I moonlight at a machine shop. I'm not a machinist, and I'm hardly a machine operator.
Machine shops need bodies to do a few things (some of which OP mentioned) loading materials, removing materials, changing broken cutting bits, cleaning the machines (especially if you're cutting something like magnesium).
Machine shops also need bodies to build jigs to hold pieces to the machine table while the machines remove material (think 5 axis pieces, especially).
Despite all of this – there's an amazing amount of automation at play. Tool changes, coolant, temperature monitoring, the list goes on. I'm currently building some automation hardware and software so we can start to fill the gap of 'dead' time where machinists are pressing buttons, loading in blanks, and are bored out of their minds.
A few years back I toured a machine shop staffed primarily by blind individuals. The Javits–Wagner–O'Day Act requires federal agencies purchase some select goods made by the blind when possible. Roughly three hundred employees, 90%+ somewhere between legally blind and fully blind, and 1/3rd or so also deaf.
Someone more skilled than I could and should fill a book with the accessibility changes they have made over 50 years of operation. Every single inch of the place had fascinating design details. All machines were laser fenced. Vacuum work holding with audio alignment sensors. All paths had a raised curb. All readouts on CNCs were capable of displaying text one 15" tall letter at a time, or reading G-code aloud or via braille. All offices have one bright wall and one dark wall to provide high contrast for sign language interpreters.
One of the sighted employees I met works in the accessibility dept. redesigning systems to be compatible with various disabilities. He had a full lab of 3d printers, CNCS, and enough other equipment to make a makerspace jealous.
The company also did some amazing outreach through their charity. The provided dogs, housing, and general mobility training (how to navigate busses, etc.).
I should check in and see how they are coping with covid.
You see, it’s quite useful to talk about it, I was looking for a charity/NGO/open-source software to sponsor and, for a specific reason, Postgres or Debian (which we use) don’t comply to our requirements. You are giving me a new group of charities/social companies to search into.
This is absolutely the coolest thing I have seen so far this year. They have more videos where individual works explain how the work from a call center operator to a machinist journeyman!
Laser fencing is the worst, pressure mats are approximately 1000x better. Our machine with the laser fence is always stopping for no reason because there was some dust. I just cut 50 sheets of MDF yea there is going to be some dust.
In woodworking the new trend is to put pads on the gantry with pressure sensors so if the machine smacks you it stops.
That is really great when you are testing new programs you can see what is happening without binoculars. The only downside is that they limit the max rapid speed on the machines so when it hits someone they are much less likely to get knocked over.
How do those systems stop hair or clothing from getting sucked into the spinning bits? Those two modes have always been the more dangerous ones for spinning tools.
I have no idea why the people in that video cut a hole in their door. Seems like a really stupid idea to me. But you can see in a normal machine that someone hasn't modified to be more dangerous you can't get to the the spiny sharp bits.
Looks like that video was a demonstration from a company that sells these machines. The hole was just cut out for doing videos like this I'd assume, not a modification for regular use.
Yea but in the US at least if OSHA, or way more likely our insurance when they doing their yearly inspection saw that they would freak out. That company looks like they are German, and from what I've heard Germany is way more strict about safety than the US.
So in order to sell that machine to me I would require a safety compliant door. My guess is a new door would be around $2800. That is an expensive hole.
What I know that you don't is that that machine is sold with a camera as an additional option. So there is already a place inside for mounting a camera. They would just need to make their own mount. Otherwise if I was just filming a demo to sell the machine I would override the door closed safety interlocks and film with the door open temporarily.
Humans are inefficient and unsafe machines, to be sure. Soft and fragile, slow to respond, 98% blind to the EM spectrum, insanely low communication bandwidth, unusually difficult to motivate, prone to acts of defiance, uncalibrated and wandering accelerometers and rotation sensors, very low precision parts, require a host of special atmospheric and environmental conditions to be met, etc.
I didn't catch anything unsafe. They have laser trip wires anywhere things might be dangerous. This is becomming common industry practice anyway because even people who can see can miss seeing something - a common problem when doing repetitive tasks is to make it faster and faster without realizing that you are getting closer and closer to something unsafe until you loose your arm.
It isn't clear how inefficient they really are. Sure it takes longer to explain to a blind person who to clean the chips off, but in the end it is just a broom (vacuum?) and the only need to do ensure they get everything. Once you train them they have it. Likewise it takes a little longer to explain where the hold downs are then to explain it, but not that much and there are good reasons for hold down placement so once the blind person learns the location there isn't much more training.
I assume they have a few sighted people around for the unique setups that are not enough like something else that it be put into brial. (deaf but not blind can be good at this)
Again back to my first point: everything they pioneer and make work is something that other industry will be interested in as well. Maybe someone with sight can do it easily today, but often once they perfect a method it is better than what the sighted person can do.
Overall I'd prefer to not be doing manual labor myself, but if you can't handle enough school to get a good degree... I'm glad places like this exist for those who need it.
Exactly. They are actually efficient enough to be quite competitive in the private sector. They have large contracts with Boeing, and they manufacture a wide variety of things like white boards, backpacks, shovels, canteens, etc.
That's weird, it sounds incredibly safe to me. I've never been in a machine shop that would be safe enough for blind people to work. OTOH, my mom was responsible for keeping factories safe for the workers who hadn't lost their hearing yet...
Many moons ago I also spent a fair amount of time in a machine shop, although never more than in an amateur capacity (I'm not sure I'd call what I did even moonlighting, mainly an enthusiast making small stuff), and I was never using really fancy machines as a result.
What struck me was how just how little was automated, how stuff that was nominally automated still had quite a bit of manual labor. I had always had in my mind that with CNC machines you just stick the metal in the vise, load the program, hit start, and you're good to go, but there's manual calibration, facing off and the like that needs to be done before every run.
For small, simple pieces I would often forgo the CNC automation and just manually make the piece myself, even when I had already made the piece in SolidWorks (and so could easily generate G code).
I've heard that for really fancy machines it's truly a push button process (as long as you feed it precision milled blanks), but I've never had a chance to actually use those.
You can get electronic probes that measure the stock automatically so you don’t need precision blanks. The probe loads in the tool changer so it can be completely automated. Automatic tool setting makes it so operators don’t need to measure new tooling in tool holders, and tooling wear can be monitored automatically.
I dabble in running a small machine-shop, and we are able to automate a good chunk of what people still do manually. The stock we use is cheap, straight from the mill stuff as we have the machine measure the stock with an electronic probe. The machines also have 20 pallets, so once the stock is loaded, the operator can leave the machine for 20x the cycle time of a part. The pallets don’t even have to be the same product, so we can queue up replacement of inventory with just the quantity that a customer ordered and offer a bunch of made-to-order parts with reasonable turnaround times.
The machines also monitor spindle vibration so they can tell if a tool looses an insert, and the tool-setter is used to check if solid tooling is still intact.
The only manual parts are taking raw materials off the suppliers truck, unloading finished parts (next on automation list), final assembly (working on automation for this), occasionally loading new tools as they wear/break and fulfillment.
The machines also monitor spindle vibration so they can tell if a tool looses an insert, and the tool-setter is used to check if solid tooling is still intact.
That's a key feature. For unattended operation, you must have good fault detection. This tends to be an overpriced extra cost option on machine tools.
There are a lot of things in industrial automation that cost more than they should. Motors with encoders, for example.
> There are a lot of things in industrial automation that cost more than they should. Motors with encoders, for example.
Not really. Motors with encoders aren't really much different in price from the alternatives at industrial sizes.
What you do see is that the transition from hobbyist machines to production machines is quite a significant price change. The jump from, say, a Tormach to a Haas is almost x10 (about $10,000 to $100,000).
This is not a surprise. Those machines are expected to run 24/7 and, if they don't, people are going to get angry and reputations come into play.
This space is optimized for production and if you're not, you're the outlier. Pre-Covid I accidentally tripped a Sunday delivery at the office from one of the machine tool suppliers--scared the hell out of me when someone knocked on my office door at 3:00PM on a Sunday. The delivery guy was just as confused--he had never delivered to such a nice office area. We both had a good chuckle about it all.
(Side anecdote: I had a really nice conversation with the late founder of Tormach many moons ago at a LittleMachineShop open house about servomotors vs steppers. He was quite blunt--the issue wasn't motor cost but customer service cost. His customer service budget would need to go up about $300K per year for about 3 years every time he introduced or changed some major machine feature. So, any change needed to earn $1M over 3 years and then be net positive after that in order to get implemented. Servomotors wouldn't pass that threshold--so unless a competitor forced him to, that upgrade wasn't happening.)
Most things I’ve found to be reasonably priced for what they are.
Most products are made with much higher quality materials than you would find in consumer gear. Also the “rated” spec for parts is expected to be understated as the “unreliable” part is one that fails in under 7-10 years of being run at the absolute limit of its spec. If a part fails in 2 years I will completely stop using that vendors products, as having a machine that is worth $1k/hr going down for even a day or two wipes out any savings.
Another factor is the total sales volume of automation products is low, so the engineering costs dominate pretty much everything in that space, even though everything is somewhat optimized to reduce engineering time (the buyers and/or vendors).
I don't think actuators are crazy overpriced really. It's true that if you are going to bother packaging them you are probably using decent-to-good quality everything, but that's what you want for non-toy projects anyway. And for toy projects if your budget is really tight no big deal to set up the encoder yourself on this one-off.
If you are careful yes. Doing job shop work (making other peoples parts) is hard to make a good profit as people expect overseas dumping rates.
The real profit comes from making your own products, as then you can force competition to also build their own factory to compete. By having lean production we can have a very large part catalog without having large amounts of inventory for low sales velocity parts.
We have at least a hundred variants that may only sell a dozen or two $100 parts per year, but they only cost us a line in the catalog to maintain the SKU. Since the products are for industrial use, most of our customers like the fact that we have in some cases been making the exact same part for nearly 30 years, which encourages them to design our products into theirs as we never discontinue products, so they don’t have to re-engineer theirs (our products in turn get used in machinery that might have a 10-100 unit/year global market).
As I understand it, the economics of machine shops are a bit weird because of the financing cost of keeping up with the latest equipment. If you have old machines, and not money to buy new machines, then you run old machines.
The most modern shops are in extremely high demand, expensive, always busy, happy to turn down work. If you can design something in a way that lends itself to conventional techniques, you can get it into a smaller shop that may have some idle capacity, or run on the "tool room" machines in the big shop. The big shops always keep a few old machines around for tooling and jig work, rework, and one-offs.
The smaller shops are also willing to take work that's not 100% detailed, even hand drawings. So you don't necessarily even need a CAD operator. "Do things that don't scale."
In kind of an odd analogy, my old high school band mate built a recording studio from cast-off obsolete equipment that he bought for pennies on the dollar, and it meant that he couldn't take the biggest jobs, but he was instantly profitable and never in debt. Likewise, I have a very small side business that involves some basic metalworking, and I do all of it with powered hand tools and jigs that I made from plywood and carbide drill bushings. My capital cost was under 100 bucks.
> I had always had in my mind that with CNC machines you just stick the metal in the vise, load the program, hit start, and you're good to go, but there's manual calibration, facing off and the like that needs to be done before every run.
Much of it depends on the volume.
If you're building a million of something, the up-front cost to automate every step of the way makes sense.
If you're only building 10 of something once every few months, paying someone to do the manual operations 10 times makes more sense than investing in all of the fixturing, programming, and other tasks.
Really though, a lot of smaller machine shops have older CNC machines that are paid for and otherwise sitting idle. Hiring someone for $60/hr to do all the manual steps can keep that machine profitable.
It's all about the tools the shop has. Sapphire probes (edge finders) aid in a lot of the process. Mastercam allows you to simulate the machine process (and you can usually simulate it completely). Every shop is different – the shop I'm at here has a few lathes, robo-drills, 5th axis, and a few older pieces of hardware (which still are able to receive G-Code).
I was definitely using edge finders (it seems like machining in general would be way way way more work without edge finders) and occasionally Mastercam.
Definitely never used a 5-axis mill or robodrill though.
I guess my main point of comparison was to laser cutters, which really were basically push button automatable in comparison.
Granted the problem is a lot easier (purely 2-D cuts) and the cutter loses a lot of precision for thick material, but any time I got to use a laser cutter I was blown away by easy everything is and kept wondering whether a similar process could ultimately take root for machining.
The forces involved in CNC machining is a large complicating factor, compared to laser (or 3d-printing). In a laser one can just throw the material onto the bed, and the results will be fine - not so on a CNC. And that one is quite inherent to the paradigm.
Other aspects could likely be simplified and automated to a larger degree than today. In many areas solutions exists, but are quite high budget - things like tool changers, material loading and unloading.
Machining can be simpler if you've got a couple precision features that can be clamped to with hard fixturing. It's all about positioning the workpiece accurately so the fixed g-code program can do its thing.
You also need bodies to listen to the tool... ever since I build my small CNC machine I’ve learned just how important it is to listen to the cutting especially when you are doing adaptive passes to make sure your software doesn’t cock up the cuts.
This is probably the biggest area where i can see machine learning be actually useful with acoustic sensors.
I’m not an experienced machinist by any stretch of the imagination but even with my relatively small experience I can tell how good or bad the finish of a cut is going to be based on the sound the machine makes.
I would assume it does especially on 6-7 figure machines but I think even with basic microphones this might be a good application for ML at the small shop and hobbyist levels and I haven’t seen that yet.
Anyone know of reasonably priced acoustic emission sensors? Even on AliExpress I could only find transducers that start around 500 USD. Would love something closer to 50...
My first CNC experience was with a (C)NC machine from the 70s. I loved it at times. It was really amazing to see the machine shops being upgraded. The elvel of automation, already back in 201/12 (last time I had any close contact with these things) is amazing.
The concept of "Hauptzeitparllele Tätigkeit" (stuff operators can do when the machine is doing its thing, one of the things German just describes so well) stuck with me ever since.
The other day, I realized our Haas CNC machine actually had a programmable coolant nozzle. You can program the nozzle to spray coolant at various locations!
Yup, it's fun to make a "wash the part off" program that drives the table back and forth a few times with the nozzle at progressive angles. Or aim right at the perimeter of a big face mill.
Depends on where you live. I recommend just going around and meeting machinists. Look into local meetings...Society of Manufacturing and such. Typically there are things like shop tours put on monthly (non-pandemic times). Talk to local machine shop instructors and/or metal suppliers and see who might be interested in hiring...they usually know everyone.
My area has many very, very large machine shops but it has a larger number of small "mom and pop" shops which just run a few machines. I have found that the people who run either type are equally as approachable...it is hit and miss if they are willing to even let you in the shop at all without NDAs and such...then there are others who have no issues with it.
Most of the really difficult to learn stuff in the field is fixturing and CAM setup. Once that is layed out it is mostly just changing parts and watching the machine run...but there is a lot to be learned while that is going on.
I've never moonlighted, but I agree about the approachability.
Some time ago I lived in an industrial city with lots of small machine shops and especially lots of Screw Machine shops. I found that they really appreciated that I was interested in what they did and were willing to show me around.
Fast forward to about a year or two ago and I had to call on a prospective customer with a sales person. I didn't know until I got there that they were a branch of a well known workholding manufacturer and our contact offered to give me a tour of their machine shop. Pure heaven: I've never seen such a collection of modern, massive machine tools. The toolchanger alone on one of these machines was bigger than most VMC's I had ever seen. Their pallets were multiple feet long (not even sure if you call them pallets at that point) and ran on external tracks from machine to machine. Easily the biggest lights-out shop I've ever seen.
That was by far the most fun part of the sales call :-)
Thanks for the reply. When / if the pandemic ends I have to move and I am seriously considering taking a few machining courses at a community college and changing careers.
At the start of COVID, I walked in, and told them I was interested. Coincidently, they needed some networking help. We ended up trading some skills, and then I just kept coming in. I didn't take no for an answer, and they liked it. YMMV.
I wrote an email to friends years ago about a side job I got at a machine shop while working in tech, it was quite well received. email in my profile, contact me and I'll send it to you.
> Those skills need to either be brought in or be taught and as soon as a person has those skills we are back at having a skilled person doing a large amount of unskilled work.
I’m completely in awe of this attitude. Are you located in the USA? I know a bunch of people in small part manufacturing (mostly metal and plastic) that are turning away work on a daily basis. There is so much demand in the USA right now that any worker with the skill to increase production rates is like gold.
I think you misunderstand the authors point. There's unskilled aspects of a job that seem like an allocation waste to have skilled labor handle. But to properly account for some costly errors and such, an unskilled worker has to be trained...making them a skilled worker that should be able to better allocate their time elsewhere than the unskilled tasks.
Based on the people I know, paying someone skilled wages to do an unskilled job is more than worth it right now.
One of the shops I know of is making small little pieces of aluminum in the central US and shipping them to Foxconn in China for assembly into computers. As fast as they can make them, which is constrained by the number of people they can get on the floor. 2020 was their best year ever, by a wide margin. Things are seriously crazy right now. If you’re wondering why, the answer is most likely tariffs. The more American value in the product, the lower the tariff once you import the computer. Nobody knows if Biden will continue this.
From what you've said semi-automation is even more crucial as the more checks and automations you have the lower the bar of entry for the people on the shop. What a company pays is up to them, but having more options is always for the better.
If you're interested in machining, check out the Business of Machining podcast. https://businessofmachining.libsyn.com/ Every week John Saunders of NYCCNC talks about his business with John Grimsmo of Grimsmo knives (they make $1000 pocket knives on cnc machines). It is an excellent podcast about machining and business.
Would you have a favorite episode or two to recommend? When someone recommends a series that sounds interesting and I see there are 200+ episodes, it's hard to know where to start!
Yep, one of the example in the Wiki is robots building robots since 2001:
"FANUC, a Japanese robotics company, has been operating as a lights-out factory since 2001.[6] Robots are building other robots at a rate of about 50 per 24-hour shift and can run unsupervised for as long as 30 days at a time."
With IIOT (industry internet of things) that definitely seems like a possible reality. However, as long as Skynet is controlled by ex-factory workers it will be a promotion for everyone.
> First, was the network setup. We needed to merge multi-site networks into a mesh topology network where all the nodes could be interconnected which was solved using a hub and spoke topology VPN. Although, it does have its own drawbacks it had a good cost-benefit ratio when it comes to setup and deployment and is working for us for now.
How is "mesh" solved with "hub and spoke"? If your hub goes offline, everything stops; the whole point of a mesh is to avoid the SPOF.
> Next, was building the dashboard which not only required the making of the UI but required figuring out how we’re going to expose and manage resources in our shop network. We decided to build the backend in the spirit of Kubernetes, so it is easy to manage at scale with the addition of new sensors, machines and full-blown sites themselves.
Indeed, as you noticed the networking setup is primitive, but for the time being it works for us. Currently, the hub is chosen to be as reliable as possible and nodes at the end of the spokes are prepared for hub down (ie. they stop safely ).
Kubernetes is more than just Docker containers. Having managed small computing clusters for a decade using custom scripts I find that the value of K8s is how easy it is to manage highly complex applications over an array of loosely connected computing nodes. So in my view k8s includes networking, orchestration and principles of how everything should be done and managed to achieve redundancy, performance and stability.
Containers are just a convenience for simple deployment on asymmetric nodes and is only part of K8s.
If you're actually running kubernetes and you say you're running stuff "in the spirit of kubernetes" it gives the strong impression you are not actually running kubernetes.
I also didn't say your networking was primitive, just that "hub and spoke" and "mesh" are two entirely different things that you seemed to have conflated entirely.
And that is correct. We are indeed not running kubernetes because it is not supported on our hardware and some assumptions were limiting. So we have a new system, but having kubernetes as our ideal.
Hub and spoke is the underlying architecture of the net, but we are achieving mesh like functionality having hub as our central router. The wording might be a bit confusing.
IPIP/IPsec and OSPF would do nicely for efficient site-to-site meshing.
But, adding docker to the equation makes it much more complicated. I'd prefer just operating the machines over IP and get rid of all remote desktop related latency all together.
We built a similar system for 3D printers. There's nothing quite like intentionally giving access to someone on the other side of the globe and having them make something for you in front of your hands and eyes. It's not technically teleporting, but close enough to have that same magical feeling.
I work with large scale flat pack furniture manufacturing. Unattended machining is possible and not uncommon. You have a large tool changer on the CNC, an auto loader to get material on the CNC, and then a stock warehousing robot to fetch the material. You just talk to one of the large equipment manufacturers and they'll sell you this sort of setup.
This seems like a strange problem to try and solve. As a former software developer I see problems every day that AI or advanced expert systems could solve I'd love to chat with these guys. If they solved some of the problems with QC or Packing the industry would be clawing at their doors to buy their products. Not to mention the outdated state of most CAM / nesting packages.
Being software developers by nature and having got ourselves into CNC we feel exactly the same. It is crazy to us how easy it is to make a dumb and costly mistake.
So we are tackling it one little thing at a time to make operators life slightly less stressful.
We're always open for a discussion so feel free to reach out ( the contacts are on the site of the blog )
This is cool stuff. I don't know the state of the art here, but it seems like making progress on remote factory operations would dramatically speed up our ability to industrialize space.
Yes! Most CNC setups will flood the cutting area with coolant. It moderates tool temperature and also helps carry away chips.
The machines have filters to recover and recycle the coolant. They might also have augurs to carry chips out of the machine.
The coolant can be topped up periodically between shifts as needed, when the waste chips are collected.
At scale, these things can be automated even further. For medium sized shops it's easy to pay someone for a couple hours of labor to do it every few days.
Typically only when cutting metal. And harder metal especially. Wood and the like will often have an spray of air directed at the cutting bit but thats simply to help clear away chips and clean the cutting surface. Helps keep a nice finish and good visibility.
Actually, it's a little different. Using coolant is the norm even when doing mild steels. You can remove material quicker, your part remains cooler, and your tools remain cleaner.
There's even cuttings bits that have integral cooling (holes in the cutting surface).
Right--I support a fair number of machine shops and they're all moving to CNC (although they're using a lot manual machines still). 5 axis is nice and there's a surprising amount of manual tooling changes.
We had a problem where one of the integral cooling loops had a tube come off and it sprayed the better part of a 50 gallon drum of coolant all over the shop before someone got over and hit the e-stop.
Pretty sure even the mildest steel is still a hard metal. Soft metal would be something like copper or aluminum IIRC. (Or lead or gold, moreso, but you probably wouldn't want to be machining those for obvious reasons.)
Yeah, watch some CNC videos on youtube or whatever! The (usually blue) liquid spraying everywhere is coolant.
Fancier machines will have a hollow spindle and can actually force the coolant through an also-hollow tool, so it emerges right near the cutting face. The simpler approach just uses a nozzle pointed in the general direction of the cutter. In between, there are programmable nozzles....
I am convinced being a skilled machinist with proper engineering background will be one of the hottest career tracks in the next couple of decades. Seems demand for small batch production is only going to increase.
Having hired subcontractors from offshore I can tell that there is always a background difference that decrease the value.
Covid has shown that a lot of office work can be done remotely, but that does not mean that all the businesses suddenly will go away offshore. It just becomes a perk of the job.
And, thus, we're close to having machine operators working from their beds :)
Machine shops need bodies to do a few things (some of which OP mentioned) loading materials, removing materials, changing broken cutting bits, cleaning the machines (especially if you're cutting something like magnesium).
Machine shops also need bodies to build jigs to hold pieces to the machine table while the machines remove material (think 5 axis pieces, especially).
Despite all of this – there's an amazing amount of automation at play. Tool changes, coolant, temperature monitoring, the list goes on. I'm currently building some automation hardware and software so we can start to fill the gap of 'dead' time where machinists are pressing buttons, loading in blanks, and are bored out of their minds.
It's such an interesting space.