It’s no secret that I’m skeptical of innovations that are labeled as “disruptive” or “game changing”. In manufacturing, there’s very little that is truly disruptive.
It takes decades to move new technologies from laboratory curiosity to the assembly line. Productivity and quality advantages are already well understood and priced into the market and manufacturers are reluctant to become early adopters of new technology for good reason: in most cases, there’s too much to lose in a new radical and unproven production technology.
Today, the Cloud computing community is pushing hard for an industrial Internet of Things (IoT) which promises to gather vast amounts of data from manufacturing processes to allow unprecedented control and analysis of everything from steel mills to chip fab operations. Will it pan out? We’ll see soon, but here’s a related and I think cautionary tale:
In the video above, I show a body control module from a Honda Accord. It came out of a relative’s Honda as part of a repair to fix non-functioning backup lamps. Now for the last half century or so reversing lights have been federally mandated.
Standard practice was to include a simple single-pole, single-throw (SPST) momentary contact switch into the transmission, which routed current from the fuse panel main bus to a couple of bulbs in the taillight clusters. Today, information from the transmission range selector assembly is sent to this processor which in turn commands current flow through the wiring harness to the bulbs.
Needlessly complex you say? Absolutely, but it’s typical of the way modern devices are engineered.
Computational capacity available today vastly exceeds the number of sensors and actuators that they control in most automotive and industrial applications, so it’s easy for today’s engineers, who are steeped in problem-solving through software, to pass every function through a computer. In this case, it works the same as the simple switch, which might be slightly cheaper by eliminating the switch itself. The cost is lower reliability and higher cost of repair when the unit fails.
When everything from CNC machine tools to the janitor’s mop contains a Wi-Fi enabled sensor suite and processors, will the systems be reliable enough to generate high confidence data?
There’s a long history of engineers rightfully questioning sensor reliability when the data isn’t familiar or expected. A single false reading can destroy a manager’s confidence in an entire sensor/detection system and it happens more often than you think.
Less complexity is the best way to fight the problem, along with self diagnostic capability to produce simple “go/no-go” self testing results.
Can we reliably operate thousands or perhaps tens of thousands of discrete sensors and processors in a factory environment? Maybe, but like a lot of production managers out there, you’ll have to show me.