Control Engineering

Link This | Email this | Blog This | Comments (0)

When is automation not appropriate?
November 12, 2007

Over the past couple of weeks I’ve toured three production plants that show different levels of factory automation. Each, in its way, was the right level of automation for its situation.

The first two facilities were at Mitsubishi Electric Corporation’s Himeji Works in (appropriately) Himeji, Japan. The company’s alternator-production line is almost completely automated. It pumps out 140,000 units per month with a work force of 22 people split into two shifts. That works out to about 6,363 alternators per production worker per month. Of those humans, perhaps half a dozen actually touch products. The rest build, install, and service automated production machines. While the production line fabricates roughly 50 alternator models, this still qualifies as a high-volume/low-mix operation. The unit volumes are sufficient to amortize a lot of non-recurring engineering (NRE) cost for each model.

This is important because automation is all about minimizing total per-unit cost. Total production cost falls into two areas: NRE to design, build, and deploy the production system; and variable cost, which captures labor, materials, and any other expenses that increase as production volume goes up. Automated systems trade increased NRE for decreased variable cost.

Theoretically, any individual manufacturing task or step can be automated. With today’s technological level, we are very close to changing the word “theoretically” to “practically.” There are very few tasks that can’t be automated.

The question is; at what cost? Manufacturing companies are fundamentally in business to make stuff for sale at a profit. Management may have other goals, such as improving employee working conditions, and being a good corporate citizen of the community, but ultimately, they are in business to make profits for stockholders. (The stockholders, who actually own the company, wouldn’t have it any other way!)

The business case for automating a task hinges on whether the variable cost reduction generates enough savings over the machine’s lifetime to pay for the additional NRE. When you can crank out a lot of units before you have to make additional NRE investments, it takes relatively little per-unit savings to cover the substantial NRE investment automation requires.

As unit volumes decline, however, it takes more per-unit savings to cover a given NRE investment. The business case becomes harder.

For example, winding copper magnet coils for alternator stators is a tedious, repetitive task that is relatively easy to automate. It is, therefore, low-hanging fruit on the automation tree. It was, in fact, the first to automate with automatic coil-winding machines coming into vogue at the turn of the 20th Century.

Installing the wiring harness and making the final connections is hardest to automate. It is one of the few tasks on the Mitsubishi production line that remains manual.

The second production line I want to talk about is also at Mitsubishi’s Himeji Works. Board assembly for the company’s line of PLCs uses a lot of automated equipment—automated screen printers to lay down adhesive, tape-and-reel-fed pick-and-place machines to add parts, and automated inspection systems to assure circuit-board quality—at the production line’s front end.

At the back end, however, final assembly, packaging and functional test happen in a semi-automated workcell inhabited by three humans. While tests are standardized and automated, handling of the boards and loading subassemblies into chassis are still manual tasks.

There, the mix of units is higher, the unit volumes are lower, and the number of variations (for example, pass, rework, or discard after test) in product flow are greater. This variability favors the nearly infinite flexibility of well-trained human technicians. Ultimately, this facility makes about six times as many units with about nine times as many technicians as the alternator line.

Kontron’s custom-product shop in San Diego, CA, on the other hand, has a very high mix and low production volumes. The facility produces a few thousand units per month with about 50 employees and nary a robot in the place. In fact, the only automation in evidence was the burn-in ovens!

For this facility, factory automation makes little or no sense. By the time a robot could be programmed to performed any of the tasks for an order, a human technician would already have completed the whole thing. A human can read a work order, pick parts out of the stock room, carry them in a tote bin to a workbench, and modify a dozen circuit boards in the time it would take to plan out a conveyor system to automatically deliver the parts, let alone install it. In high-mix, low-volume situations, human flexibility is king!
Look for more about advanced automation in Japanese industry in the December 2007 issue of Control Engineering.

Posted by Charlie Masi on November 12, 2007 | Comments (0)