Latest / Mechanical Engineering Made Simple / (#93) Blast from the Past - Simulating Reality Before Cutting Metal

Transcript

(#93) Blast from the Past - Simulating Reality Before Cutting Metal · Mechanical Engineering Made Simple

0:00
Welcome back to mechanical engineering made simple.
0:02
It's great to be back and today we're doing something a little
0:06
different. We're sort of stepping away from
0:08
the shop floor for a bit. We really are.
0:11
You know, when people think about mechanical engineering,
0:13
the image that usually comes to mind is very physical, very
0:17
tangible. Oh, absolutely.
0:19
You think of a wrench, a lathe, maybe a big schematic rolled out
0:22
on a drafting table. Exactly.
0:24
The smell of machine oil and metal filings.
0:27
It's all about the hardware, the stuff you can touch and hold.
0:30
But today, we're putting all that aside.
0:32
We are. We're moving beyond the drafting
0:34
board, beyond the wrench, and into the digital realm.
0:38
We're doing a deep dive into the digital revolution that
0:41
completely, and I mean completely, changed how things
0:44
get made. This isn't just about new tools,
0:46
it's about a new way of thinking it.
0:48
Really is. We're looking at the whole
0:50
ecosystem, from the philosophy of how teams even work together,
0:54
all the way to these incredible virtual worlds where products
0:58
get tested for a single atom of them exists in reality.
1:02
It sounds like science fiction, but it's the day-to-day reality
1:05
for engineers now. And to guide us, we're analyzing
1:08
a specific stack of research that really breaks this down
1:11
into four key pillars. Right, the first one is the
1:15
philosophy itself. It's called even Current
1:17
engineering or CE. It's basically the story of how
1:20
engineers stopped working in isolated silos that.
1:22
Famous over the wall approach. We'll get into that.
1:25
Then we have the technologies that made it all possible, the
1:27
digital glue, if you will. Things we take for granted now,
1:31
like e-mail and shared databases.
1:33
But we're revolutionary at the time.
1:35
And from there we moved to the digital canvas itself, Computer
1:38
Aided Design, or CAD. How we went from drawing lines
1:42
to building intelligent 3D models.
1:44
Finally, the ultimate pay off virtual reality and
1:47
optimization. This is where we step into the
1:50
matrix, so to speak, to test and perfect these designs
1:53
mathematically. So our mission today is to
1:55
understand how engineers use this digital toolkit to save
1:58
staggering amounts of time, money, and, let's be honest,
2:01
their own sanity, all by simulating reality before they
2:04
even think about cutting a piece of metal.
2:06
It's the journey from educated guessing to calculated
2:10
perfection. It's a fascinating story.
2:12
It really is. So let's start at the beginning.
2:16
Before the fancy software, before the VR goggles.
2:19
You have to change how people think and work.
2:22
Let's talk about concurrent engineering.
2:25
The sources peg this as a phenomenon of the 1980s.
2:28
It was a huge phenomenon, and it grew out of a really serious
2:31
problem. To understand it, you have to
2:33
look at where it started, which, as is often the case with these
2:36
big technological shifts, was in the US Department of Defense.
2:40
The DoD, right? They have the complex projects
2:42
and the budget to try and fix them.
2:45
Exactly. So in the 80's the DoD is
2:48
looking at its procurement process for these incredibly
2:50
advanced defense systems. Fighter jets, submarines, you
2:53
name it, and they realize they have a fundamental systemic
2:56
flaw. Which is what?
2:58
The people designing the thing were completely disconnected
3:00
from the people who had to actually build the thing.
3:02
The classic divide the ivory tower designers versus the
3:06
factory floor crew. A chasm?
3:08
Really. You had brilliant engineers
3:10
designing these systems to be aerodynamically perfect,
3:13
theoretically flawless, but they weren't thinking at all about
3:17
what the source calls manufacturability.
3:20
Can you actually build it? Can you build it?
3:21
Can you build it on time? Can you build it for a
3:23
reasonable cost? The answer too often was no.
3:27
A design would be finalized, signed off, and then it would
3:30
hit the factory floor. And that's where we get this
3:32
term. The source uses factory shock.
3:35
It's a perfect term. Yeah.
3:36
The design lands and the manufacturing engineers look at
3:39
it and just their jaws drop. They see a part that requires 5
3:43
axis machining when a simple stamped piece would do, or a
3:46
bolt that's placed in such a way that you'd need a child's hands
3:49
and a custom built wrench to tighten it.
3:51
So they're discovering these massive shortcomings, but it's
3:54
way too late in the game. Way too late.
3:56
The tooling has been ordered, the materials are on their way.
3:59
The production line is being set up to go back and change the
4:02
design at that point isn't just expensive, it's catastrophically
4:05
expensive. It causes huge delays.
4:08
OK, so the DoD sees this is a massive drain.
4:12
This leads to a formal report from the Institute of Defense
4:15
Analysis where they actually define this new idea.
4:17
They do. They give us the first real
4:20
formal definition of concurrent engineering, and it's a bit of a
4:24
mouthful, but the keywords are right in there.
4:26
They call it a systematic approach to the integrated
4:30
concurrent design of products and their related processes.
4:33
Integrated and concurrent. Those are the magic words
4:35
through. Everything concurrent just means
4:37
at the same time. So instead of the old way, the
4:40
classical mold of engineering. Which was sequential step A,
4:44
then you finish step A, then you start step B.
4:45
Exactly. Designer finishes, throws it
4:48
over the wall to the manufacturing engineer.
4:50
Manufacturing engineer figures out how to build it, then throws
4:53
it over the wall to the quality control team.
4:55
It's a relay race. I love that over the wall
4:58
imagery, it implies you don't even see the person you're
5:00
handing off to, you just lob it over and hope for the best.
5:03
It's almost adversarial, but CE says no, we're not doing that.
5:08
The core idea is parallel working from the absolute
5:12
outset, from the first sketch on a napkin, you have everyone in
5:16
the room, everyone. Everyone.
5:18
You don't just consider design, you consider all elements of the
5:21
product's entire life cycle, quality, cost, the schedule,
5:26
user requirements. And this is the one that really
5:28
stood out to me from the sources, even disposal.
5:31
Right, I have that note highlighted too, the idea that
5:34
in CE the customer isn't just the person who buys the final
5:37
product. That's a huge mental shift.
5:39
The customer is anyone downstream who has to interact
5:42
with what you've made. The manufacturing team is your
5:45
customer. They need a design that's easy
5:47
to build. The maintenance crew is your
5:48
customer. So if you design A car engine
5:51
where you have to remove three other components just to change
5:53
an oil filter. You've failed your maintenance
5:55
customer, and that failure has real costs, means higher repair
5:59
bills, customer dissatisfaction, A damaged brand reputation, all
6:03
because the designer didn't think about the person who'd
6:05
have to service it 10 years down the line.
6:07
And that includes the disposal team, who has to figure out how
6:10
to safely and efficiently recycle it at the end of its
6:12
life. Precisely.
6:14
Are there toxic materials? Can it be easily disassembled?
6:18
These are design questions in a CE framework.
6:22
But getting people who are used to working in their own little
6:25
kingdoms to suddenly collaborate must have been challenging.
6:28
You're talking about a major culture clash.
6:30
Oh, it was a huge battle. The sources talk about several
6:33
major barriers. The first, obviously, is that
6:35
over the wall mentality. It was just how things were
6:38
done. It was comfortable.
6:39
But then there's something they call the cost paradox.
6:42
This sounds like where the accountants get nervous.
6:44
This is where management gets very nervous because on paper,
6:48
concurrent engineering looks more expensive upfront.
6:51
How so? You need bigger teams from day
6:54
one. You're having more meetings,
6:55
You're paying a manufacturing expert, a quality assurance
6:58
person, maybe even a materials disposal specialist to sit in on
7:02
a conceptual design meeting. And a manager looking at the
7:05
quarterly budget just sees salaries and meeting room costs
7:07
going up. It looks like waste.
7:09
It looks like pure inefficiency. Why are all these people here?
7:12
The designers haven't even designed anything yet.
7:15
But that's the paradox. You spend that money upfront to
7:17
save 10100 times that amount down the road.
7:21
Because you catch the untightenable bolt problem in
7:23
the meeting, not after you've built a million of them.
7:26
Exactly. You avoid those catastrophic
7:28
factory shocks. You have far fewer engineering
7:31
change orders late in the process, but you're incredibly
7:34
costly. The product gets to market
7:35
faster, which is a huge competitive advantage.
7:38
It's the ultimate measure twice, cut once.
7:41
There's another conflict, mentioned a more human one.
7:44
The clash between command style management and what they call
7:47
team empowerment. Yes, this is about power and
7:50
control. Traditional engineering was very
7:52
hierarchical. A senior engineer makes a
7:55
decision and it flows downward. CE requires a much flatter
7:59
structure. The sources describe the CE team
8:02
as a flat entity or maybe A2 level structure with a leader
8:05
and some sub teams. Right.
8:07
It's not about command and control, it's about consensus
8:10
and collaboration. Information flows horizontally
8:13
between disciplines, not just up and down the chain of command.
8:16
And that can be very threatening to old school managers.
8:19
And also to old school engineers, it seems they talk
8:22
about the clash between rigorous analysis and parallel work.
8:26
This is a deep seated cultural thing for engineers.
8:29
The classical training is all about rigorous step by step
8:32
analysis. You don't move on to the next
8:34
step until you have complete, verified, perfect data from the
8:38
previous step. You want certainty.
8:40
You crave certainty, but in CE you can't have it.
8:43
You're working in parallel. The team designing the chassis
8:46
is working at the same time as the team designing the engine
8:49
that goes in it. So you're working with tentative
8:51
data, incomplete information. You have to, the engine team
8:54
might say, OK, we think the engine block will be this size
8:57
±5%, and the chassis team has to take that incomplete data and
9:02
start their design work. That sounds terrifying for an
9:04
engineer who likes precision. It requires a massive amount of
9:08
trust. It's a high trust environment.
9:10
You have to trust that your colleague in the other
9:13
department will tell you the instant their data changes.
9:16
If the engine designer realizes the block has to be 6% bigger,
9:20
they can't sit on that information.
9:22
They have to communicate it immediately.
9:24
Because if they don't, the chassis won't fit and the whole
9:27
parallel process collapses. It breaks down completely.
9:31
The workflow is described as these long intervals of focused
9:34
parallel work, punctuated by these short, intense bursts of
9:38
communication to sync up and produce consensus.
9:41
Work sync, work sync. It's a rhythm.
9:44
It is. But for that rhythm to exist,
9:46
you need more than just good intentions.
9:48
You can't just throw people in a room and say collaborate.
9:50
You need the infrastructure. You need the technology to make
9:52
it all happen. Which is the perfect segue to
9:55
our second pillar, the infrastructure, the technologies
9:58
of collaboration. How do you get all these
10:01
different experts who might not even be in the same building
10:03
talking to each other effectively?
10:05
Well, the sources give us a bit of a fun historical tour here.
10:08
Start to the basics. Let me guess, e-mail?
10:11
E-mail. The source, written a while ago,
10:14
notes that it had already been in use for 1/4 century, but they
10:17
talk about the move from just plain text to multimedia mail
10:22
being able to embed images and data.
10:25
And they mentioned a specific product that I have not heard in
10:28
a very long time, Lotus Notes. A true blast from the past, but
10:34
you have to understand how groundbreaking that was.
10:37
It wasn't just a messaging system, it was about creating
10:40
structured shared databases for communication.
10:43
So it's not just sending a message that gets lost in an
10:46
inbox, it's contributing to a central pool of knowledge that
10:50
everyone on the project can access and search.
10:52
Precisely, it's about creating a project memory.
10:54
But then things get even more interesting when we get to what
10:57
the source calls computer supported meetings.
10:59
This is where the Xerox collab comes in.
11:00
I was fascinated by this. The collab project at Xerox
11:03
Parerox was so ahead of its time, they basically invented
11:06
the concept of the electronic chalkboard.
11:07
And they had that great acronym for it, WASIWIS.
11:10
Wasiwis, what you see is what I see.
11:12
It seems so obvious to us now. We do it every day on Zoom or
11:16
Teams with screen sharing. But the idea that multiple
11:20
people could be looking at a digital workspace, all seeing
11:23
the exact same thing in real time and all being able to
11:26
interact with it, that was a game changer.
11:29
It absolutely was. It meant the design review
11:31
meeting was no longer just people pointing at a physical
11:33
blueprint. It became a dynamic,
11:36
collaborative design session. But there were even more
11:39
advanced tools. The ones for argumentation like
11:42
ark noter and shibas. Right.
11:44
These tools recognized a fundamental truth about
11:47
engineering. It's a series of debates and
11:49
decisions. It's not just about drawing
11:51
pretty pictures, it's about arguing over trade off.
11:54
Exactly. You have what the source calls a
11:56
Tiger team. This multidisciplinary group of
11:58
experts. Material scientist says we
12:01
should use composite. The manufacturing engineer says
12:03
composite is too expensive to machine.
12:06
The structural engineer says the alternative, aluminum, isn't
12:08
strong enough. You have a conflict.
12:11
And these tools were designed to structure that argument.
12:14
Yes. Instead of it just being a
12:16
heated discussion where important points get lost, these
12:19
systems provided A framework. You formally propose an idea,
12:23
use composite. You then attach arguments for
12:26
and against it. You document the final decision
12:29
and, crucially, the rationale behind it.
12:32
The rationale that connects back to what the source calls
12:35
corporate history and memory. This is one of the single
12:38
biggest, most persistent problems in long term
12:40
engineering projects. The problem of lost rationale.
12:43
You've got a blueprint from 15 years ago, you're doing a
12:46
redesign and there's a weird bracket on it and nobody,
12:49
absolutely nobody in the building knows why it's there.
12:52
Exactly. The drawing shows you what it
12:54
is, the geometry, but it doesn't tell you why.
12:57
Was it there to solve a vibration issue that no longer
13:00
exists? Was it to hold a sensor that was
13:02
later removed? You don't know.
13:03
And because you don't know, you're afraid to remove it.
13:05
Because maybe it's critically important.
13:07
So you leave it in adding unnecessary cost and weight to
13:11
the new design, all because the original intent was lost.
13:14
That's why they started developing electronic design
13:16
notebooks, or Edns. To capture the thinking, not
13:20
just the final geometry. To capture the conversation, the
13:23
trade-offs, the why, There's a brilliant example in the text
13:27
about this. They talk about a system for
13:29
constraint management and the acid test for monitoring a
13:32
project's health. Another great acronym?
13:35
Let's breakdown ACID. It stands for Access, Choose,
13:39
Interpret, and Display. For a project leader to manage a
13:42
complex design, they need a system that lets them do 4
13:46
things. First, access all the relevant
13:49
information, past and present. OK, get the data.
13:51
Second, choose the specific parameters they want to look at.
13:54
You don't want to be flooded with everything. 3rd and this is
13:57
the key. Interpret that information to
13:59
understand what's really going on and finally display it in a
14:02
way that's intuitive and easy to grasp.
14:04
And the specific example they used to illustrate this was
14:06
about a jet engine blade, right? The strength constraint.
14:09
It's a perfect example. So imagine you're the project
14:12
leader. You pull up your project
14:13
dashboard. The system flags an alert blade
14:17
strength constraint violated. OK, that's the data, but it's
14:21
not enough. It's the what, not the why.
14:23
Exactly. A good system lets you click on
14:26
that and interpret it. You dig deeper and the system
14:29
shows you who is violating the constraint.
14:32
It's the aerodynamics engineer and it shows you that he's done
14:35
it 10 times in the last two days.
14:37
OK, now it's getting interesting.
14:39
It's not a random error, it's a pattern.
14:41
It's a pattern, and the Electronic Design Notebook tells
14:44
you why. The notes show that the
14:46
aerodynamics engineer is relentlessly trying to tweak the
14:50
blade shape to optimize the airflow path through the
14:52
turbine. He's not making a mistake, he's
14:55
pushing the boundaries of his own discipline to improve
14:57
performance, but it's putting him in conflict with the
15:00
structural requirements. Precisely.
15:02
Now, as the project leader, you have full context.
15:05
You don't just see an error, you see a design conflict between
15:08
aerodynamics and structural integrity.
15:10
You know exactly which two engineers you need to get in a
15:12
room together to find a compromise.
15:14
That is effective data-driven management, and it's impossible
15:18
without this kind of digital infrastructure.
15:20
That makes perfect sense. So we've got the philosophy of
15:23
teamwork, concurrent engineering.
15:25
We've got the digital infrastructure to let them talk.
15:28
Now we need the canvas, the digital space where the design
15:31
actually takes shape. Let's get into Computer Aided
15:33
Design CAD. Right.
15:35
And it's important to remember that CAD has been around in some
15:38
form for a very long time. The source gives a nod to the
15:41
SAGE air defense system way back in the 1950s.
15:45
Using light pens on big CRT screens.
15:47
It looks primitive now, but that was the genesis.
15:50
It was. It established the fundamental
15:52
connection that CAD is all about, linking the intuitive
15:55
creativity of the human hand with the raw processing speed
15:58
and precision of a computer. But modern CAD is so much more
16:02
than just electronic drafting. It's not just about replacing
16:05
the pencil and T square with a mouse and keyboard.
16:07
Not at all. It's about building an
16:08
intelligent mathematical model of an object.
16:12
To understand that, we have to look at how these systems
16:14
actually modeled geometry. The earliest method was the
16:17
wireframe model. The skeleton approach.
16:19
Exactly. You're just defining the edges
16:21
of an object. It's basically a 3D stick
16:23
figure. The main advantage was that it
16:26
was computationally simple. In an era of slow computers,
16:30
that mattered a lot. But it had a massive, fatal
16:33
flaw. Ambiguity.
16:34
Ambiguity. If you look at a wireframe cube
16:37
on a screen, your brain can sometimes flip it inside out.
16:41
You can't definitively tell which face is the front and
16:44
which is the back. It's like one of those optical
16:46
illusions. It is, and for a computer that's
16:49
a disaster. If it doesn't know which surface
16:51
is which, It can't calculate things like surface area,
16:54
volume, or mass. It can't tell if two parts are
16:57
interfering with each other. The model doesn't know it's a
17:00
solid object, it just thinks it's a collection of lines in
17:03
space. So we needed a better way, which
17:06
brings us to solid modeling. Solid modeling.
17:08
Now the computer understands that this is a solid object with
17:10
an inside and an outside. The source breaks this down into
17:14
two main methods for creating these solids, CSG and B Rep.
17:18
Let's start with CSG. That's constructive, solid
17:20
geometry. This is the one that feels like
17:22
playing with digital building blocks.
17:24
That's the perfect analogy. You start with a set of simple
17:27
predefined shapes called primitives.
17:30
Things like a cube, a cylinder, a sphere, a cone.
17:34
And you combine them using boolean logic.
17:36
Yes, the same AND or not logic that runs all of computers but
17:40
applied to geometry. The main operations are union,
17:43
subtraction and intersection. So if I want to make a simple
17:46
part like a block with a hole through.
17:48
It you'd start with AQ primitive, then you'd create a
17:50
cylinder primitive. You move the cylinder so it
17:53
passes through the cube where you want the hole.
17:55
Then you perform a subtraction operation, cube minus cylinder.
17:59
The computer calculates the resulting geometry and you're
18:02
left with a block with a perfect hole in it.
18:03
It's like virtual machining. You're removing material.
18:06
Exactly. And union is just adding things
18:08
together, fusing them to one part.
18:11
Intersection gives you only the volume that two objects share.
18:14
It's a very powerful and intuitive way to build many
18:18
types of mechanical parts. OK, so that's CSG.
18:21
What about the other method B Rep?
18:23
B. Rep stands for boundary
18:25
representation. Instead of combining primitives,
18:28
this method defines an object by explicitly describing its skin
18:32
or its boundary. So you're defining all the
18:34
faces, edges and vertices. Correct.
18:37
You're essentially creating a mathematical description of the
18:39
object's surface. It's a more complex data
18:42
structure, but it's also more flexible, can represent much
18:45
more complex freeform shapes like the body of a car or the
18:49
wing of an airplane that would be very difficult to build out
18:51
of simple cubes and cylinders. So most modern CAD systems
18:55
probably use a combination of both.
18:57
Yes, many are hybrid systems. You might use CSG like
19:01
operations to create the basic form and then use B Rep tools to
19:04
add complex features like flays and chamfers.
19:07
Now the source gets into another important distinction,
19:09
parametric versus variational modeling.
19:12
This feels like the step where the model gets smart.
19:15
It is absolutely the step where the model gains intelligence.
19:18
This is a huge leap beyond just being a static 3D shape.
19:21
Let's talk about parametric modeling first.
19:23
It's all about defining relationships and constraints
19:26
between different parts of the geometry.
19:28
So it's not just a collection of lines and faces, it's a system
19:31
of rules. A system of equations, really.
19:34
Think back to our block with a hole in a dumb solid model.
19:39
If you make the block wider, the hole just stays where it was.
19:42
It might not be in the center anymore.
19:44
Right, you'd have to manually move the hole to recenter it.
19:46
In a parametric system, you don't just place the hole, you
19:50
define a rule. The center of this hole must
19:52
always be at the geometric center of this block's top face.
19:56
That rule is stored as part of the model.
19:59
So now if I make the block wider.
20:01
The computer automatically resolves the equations.
20:04
It finds the new center of the face and moves the hole to match
20:07
the design intent. The hole is in the center is
20:09
preserved. That's incredibly powerful.
20:11
It saves so much rework. It's fundamental to modern
20:14
design. This concept is taken even
20:17
further with what's called Feature based modeling.
20:19
This is another layer of intelligence.
20:21
This is where we start thinking in terms of primitive shapes
20:24
like cylinders and start thinking in terms of engineering
20:27
features like holes. Exactly.
20:29
You don't tell the computer to subtract a cylinder.
20:31
You use a hole command and you can specify what kind of hole.
20:35
Is it a through hole? A blind hole?
20:37
Is a countersunk? Is it threaded?
20:40
And the computer understands the engineering meaning of those
20:42
terms. Yes.
20:43
So if you place a through hole in a part, and you later make
20:46
that part thicker, the hole automatically gets deeper to
20:50
maintain its through nature. The model captures your design
20:53
intent, not just the resulting geometry.
20:56
It knows you wanted a hole that goes all the way through, no
20:59
matter the thickness. OK, this is amazing stuff.
21:02
But underneath all this intuitive feature based
21:04
modeling, the computer's just crunching numbers.
21:07
The source touches on the math behind the graphics, how the
21:10
computer actually moves things around using matrix algebra.
21:13
It's worth a quick look because it's so fundamental.
21:15
Every single point or vertex in your 3D model is stored in the
21:20
computer's memory as a set of coordinates, an XY and Z value.
21:25
So a simple cube might just be 8 sets of coordinates.
21:28
Right. And when you, the user, want to
21:30
move that cube on the screen, the computer performs a matrix
21:33
operation on all eight of those coordinate sets simultaneously.
21:37
So let's take the simplest one, translation, just moving it.
21:40
That's basically just matrix addition.
21:43
You define a translation matrix which the source writes as T
21:46
equals the matrix containing delta X and delta Y.
21:49
In 3D it would have a delta Z as well and you just add that delta
21:53
that change to the original coordinates of every point.
21:56
New position equals old position plus the change.
21:58
Simple enough. What about scaling?
22:01
Scaling is matrix multiplication.
22:02
You define a scaling matrix S and you multiply the coordinate
22:05
matrix by S The source shows this as I prime equals I * s If
22:10
your scaling factor is 2, all the coordinates get multiplied
22:13
by two and the object double S in size.
22:15
Then rotation is the most complex 1 I assume.
22:17
It is. The rotation matrix involves
22:20
trigonometry, sines and cosines to calculate the new coordinates
22:23
as the object pivots around an axis.
22:27
And the amazing part is that when you're smoothly spinning a
22:29
complex model on your screen with your mouse, the computer is
22:32
performing millions of these matrix multiplications every
22:35
second. It's incredible to think about
22:37
the horsepower required, but all this complexity creates a
22:41
problem. If I'm using brand X CAD
22:43
software and you're using brand Y, how do we share these complex
22:47
intelligent models? The interoperability.
22:49
Problem It's a massive, persistent headache in the
22:52
industry. For a long time, it was a huge
22:54
barrier. You were locked into one
22:55
vendor's ecosystem. So standards had to be developed
22:59
neutral file formats. Exactly.
23:01
The source mentions a couple of key ones.
23:03
The first big one was IG ES, which stands for the Initial
23:06
Graphics Exchange Specification. This was an early ANSI standard
23:10
in the US. And it was a good first step
23:12
that had its problems right. It did.
23:14
IG ES was pretty good at translating basic geometry
23:16
surfaces and wireframes, but often struggled with the
23:19
complexities of solid models and could lose some of the
23:22
intelligence. It could be ambiguous.
23:23
Which led to the development of step.
23:25
Right step. The standard for the exchange of
23:28
product model data. This is the big international
23:31
standard and ISO standard. It was designed from the ground
23:34
up to be much more robust. It's not just about transferring
23:37
the shape. It's designed to transfer the
23:39
full product model, including the parametric intelligence,
23:43
feature information and other metadata.
23:45
It's the lingua franca of modern CAD.
23:48
OK, so using all this tech, we've designed our part, we've
23:52
collaborated on it with the whole team, it exists as a
23:55
perfect intelligent digital model.
23:57
But now we want to know does it work in the real world?
24:01
But we're not ready to build it yet.
24:03
Now we enter the simulation. We step inside the model.
24:07
This brings us to our fourth pillar, virtual reality and
24:10
engineering. VR The source has a great simple
24:12
definition. It calls it an oxymoron.
24:15
Reality that does not exist. It's a good way to think about
24:17
it, but for an engineer it's really defined by two key
24:20
characteristics, immersion and interactivity.
24:23
Immersion is the feeling that you are in the digital world,
24:25
not just looking at it on a screen.
24:27
And interactivity means that the world responds to you.
24:31
When you move, your viewpoint changes.
24:33
When you reach out and touch something, it reacts.
24:37
Those two things together create the sense of presence that
24:40
defines VR. And to make that happen, you
24:43
need some pretty specialized hardware.
24:45
The source calls them the toys. Let's talk about the input
24:48
devices first. Right.
24:49
How do you get your actions into the virtual world?
24:52
The source gives a great nod to the Nintendo Power Glove, which
24:56
many of our listeners might remember.
24:57
For the home enthusiasts. For the home enthusiasts, but in
25:00
a professional engineering lab, you'd be using something much
25:03
more sophisticated like the Cyber Glove.
25:05
These are gloves lined with fiber optic cables or electrical
25:08
sensors that can precisely track the bend of every joint in your
25:12
fingers. So your virtual hand can
25:14
actually make a fist or point. Exactly.
25:16
Then you have tracking devices like those from Ascension.
25:19
They use magnetic or optical systems to track the position of
25:22
your head and hands in 3D space. And they mentioned Spaceballs
25:26
and six DOF mice. Six DOF 6° of freedom.
25:30
That's right, it means it can track not just position left,
25:34
right, up, down, forward, back, but also orientation, pitch, Yaw
25:39
and roll. It gives you complete freedom of
25:41
movement in a 3D space. OK, so that's how we get
25:44
information in. What about getting it out?
25:47
How do we see and feel this virtual world?
25:50
The most famous output devices, the HMD, the head mounted
25:53
display, big goggles that block out the real world and put
25:56
screens right in front of your eyes.
25:58
That gives you total immersion, but the source notes they often
26:01
had drawbacks like being heavy or having low resolution,
26:04
especially the early ones. They did.
26:06
An alternative for desktop use is stereo glasses.
26:10
These are shutter glasses that flicker with each lens opening
26:13
and closing in perfect sync with the monitor, which is displaying
26:16
alternating images for your left and right eye.
26:18
It tricks your brain into seeing a true 3D image popping out of
26:21
the screen. But the output device that just
26:23
blew my mind was Haptic interface.
26:25
They mentioned the phantom device specifically.
26:27
Haptics is the science of touch, and this is where VR gets really
26:30
wild. The Phantom was an early robotic
26:33
arm that you would hold and it would provide force feedback.
26:36
O you can feel the virtual objects.
26:38
You can feel them. If you're moving your virtual
26:40
hand and you push it against a virtual wall, the robotic arm
26:44
will physically push back against your real hand.
26:47
You can feel the shape, the texture, the resistance of an
26:50
object that doesn't actually exist.
26:52
The applications for that in engineering are just staggering.
26:56
You could test the ergonomics of a new dashboard design by
26:59
actually feeling how far the reaches to the radio knob.
27:03
Or you could simulate a complex assembly task.
27:05
Can a mechanic actually fit their hand and a wrench into
27:08
this tight space to remove a part?
27:11
With haptics, you don't have to guess, you can feel it.
27:14
So we have all this hardware and the source says it gets
27:16
configured into a few different types of VR systems.
27:19
Broadly, yes, you have the fully immersive systems.
27:22
The classic example is the San BE, which is a room where the
27:25
walls and sometimes the floor and ceiling are rear projection
27:28
screens. Stand in the middle, put on the
27:30
glasses, and you are completely surrounded by the virtual world.
27:33
Great for architectural walkthroughs or driver training
27:35
simulations. Then there's the other end of
27:37
the spectrum, desktop VR. Right, they call it a window on
27:40
a world is much cheaper. You're using a standard computer
27:43
monitor but with stereo glasses to get the 3D effect.
27:46
It's less immersive, but perfect for things like CAD model
27:50
inspection. And there are hybrids in
27:52
between, like a virtual workbench where a 3D image is
27:55
projected down onto a table. So why do all this?
27:58
Why invest in this expensive complex technology?
28:01
Is it just a cool gimmick? I'm guessing the answer is
28:04
money. The answer is always money.
28:07
It all comes back to a statistic mentioned in the source that is
28:10
just eye watering. 60 to 70% of a product's total life cycle
28:14
cost is committed and locked in as soon as the initial concept
28:18
design is finalized. Wow.
28:20
Let that sink in. By the time you have the basic
28:23
shape and concept, most of the money has in effect already been
28:27
spent. Exactly because that concept
28:30
dictates the materials you'll use, the manufacturing processes
28:33
you'll need, the complexity of assembly, everything.
28:36
So if you make a mistake make at that early stage, it's an
28:38
incredibly expensive mistake. And virtual prototyping is about
28:41
catching those mistakes before you ever cut a single piece of
28:44
metal. That's the whole game.
28:46
The source gives an example of a virtual lathe.
28:49
You can put your digital part in a digital machine tool, program
28:52
a digital cutting tool and watch the simulation.
28:56
You can see is the tool will chatter if it will collide with
28:59
the part holder, if the surface finish will be acceptable, all
29:03
without wasting any material or machine time.
29:05
They also mentioned a system called Coverts, which sounds
29:08
like something out of Star Trek. The conceptual virtual design
29:11
system. This was a research system where
29:14
the designer could use voice commands and hand gestures
29:16
inside a VR environment to create and manipulate shapes.
29:20
Literally saying computer make that cylinder longer while
29:24
stretching it with your hands. So it's about making the design
29:27
process more intuitive and catching those fundamental
29:29
conceptual errors right at the very beginning where it's
29:32
cheapest to fix them. That's the billion dollar idea
29:34
right there. OK, so we simulated the world to
29:37
see if our design is usable and buildable.
29:40
But is it the best possible design?
29:42
Is it the cheapest, the strongest, the most efficient?
29:46
For that, we need to go beyond just simulation.
29:48
We need to get into the brains of the operation.
29:50
This is our final segment and it's where the heavy math lives.
29:54
Design optimization. What is the core concept here?
29:57
What are we trying to do with optimization?
30:00
In short, optimization is a mathematical way to find the
30:03
best solution from among an infinite number of possibilities
30:07
without having to test them all. Because testing them all would
30:10
be impossible. It would literally take longer
30:12
than the age of the universe for any non trivial problem.
30:16
So to do optimization you need three things.
30:19
A mathematical model of your system, a performance criteria
30:23
you want to maximize or minimize, and a set of
30:26
independent variables that you can change.
30:27
And the source gives us 2 fantastic, really concrete real
30:31
world examples to show how this works.
30:33
I want to walk through them. First up, the oxygen supply
30:36
system for a steel mill. A great classic systems
30:40
engineering problem. You have a basic oxygen furnace
30:42
in a steel plant and it needs huge blasts of pure oxygen at
30:46
specific times. The question is how do you
30:48
design the most cost effective system to supply that oxygen?
30:51
And the system has three main components.
30:53
Right. You have an oxygen plant that
30:55
actually produces the gas from the air.
30:57
You have a big storage tank to hold it and you have a
31:00
compressor to pressurize the gas so you can fit a lot into the
31:03
tank. And right away you can see the
31:04
trade-offs. It's all trade-offs.
31:07
You could build a gigantic oxygen plant that can produce
31:10
gas as fast as the furnace needs it.
31:13
That way you wouldn't need a big storage tank, but a huge plant
31:16
has a massive upfront capital cost.
31:19
OK, option 2, you could build a much smaller cheaper plant that
31:23
just runs constantly 204 seven slowly producing oxygen.
31:27
But if you do that, you need a massive storage tank to act as a
31:30
buffer to hold all that gas until the furnace needs a big
31:33
blast of it. And a huge high pressure tank is
31:36
also very expensive. And to get all that gas into the
31:38
tank, you need a really powerful compressor, which is not only
31:41
expensive to buy, but also cost a fortune and electricity to
31:45
run. Exactly.
31:46
Every decision you make affects the cost of the other
31:49
components. So the engineers defined the key
31:52
variables they could play with the production rate of the
31:55
plant, which we can call F, the capacity of the tank V, the
31:58
capacity of the compressor H, and the maximum storage pressure
32:02
P. And then they wrote a giant
32:04
equation for the total cost. The cost function.
32:06
This is the heart of the optimization.
32:08
The source describes the annual cost equation and it's basically
32:11
just adding up all the different costs.
32:13
If you have the plant cost, which is some fixed cost a 1 + a
32:17
variable cost, that deends on the roduction rate a two times.
32:20
FOA one plus a 2 * F. Then you add the tank cost.
32:25
The source notes that the cost of pressure vessels often scales
32:28
with volume raised to some power, so they model it as B 1 *
32:32
V raised to the power of B2. OK, so a term for the plant, a
32:36
term for the tank. Then a term for the compressor
32:38
which is similar B 3 * H raised to the power of B4.
32:41
And finally you have to add the operating cost, the electricity
32:44
to run that compressor. That's a term like n * B five
32:47
times T 1 * H. So you have this big long
32:50
equation that adds up the cost of the plant, the tank, the
32:54
compressor and the power, and the goal is to find the specific
32:57
values for FVH and P that make the result of that equation as
33:02
small as possible. That's it.
33:03
You feed that equation and the variables into an optimization
33:07
algorithm, and the computer crunches the numbers.
33:10
It will methodically test different combinations, always
33:13
moving in the direction of lower cost, until it finds the sweet
33:16
spot, the optimal ballast between all those competing
33:19
factors. It's amazing.
33:20
It turns a complex design problem into a solvable math
33:23
problem. The second example is the welded
33:26
beam, which feels much more like a classic mechanical design
33:29
scenario. It is.
33:30
This is a problem every mechanical engineering student
33:33
is probably seen in some form. You have a steel beam that's
33:36
welded to a rigid support and it needs to hold a 6000 LB load at
33:40
its end. And the goal is simple, minimize
33:42
the cost. Minimize the total cost, which
33:44
is the cost of the material used plus the cost of the labor to
33:47
Weld it. So what are the variables we can
33:49
change here? There are 4 design variables,
33:52
the thickness of the Weld H, the length of the Weld L, the
33:56
thickness of the beam itself T, and the width of the beam B.
34:01
And just like before, there's a cost function.
34:03
Yep. The source lays it out.
34:05
F of X equals well. There are two terms for the
34:08
Weld. The first is H ^2 * L which is
34:11
related to the cost of the Weld material.
34:13
Then you add C 3 * H ^2 * L which is the labor cost with C3
34:18
being a cost coefficient. OK, so material and labor for
34:21
the Weld? And then you add the cost of the
34:23
beam itself, which is C 4 * T * B multiplied by the sum of L + L
34:28
That just represents the cost of the steel bar stock.
34:30
But wait, if that's the equation, the cheapest solution
34:33
would be to make HLT and B all equal to 0.
34:37
Just use no material. But you can't, because you have
34:40
rules you're not allowed to break.
34:42
These are the constraints. This is the most important part
34:44
of real world optimization. The design has to actually work,
34:47
it can't break. It has to be safe, so the
34:50
engineers add several constrained equations. 1st, the
34:53
shear stress in the Weld, which we'll call Tau, cannot be
34:56
greater than the maximum allowable design shear stress
35:00
for that material cod. So the Weld can't RIP apart.
35:03
2nd, the bending stress in the beam Sigma cannot exceed the
35:07
maximum allowable normal stress Sigma.
35:10
The beam itself can't snap. Third, the beam can't be so long
35:13
and skinny that it buckles into the load, so the buckling load
35:17
PC has to be greater than the applied load.
35:19
So the optimization algorithm has to find the cheapest
35:22
combination of HLT and B that satisfies all of those safety
35:27
constraints simultaneously. Precisely.
35:29
And the source notes that calculating those stresses,
35:31
particularly the shear stress in the well the Tau of X, is not
35:34
trivial. It's a combination of a primary
35:36
shear stress Tau prime and the secondary torsional stress Tau
35:39
double prime. Some fairly heavy math.
35:41
But the beauty is the engineer doesn't have to solve it by hand
35:44
1000 times. The computer does the hard work.
35:48
It explores the design space, constantly checking.
35:51
Is this design cheaper? Yes.
35:53
Is it safe? No.
35:54
OK Backup. Is this design safe?
35:56
Yes. Is it cheap?
35:58
Not really. Let's try to make it a little
36:00
thinner. Until it finds the design that
36:02
is right on the edge of the allowable limits, the absolute
36:06
cheapest, lightest beam that will do the job safely.
36:09
That's the power of optimization.
36:11
But what if you have goals that are in direct conflict with each
36:14
other? The source brings up a final
36:16
concept here, goal programming. This is a really interesting and
36:20
practical area. It's first situations where you
36:23
have multiple competing objectives.
36:24
The classic example they give is from machining.
36:26
You want to maximize your metal removal rate.
36:29
You want to cut the part as fast as possible to make more parts
36:32
per hour. But you also want to maximize
36:34
your tool life. You don't want to be constantly
36:36
stopping the machine to replace a worn out drill bit because
36:39
that cost time and money too. And those two goals are in
36:42
direct opposition. The faster you cut, the hotter
36:45
the tool gets and the faster it wears out.
36:48
Cannot have the absolute maximum of both at the same time.
36:51
It's impossible. So goal programming changes the
36:53
question. Instead of trying to maximize
36:55
both, you take goals. See my goal is a metal removal
36:59
rate of X and my goal is a tool life of Y.
37:03
And then you try to get as close as possible to those goals.
37:05
Exactly. The mathematics is set up to
37:07
minimize the deviations, the underachievements or over
37:11
achievements from those goals. These are often written as D
37:15
minus and D plus. You're trying to find the
37:17
solution that has the smallest total disappointment.
37:20
I like that you're mathematically minimizing the
37:22
engineer's disappointment. A very elegant way to solve a
37:26
very real problem. It is.
37:27
It's a pragmatic way to handle the messy reality of conflicting
37:30
requirements. So if we take a step back and
37:32
look at the whole journey we've been on today, it's really quite
37:36
incredible. We started with the old world,
37:38
the over the wall philosophy where everyone worked in
37:41
isolation. Then we saw the rise of a new
37:43
philosophy, concurrent engineering, focused on teamwork
37:46
and a life cycle view. We looked at the digital glue
37:49
that held those teams together, the communication and
37:52
coordination technologies. We moved to the digital canvas
37:56
of CAD, where we learned to build intelligent models instead
37:59
of justice. Dumb drawings.
38:01
From there we stepped inside those models with virtual
38:04
reality to test them in a human centric way and finally we use
38:09
the pure logic of design optimization to mathematically
38:12
perfect them. It paints a clear picture.
38:15
Modern engineering is not just about physics and materials
38:17
anymore. It's about information
38:19
management, it's about simulation, and it's about
38:21
applied mathematical logic. The wrench and the hammer are
38:24
still essential, but they're now guided by this immense mountain
38:28
of data and computation. The physical act of making
38:31
something is now the final step in a very long, very deep
38:35
digital process. It is.
38:36
And that brings me to a final provocative thought for you to
38:39
take with you. We've talked all about this
38:41
incredible power of simulation, virtual prototyping,
38:44
mathematical optimization, having a perfect digital memory
38:48
of every decision. But are we in danger of
38:51
optimizing away the happy accidents?
38:54
That's a really interesting question.
38:56
What do you mean? Well, so many great inventions
38:58
in history came from mistakes from serendipity.
39:02
Someone tried something that shouldn't have worked, or mixed
39:04
the wrong things together and discovered something amazing
39:07
like penicillin or post it notes.
39:10
If we only ever build what the computer has already confirmed
39:13
is the most efficient or optimal solution.
39:16
Do we lose the opportunity for those unexpected breakthroughs?
39:20
Exactly. If we only ever cut the metal
39:23
that the simulation says is perfect, do we lose the chance
39:27
to discover something new from a design that looks wrong on paper
39:30
but has some unforeseen benefit in the messy, unpredictable real
39:35
world? Are we designing ourselves into
39:37
a perfectly efficient box and losing the random spark of true
39:41
invention? It's the paradox of perfection.
39:44
If you eliminate all risk and error, you might also eliminate
39:46
all surprise. Something to think about?
39:49
Well, that brings our deep dive into the digital transformation
39:51
of engineering to a close. This was a fantastic journey.
39:55
It was a pleasure, as always. And remember, if you haven't
39:58
already, get over to our YouTube channel Mechanical Engineering
40:01
Made Simple, where we post hands on how to videos, visual
40:05
breakdowns, and practical engineering walkthroughs.
40:07
You can actually use a family that solves together, stays
40:10
together.