Plastic Injection Molding (2015) [video]
youtube.com
youtube.com
My plant was in Wisconsin but wasn't heated during the winter when the lines were running. Melting plastic takes so much heat that we were the main electricity consumer in my town.
Because there isn't a check ring in place while the screw is turning, plastic is always trying to slip backwards over the flights of the screw. You have not just heater bands around the barrel, but also cooling water, which makes the plastic less melty so the flights can push it forward without slipping back. The more recycled material you use, generally the cooler you need the barrel because those little flakes melt faster and slip more than virgin pellets.
The many challenges of fighting reality to get your plastic to come out correctly mixed and without bubbles build a lot more character than similar paying jobs like driving a forklift. Facing the same problems repeatedly helped me develop the quick-access note system I still use for debugging and syntax. I wouldn't be half the programmer I am without the life skills from manufacturing. The only problem with being a line operator was that most permanently solvable problems have been engineered out already, so being clever and organized about solving things didn't produce nearly as much value for the plastic factory as it does now that I'm a programmer.
Even though he’s never done software before, we’ve had lots of interesting discussions about debugging because the process is the same. It’s just that my machines of choice are abstract and his are large, physical and dangerous.
The amount of helpful stuff grew fast and I had to keep erasing my notebook and reorganizing it to find things. Settings were organized by machine number, defects by symptom, job changes as a checklist, alarm responses as a flowchart. (You only had 3 minutes before your line would shut down when the 'out-of-pellets' alarm went off, so you had to consider only solutions that might work).
The machine produces a half-ton of plastic every hour so my lookups had to get more and more efficient. Every second you spend looking stuff up means more plastic you have to pick up and throw in the scrap box. I switched to a Word document I could print out and bring with me. I used Word's four levels of headings and the "Generate Table of Contents" feature so I could find the exact page with my issue in seconds. I kept the most important six pages folded up in my front pocket for immediate access. Things like the stacking table ceasing to lower so that the plastic would jam up within minutes.
Every day when you make plastic you fail and waste money, it's very challenging. The consequences are much more tangible than in programming -- orders don't get on trucks, people have to roll up hundreds of pounds of plastic off the floor, the line goes down and you have to spend an hour sweating to get it back up again. So every day I fixed my notebook so that day's timewasters would have been solved faster. Write down how to fix things without calling maintenance, record the solution that worked and not the five that didn't, add a step to a checklist.
So what this did that carries over to programming is it makes you start using your notes as an extension of your memory. There were fixes I wouldn't use for months but could instantly access by the situation (even though I didn't have Ctrl+F). Because of the speed of the lookup, I wouldn't even bother remembering these things at all, which gave me more working memory.
Now I have 280,000 words of notes about programming, but it's not like college note-taking where you'd have to skim pages and pages to find what you need. There's a Python.docx, Concurrency.docx, Testing.docx. It's all organized by headers like "Design patterns", "refactoring conditionals", "String.format expression syntax". That way if I can't remember exactly what I wrote to Ctrl+F, I can still get there very fast -- and see all the other related notes beside it.
They're all on AutoHotKeys so that I can just Ctrl+Alt+D to open "debugging.docx" and search "ConcurrentModificationException" and see exactly what the typical errors I make are that cause this exception and how I solved them last time.
In the end, just like I was able to move to any line in the factory and run it as familiarly as if I had been there for months, I can move from writing a context manager in Python to doing conditional inserts in SQL and recover all the expertise I ever had in under a minute. It's great at my job which is full stack from Bash to Javascript.
The same approach helped me revitalize our support wiki. Walls of text became "if this, click to expand. If that, go to page X". Related issues got stored together so you can go up the hierarchy a level if one approach fails and try others. Information got moved to right when you need it instead of buried on some other page.
I feel like people who use Confluence and text notes to expand their quick-access memory like this are kind of "digital-ready" -- it's like our brain is expandable with an SD card slot that others don't have. Good notes let you crystallize a bit of knowledge every day so you have more room to learn something new the next day.
Great stuff
Glad at least two people will see it.
Driving a car without shock absorbers is also a real experience! With alternating braking and accelerating, you can build up an oscillation that's big enough to bounce the wheels off the ground with almost no forward speed.
And now she can run around the house finding the ejector pin witness marks on everything.
If you can find them somewhere, hopefully she'd get a kick out of them. :)
The worker let us cut one or two of those ejector thingis that are attached at the end. He even let us keep one, we called it “the cup with a tail”.
Fun times, so easy to be amused as a kid!
Aluminum baseball bats have a similar manufacturing method - https://www.youtube.com/watch?v=didmRLz4vfU
Topics this video didn't cover include friction/slippage along the barrel wall, circulation in the melt between each screw flight, the thrust bearing at the back end of the screw that takes the load from pushing the plastic forward, two-stage screws that drop the pressure in the middle to let gases be removed from the melt, how the plastic cools differently at the edges and center if it is crystalline/amorphous, and more.
I remember as a kid my dad once got some sort of plastic that you made by mixing two liquids (they were very liquid, so not epoxy), and I made molds of things out of putty (or clay?) and then "copied" them with the plastic. Any idea of what this room-temperature binary-plastic could be?
My usual recommendation to people trying to start something hardware related is to do as much as you can in the beginning to make "sales", with sales being defined as whatever you can get that proves someone would actually pay you money for your widget. Prove you have a market before you pay for expensive tooling. That process can vary, but crowdfunding is one example.
Some rough rules of thumb if all you are interested in is cash outlay:
need 2 parts: 3D print or CNC
need 10 parts: silicone mold or CNC
need 200+ parts: injection mold
If lead time to first part, part uniformity, or uncertainty about design changes are factors that might also swing you toward or away from injection molding.
For injection mold tools, assuming the part has a normal level of complexity:
2cm cube: ~$3k, 5 weeks to first shot
15cm x 5cm x 2cm: ~$4k, 5 weeks to first shot
25cm x 15cm x 5cm: ~$8k 8 weeks to first shot
First shot means the first time the tool is tested to make samples. Generally there is a sample approval and testing process you have to go through before any remaining tweaks are made + the final mold texture or polish is added. I find the total time has more to do with how organized and diligent the client is in responding, but assuming nobody drags their feet we generally can be production ready in another 2 weeks or so.
We only work with production tooling (hard steel, lasts a long time). From checking around, if you use aluminum tooling or other "cheap" fast turn prototyping stuff the price doesn't seem to be any less, and in many cases is more. Tooling made in America is usually significantly more... maybe 1.3 - 3x more.
If you only need 2000 or so for a simple part to start, and are willing to pay 2-3X the part price, Protolabs is a good option.
Another low tooling cost, low volume option that might be worth looking into is cast urethane.
However, for a 2-up mold for something the size of a game controller, with a 0.5mil shot life, with maybe a slider or two, you are looking at $8-10k at many of the molders I've worked with.
I'm not seeing these on my MacBook. Is that because the parts are not injection molded, or is it because these features have been sanded off or done in a clever way so that it's not easily visible?
The original unibody aluminum Apple TV remote is a masterpiece.
A lifetime ago, we were working with Foxconn and a colleague managed to sneak onto an Apple floor and take a look at some of their tools -- he was gobsmacked at what they were capable of. The stuff of industrial design engineers' dreams.
Amusingly, the cheap plastic resin chair shown prominently in the video probably isn't injection-molded. Those are usually formed from a flat sheet.
This guy has open source plans for garage-scale plastic recycling machines. If you’re interested in alternatives to new plastic from china https://preciousplastic.com