personal notes on fully automated manufacturing in the US
Something a little different. Going to write down some firsthand observations from a couple of visits to manufacturing automation integrators and production facilities as part of my work. I’ll try to paint a picture of how modern fully-auto manufacturing looks.
The connection to politics is from the Trump Administration experimenting with some protectionist trade policy. This raises the question of how a revival of a domestic manufacturing economy would look. Obviously, it would be enabled by extensive automation.
In a fully auto situation, the manufacturing facility has many individual production lines. Annual output for each line is typ. 1 million to 10 millions of parts per year. In extreme cases hundreds of millions, but that’s for simple disposable consumer goods (e.g., bathroom items). But for the jobs that are money-makers for engineers like me, the part of the higher-volume manufacturing world that is interesting is specialty automation subsystems going to medical or diagnostic products. So it’s likely single-digit millions of parts. The other big cash cow for an engineer is defense, which I generally don’t get involved in. Also, I’m talking about manufacturing of small assemblies here – typically hand-held of smaller.
A fully auto plant will go from raw plastic resin pellets to finished goods in boxes, or in racks of boxes. Picture a football-field sized room with the best air conditioning you’ve ever seen and all kinds of hoses coming in from the walls. Rows and rows of molding machines, paired with robots. Hoses and cables bring plastic pellets, coolant, and POWER to the molding machines.
The molds open and close every few seconds, and a handful of parts are picked out by a commercial robot with custom grippers, and placed into a specially designed carriage on a commercially purchased conveyor-belt system. The conveyor system then advances the molded parts through a series of stations which either perform an assembly operation, an inspection, or a sorting operation (such as removing rejected parts).
Additional parts from separate molding processes may be fed in. Some of the feeders I’ve seen are these crazy tub-like 1-piece stainless steel weldments, where thousands of plastic parts from another process are dumped in and start at the bottom of the tub. The weldment vibrates and the parts march uphill from the vibration, and come out perfectly oriented and lined up single file… One of those things you just have to see…
Anyway at various points in the production line, there may be inspection of the raw molded parts, clean-up of sharp edges (flash) from molding. There is likely joining of plastic parts by gluing, welding, snapping/pressing, etc. Installation of electrical connectors or circuits, metal inserts, springs, seals, lubricants, gels, liquids. There may be labeling, engraving, marking with a laser. There will surely be computer-vision type checks for defects of the assembly. There may be physical tests (like squeezing the trigger on a plastic assembly that has a trigger and measuring the force it takes, or testing the integrity of a plastic snap). There will be extraction of finished parts, and likely of partially finished parts, for sampling as part of additional QC outside the production line. There will be extraction of rejected parts — having access to rejected parts, sorted by the operation which caused their rejection, is incredibly helpful. The finished parts end up in bags/boxes/trays, possibly further combined into racks, bins, or pallets. At some point a person comes to take away boxes/bins/racks/trays/pallets.
A football field sized room with several dozen lines might have half a dozen operators walking around, and a handful of technicians in case the machines complain and something breaks down. Mostly though, as long as the manufacturers follow the instructions the automation designer gives them and do their routine maintenance, these things just sit there and crank out parts, with an acceptably low reject rate. The hard part is setting it all up in the first place.
The amount of labor embodied in this kind of production process is enormous. But it is contained in the construction of the production line itself, and in the mostly modular components from which the production line is built. I’ll break it down into some categories, and try to say where the embodied labor is typically from, for US systems I’ve seen (a small slice of what there is!).
1. Generic automation components (general motion components like slides, grippers, pneumatics, motion/position sensors, pressure sensors, modular framing, guarding, safety/interlock components, controllers/PLC’s/HMI’s/indicator-lights, network/comm components, tubing, cabling, modular electrical components of all kinds) — in the global market, all of the above dominated by Japan and Germany. For US production lines, the motion and pneumatic components are JP/DE, while the controllers and modular electrical are universally domestic. This is market is locked in very hard.
2. Tooling
2a. Major tooling 1 (molding machines, robots, conveyors, vision systems, presses, ultrasonic welders, laser markers) — dominated by Japan/Germany
2b. Major tooling 2, more specialized stuff (liquid/gel/glue/lubricant handling, bulk part feeders, finished part counters/baggers/boxers, rejected part extractors, tray/magazine handlers, palletizers, etc) — typ. domestic.
3. Systems integration for the automatic production line (engineering, assembly and commissioning of production lines, design and fab of custom mechanical parts such as conveyor carriages, robotic grippers or end-attachments, mounting/alignment/adjustment plates — typically hundreds of custom mechanical parts in a production line. Last but not least, PLC programming) — typ. domestic, but I just saw an entire line design and build outsourced in whole to Canada. Subassemblies and especially part-level mechanical fab increasingly shopped out to the lowest bidder in the global market.
4. Mold Design and fabrication — typ. domestic
5. Additional design of the finished product, required to facilitate fully-auto engineering — typ. domestic – the IP is carefully guarded prior to product reaching the market. Typically product design team has no communication whatsoever with engineers designing the manufacturing process. I worked on this side too but not on anything destined for fully automated production. Low volume products are an entirely different process to go from design to build.
6. Production process/flow design, high-level design of the automated tooling line (i.e., figuring out requirements to which the integrator will build the automated line) — typ. domestic, but in part done by integrator so may be outsourced. The situation I have seen so far is that the requirements are typically developed jointly between the “customer” (i.e., company whose product is being manufactured), the integrator, and the manufacturer who will operate the production line. Really this is an absolutely critical part. If there is novelty in the product being manufactured (i.e., there is an assembly step using a tool or process that the integrator and/or production line requirements designer does not have experience with), then there will be trial and error, a pilot production phase, which is more likely to be fully in the hands of whoever drives the detailing of the production line requirements. That is, the manufacturer and/or the customer (company who owns the product). Sometimes this 3-way dance works, sometimes it fails. Considerable management involvement here. Pilot production projects are another aspect of this, as mentioned. Those can be fun, esp. if something new is being developed. After that, the integrators come in and they do the heavy lifting in terms of man-hours spent detailing and executing the design/build/some-testing of the production line.
7. Validation of the whole production line (an extensive and ongoing task) — typ. domestic, may be required that it is domestic, by regulations if medical or defense.
8. Operation of the production process — depends on the industry. For the high value, low volume stuff, domestic. For consumer low value hi volume stuff, plants set up in Mexico or China.
Trends going forward — starting to see the more generic engineering get outsourced. Talking about the orchestration of the moving parts of a production line, which is a lot of work but relatively routine. Custom component fab is definitely getting outsourced, that’s well underway already. Electrical fab at all levels of integration has been outsourced for a long time now. Increasingly modular design means electrical design is on its way out too.
Conclusion:
A domestic manufacturing revival would be very cool, but not only will not create more than a handful of blue collar jobs, it will not even create that many technician or engineer jobs.
Like his predecessors, Trump is barking up the wrong tree. If manufacturing is to again become a significant part of the economy, the issue of redistribution of wealth remains. If manufacturing does not again become a significant part of the economy, the issue of redistribution also remains.
Pete's Politics and Variety 