Transcript: Lamp stack of manufacturing - Make:HIW
1. +Intro:
"Hi everyone, I'm Nick Pinkston. I was the founder of a 3D printing and
injection molding startup called CloudFab that was actually just acquired,
and previously I co-founded a community-workshop called HackPittsburgh.
They asked me to come out today to take off where the previous speakers
left off - to show a vision of what the future of manufacturing might
actually look like."
"So to start off, I want to talk about an industry that, even after the
industrial revolution ran its course, worked at a painfully slow pace. It
used 100% manual labor with just simple hand tools, but in just the past 50
years the industry went from nearly 100% manual to nearly 100% automated
with engineers directly designing and maintaining the means of production.
So what is this industry? Well, I'm being a little coy because it's
actually the computation industry.
We went from slide rules to laptops in 50 years, because computer
scientists discovered certain techniques that put the engineers in the
driver seat when the machines started taking care of themselves. I'm here
today to show you that this isn't limited to the production of processed
data like many imagine, but that nearly all these techniques can be brought
into the realm of manufacturing to fully empower hardware companies from
startups to multinationals.
+Intro Alt
"So this is the Jacquard loom, and "
+Mind the Gap:
"We have a gap in production today - on the one end you've got mass
producers like Toyota that use very advanced automation to make one part
millions of times super cheap, and on the other you've got job shops that
can work on a much smaller scale but the parts are too expensive for most
markets.
So this is problem that I've been trying to solve for a while. There are
the techno-utopians who think that 3D printers in a few years will be able
to replace the mass production processes, but in reality these processes
are going to be with us for a while - so we need to find ways to lower the
barriers of accessing this equipment affordably.
In the world of software, we've got the same problem with servers. You used
to have to buy physical machines and maintain them yourself. Fast forward
to today, and we've got cloud computing which allows you to scale your
2. company within minutes. Today there's billions of dollars of servers
sitting in data centers worldwide for the production of data. In the near
future, there will be production centers doing exactly the same for
physical production.
+Interfaces Matter:
"The problem manufacturing has right now is that our interfaces suck -
you're playing telephone between the designer to the guys on the factory
floor. Here's a continuum of different manufacturing interfaces:
- We've got Google on the full-manual route of search for companies.
- MFG.com - puts you in touch with factories, check out reviews - but not
much more.
- PCH gives you access to a ton of experience and connections by using
experience people and systems.
- ProtoMold - can do low-volume production of simple parts using design
files and minimal human interaction.
- Ponoko & CloudFab - are as close as you get to full-auto production where
you input CAD and get parts back.
As you go down this list - you'll see that designers and engineers get
closer and closer to the production process. The trend is that over time
more and more of production will be fully accessible through interfaces
that give the designers and engineers the most direct interface and
transparency to production.
+Making a Manufacturing "Stack":
Most websites today run on some version of what's called a "LAMP Stack" -
which are layers of software between the server hardware and the internet
apps that we all use. The idea is the each layer takes care of a certain
set of tasks - the operating system is the foundation, there's a database,
the application code, and the interface with the internet.
This concept of building task-specific layers on top of hardware can work
just the same for manufacturing hardware too - just like servers today. So
let's get into the meat of this talk by looking at the overall vision of
what this looks like.
- Abstract: Make the machines easy to interface with.
- Automate: Remove the repetitive tasks.
- Network: Link the machines together to build an flexible assembly line.
- Compile: Automate the conversion of the design into machine commands.
- Debug: Catch errors quickly.
3. +Abstract:
What I mean by abstract here is to build an interface that the computer can
work with - making the physical reality in data. It's like when a person
becomes a line in a database somewhere. They've been abstracted, but now
you run algorithms on them. We can talk about the big data of manufacturing
if we have time at the end.
In practical terms, we need to take things like a milling machine and make
an interface that let's it be controlled fully by a computer. So we can
change tools, load parts, send it designs, etc. - all without needing to be
next to the machine.
+Automate:
The Golden Rule of Software: Don't Repeat Yourself!
In software, you write a piece of code and then turn it into a simple
command to be used later. In manufacturing, we're repeating ourselves every
time we make something like a mold - using very similar steps with slight
modifications.
Now that we've got machines abstracted - we're able to write programs that
remove this repetition so engineers and operators can focus on higher level
tasks without worrying about specifics unless they want to.
+Network:
So we've got one machine that mostly runs itself - imagine if we did this
for each production tool? For instance, having a mill, EDM, pick-n-place,
board fab, and robotics would allow you to mostly automate the process of
basic electronics products. Such a system - due to the baked-in automation
- would be a rapidly reconfigurable production line.
+Compile:
So I'm sure the more experienced amongst you are asking - well maybe the
machines run themselves - this largely happens already (like Toyota), but
who's going to program and configure them?
This was a problem for early programmers as well. They were handwriting
code at a level just above binary, and it was something a mathematician
needed to do to get any efficiency. What happened though is we invented
whats called a Compiler - it turns a human-readable code into binary code
specific to type of computer you specify.
4. In manufacturing we have what's called CAM. It's like a compiler. It takes
CAD and turns it into G-Code - the language of milling machines. We need to
make SuperCAM that improves over time to able to be this better than humans
ever could. This is exactly how compilers were built for computers - slowly
they evolved super-human strength. This allowed engineers to spend more
time solving the problem and less dealing with the computer hardware.
+Debug:
The beautiful thing about software is that you type a bit of code - and you
run it in seconds to get feedback on what's wrong. In hardware, this is
loop measured in weeks or months. This has to get faster in order for these
systems to be as accessible as we need them to be.
Now that we see that our production line is fully abstracted, automated and
networked - we're actually able to do our setup in minutes to hours, and as
simulation tools get better - we'll be able to simulate production as
quickly as we debug software today.
+Recap
So to recap:
- manufacturing methods today are actually pretty good - the bottleneck is
that our process was designed for producing a million of something.
- However, we now have the technology to build a system that makes
production far more automated and empower our designers and engineers to
push the limits of whats possible. This lowers the barriers to building
product, speeds time to market, lowers setup & inventory costs.