04
systems. It explicitly relies on the
44:06
engineer utilizing these habits of mind.
44:08
How so? It states that you can estimate
44:11
a mean time to dangerous failure of 150 years for a hydraulic
44:15
component, which is incredible reliability, but only if the
44:19
designer applies what they call systematic avoidance measures.
44:23
Meaning the designer has to proactively anticipate the
44:26
chaos. What kind of measures?
44:28
Things like over dimensioning critical components by a factor
44:31
of at least 1.5, ensuring complete chemical compatibility
44:35
to prevent unexpected corrosion, and designing robustly against
44:39
electromagnetic interference. The International Safety
44:42
Standard literally assigns A numerical reliability value
44:46
based on the assumption that you, the engineer, utilize
44:49
systems thinking to foresee and mitigate unknown risks.
44:53
OK, We have covered an incredible amount of ground
44:55
today. We've exposed the massive gaps
44:57
in requirements, we've broken down the friction between global
45:00
GD and T standards, we look at the physical realities of
45:03
manufacturing and materials, and we've highlighted the mental
45:06
toolkit needed to survive. Let's bring this home by
45:09
answering the most common questions you, our listeners,
45:11
are asking based on these exact pain points.
45:14
I am ready. Fire away.
45:15
Question one why did my perfectly dimensioned, perfectly
45:19
inspected part still fail in service?
45:22
This is the classic heartbreak every engineer experiences at
45:25
least once. It is a rite of passage, and the
45:29
answer almost always traces back to the fundamental difference
45:32
between a static model and a dynamic reality.
45:35
A blueprint or a CAD model is a static snapshot.
45:39
Service Life is a highly dynamic movie.
45:42
That's a great way to put it. Your part likely encountered A
45:44
variable you simply didn't account for.
45:46
In your worst case assumptions. Perhaps the user subjected it to
45:49
high frequency vibration causing fatigue over millions of cycles,
45:53
or incredibly commonly a micro environmental factor was
45:57
ignored. Like the chemistry we talked.
45:58
About exactly. For example, your stainless
46:00
steel bracket was perfectly strong for the mechanical load,
46:03
but it was operating in a chloride rich atmosphere.
46:06
Maybe the factories near the ocean which led to stress
46:09
corrosion cracking. The drawing was flawless.
46:11
Your assumptions about the reality it lived in were flawed.
46:14
That highlights the problem finding habits so well.
46:17
You have to anticipate the environment, not just the load.
46:20
OK, question 2. What is the single biggest
46:23
mistake engineers make when sending designs to overseas
46:27
manufacturers? Assuming that engineering
46:29
symbology is a universal, globally understood language, as
46:33
we cover deeply with the clash between ASME and ISO standards,
46:37
if you use North American rules but the overseas shop operates
46:41
on international rules, the foundational definitions of size
46:44
and form will conflict. Rule #1 versus independency.
46:47
Exactly. The biggest mistake is failing
46:50
to explicitly state the governing standard in the title
46:53
block of the drawing. If you don't boldly declare
46:55
whether rule number one or the principle of independency
46:57
applies, you are essentially leaving the contract open to
47:00
interpretation. And when the parts show up and
47:03
don't assemble, you are legally and financially liable for the
47:06
mistranslation. It's the Spanish dictionary for
47:08
the Italian book. You have to explicitly tell them
47:10
what language you are speaking before they start reading.
47:14
All right, final question question 3.
47:16
How do I stop moving targets and changing requirements from
47:20
ruining my project deadlines without making the customer
47:24
furious? This requires a massive adaptive
47:27
mindset shift. You have to stop treating
47:30
changing requirements as a failure of the customer or a
47:32
breach of contract and start treating them as a natural,
47:36
inevitable law of the engineering environment.
47:39
Don't fight the chaos. Do not fight the chaos.
47:41
Engineer a process that can absorb it.
47:43
Build iterative loops and prototypes into your project
47:46
timeline. Use the problem finding habit
47:48
early on. Sit down with the client and dig
47:50
past their stated wants down to their actual functional needs.
47:54
Get to the root. Establish what they functionally
47:56
need the system to achieve rather than just locking in
47:59
their initial guess at a solution.
48:01
The inevitable late stage changes will be minor course
48:04
corrections rather than total project derailments.
48:07
That is incredibly practical advice.
48:08
Flexibility is not a failure. Flexibility is a design
48:12
requirement. Absolutely.
48:13
You cannot engineer the human element out of the process, so
48:16
you must account for it. We spend so much time, energy,
48:20
and money trying to engineer the mathematically perfect part.
48:23
But given everything we've talked about today, from the
48:26
unpredictable fails team to the microscopic drill wobble, the
48:29
subjective CMM software, the size effect in ceramics, and the
48:33
reality of stress corrosion cranking, what if the ultimate
48:36
engineering skill isn't designing for perfection, but
48:39
designing for forgiveness? Think about how you can make
48:42
your next design, your next drawing, or your next customer
48:45
interaction more resilient to the inevitable messiness of
48:48
reality. If you've got something useful
48:50
out of today's eisode, hit follow and drop a like.
48:52
It helps grow the show and keeps the engineering knowledge
48:54
flowing.