Factories are short of people: Schaeffler's robots, shifts and an order calculated in 65 seconds

This week the manufacturing news circled one topic from several sides: Schaeffler opened a plant in China where humanoids assemble parts for other humanoids, the US Midwest is short of forklift drivers, and a British coatings contractor can see 48 to 72 hours of work ahead. For a manufacturing owner this topic has a down-to-earth layer: how many engineer hours go into recalculating an order that a program could handle.

What came out

On October 23, Manufacturing Today reported that Schaeffler opened a plant in Taicang where humanoid robots assemble components for other humanoids. On October 20 the same publication described how Midwest manufacturers are covering the forklift operator shortage through dealer partnerships and new equipment, and the same day it looked at why ordering custom components loads procurement harder than standard parts does. All three pieces are short, and I'm not taking details from them.

On October 23, Engineer Live published a conversation with Carl Allison, director of Plastic Coatings in Kingswinford: an industrial coatings contractor since 1952, 80 people, round-the-clock production. The main problem, in his words, is unpredictability: 48 to 72 hours of work ahead, and if nothing comes in tomorrow, on day three people have nothing to do. Finding people who understand ovens and conveyors, he says, is hard.

On October 19, Engineer Live reported that MathWorks extended its TÜV SÜD certification for MATLAB and Simulink to AI tools under the automotive standard ISO 26262: agents create and edit models, run simulations and generate tests, while the engineer stays responsible for checking the result.

On October 20, CNews reported that the implementation center Kompyutery i Seti (“Computers and Networks”) completed the automation of shift planning and staffing for a manufacturing enterprise. I don't have the text of that news item, only the headline and the teaser, so I won't say anything about the result.

What a contractor with a two-day horizon has in common with your shop

Let me explain with a refrigerator. It consists of doors, a compressor, hinges, bolts and dozens of other parts. The hinges have to be mounted, some parts cut out, the upper and lower doors painted. If orders for a refrigerator, a microwave and a stove are running at the same time, the same parts and the same painting show up in all of them, and there is no sense sending it to the contractor in three batches. For a contractor like Plastic Coatings, that is the difference between a day with work and a day without.

To put such batches together, somebody has to know, for every order, where, how much and of what. If that knowledge lives in the heads of two design engineers and in Excel, it stretches to as many orders as those two people can recalculate. The shortage of people in manufacturing often starts in the design department: the engineer is calculating bills of materials instead of designing.

How it went at a manufacturing company

From February 2024 the technical director of a manufacturing company and I prepared the requirements, active development started around April, in the summer, July or August, I handed over a fully working program, and I kept refining it until the end of 2024. The customer describes a product's composition in Excel: one product is made up of many specification files in a fixed template, and the 1C catalog has tens of thousands of items. Before the program, all of it was put together by hand.

The program takes the specifications and the design documentation and answers the question “where, how much, of what”: it breaks the product down into parts, calculates materials for the chosen quantity, checks them against the 1C catalog, takes processing notes into account and flags what is missing. It merges several orders on its own, so that identical parts and identical processing go to the contractors together. The program handles one order in 65 seconds on average; before, it took 4 to 5 hours. 11 people use it every day, each saves about 4 hours a day, which is about 900 hours a month. I put the logic on a server on Node.js and Express so the program can't be copied and carried off; the design engineers work through an Electron app.

I haven't worked with robots or humanoids and won't judge Schaeffler. For a company with 11 engineers and Excel specifications, the cheapest way to free up people is in software.

What I would do

1. Count how many hours a week go into recalculating orders, who does it and in which file. At the manufacturing company above it was 4 to 5 hours per order.
2. Take one real product and walk its path from specification to the jobs for the shops and the contractors: that is where you see which rules need agreeing with the process engineers first.
3. Keep 1C as the accounting system and put the calculation next to it, with a check against the catalog.
4. Build in merging orders by type of processing from the start, and discuss it with the contractor in advance.
5. If the calculation logic is valuable to the company, keep it on your own server and give people the interface.

Related reading

On the signs that spreadsheets are no longer enough: the article When a business needs a system instead of spreadsheets. Service areas: the business as a single system, 1C integration and software systems for food production. The project itself is described in the case order calculation in 65 seconds instead of 4 to 5 hours. On the same topic: where a spreadsheet is still enough and where I suggest a database and what reaches the greenhouse and the winery from the news about transit through Georgia. Earlier I looked at how to connect equipment, data and a working decision in manufacturing. If your design engineers and procurement put specifications into jobs by hand, tell me about one product: Discuss your task.