The Inductive Sensor Retrofit That Cost Us a Shift — and the Checklist It Spawned

June 7, 2024, 11:43 p.m. My phone rang, and before I picked it up I knew roughly what I was going to hear. The operator on the other end said it anyway: “R-402 is showing both open and closed.”

The valve was physically closed. The DCS did not believe it, and the interlock logic would not let the batch proceed until I bypassed the position check. Again.

That valve had failed the same way three times in fourteen months. The first failure was moisture inside a mechanical limit switch housing. The second was a bent lever arm. The third was tonight. We were done replacing those switches one at a time.

I have worked in instrumentation since 2012, and I have made enough expensive mistakes to keep a checklist notebook. This sensor project went straight to the top of it.

Two projects, one bad assumption

During the same outage window, we were adding pressure monitoring to the pump-and-tank area. For that part of the project, I specified Emerson 3051S pressure transmitters. It was an Emerson brand decision, but not a blind one: the documentation is detailed, the support model is practical, and when we needed a straight answer about HART addressing, Emerson support sent a clear, specific reply. They answered a second question during commissioning, too. That project ran the way process work should run: quietly.

The sensor portion of the same job was the follow-up. The mechanical limit switches on 14 actuated valves were being converted to inductive sensors. The idea was reasonable: replace a lever-and-roller mechanism with something that had no exposed moving parts, no roller to jam, and no housing to fill with water.

The actuators already had an internal cam. Mounting an M18 inductive sensor to detect that cam at the end of travel was not an exotic application. It should have been a two-day job.

I ordered 20 M18, three-wire, 24 V dc, normally open inductive sensors — about $3,180 with mounting brackets. I checked the supply voltage, the thread size, and the cable length. The datasheet line that said “rated sensing distance: 4 mm” I treated as a checkbox. That was my mistake.

The sensors passed the bench test. We mounted them one by one, set them up, and ran the air stroke test. All 14 valves gave feedback. Four of them were borderline, but they passed well enough that we told ourselves we would revisit them after startup.

Then the process engineer adjusted the closing stops on four valves to get a tighter seat. It was a mundane field adjustment. It moved the cam by a few millimeters. And a few millimeters was exactly the margin we did not have.

The handheld multimeter said the sensor was fine

When the first two feedbacks began to drop, I did what most of us do: I reached for a handheld multimeter and started checking the electrical side. Supply voltage was stable at 24.1 V dc on every valve, connections were clean, and each sensor toggled normally when I passed a steel wrench in front of the face. I swapped sensors. Same result. I moved wires to a different digital input. Same result.

The sensors were not the problem. But it took me most of a shift to accept that, because the electrical checks all looked good.

Finally, I measured the actual gap between the sensor face and the actuator cam with a set of calipers. It was 4.8 mm at the close position. The datasheet rating of 4 mm means something, but it does not mean “switches reliably on any metal at 4 mm.” The standard sensing distance for an inductive sensor is based on a low-carbon steel target. Our actuator cam was 316 stainless steel, and with that target the practical detection range was much shorter. In our own bench test, the reliable trigger point was closer to 2 mm. We were asking these sensors to work from more than twice that distance.

The fix ended up being simple: shim the sensors closer, re-measure every gap with the calipers, reset the stroke stops in the right order, and re-test the feedback through the DCS. No sensor was defective. No digital input card was faulty. The expensive part was the false diagnosis and the missed production window. I put the cost at about $7,200 in labor and downtime when I wrote the post-incident note.

About the calipers, and where Starrett makes them

That night is also why I now keep a decent set of calipers in the instrument shop rather than borrowing one from the machine shop. If you have searched “where are Starrett calipers made” before buying a set for your own toolbox, you probably already know the answer is not a single clean line. L.S. Starrett was founded in Athol, Massachusetts, in 1880, and the company is still closely tied to that town. But the modern Starrett catalog is broader than the old “Made in USA” reputation, and not every product in it is built in the same place.

As with most tool brands today, the honest answer is to check the specific model and its country-of-origin marking before you buy. For our work, the more important lesson was not the brand on the jaw. It was that a caliper has to be trustworthy enough to make you believe the measurement instead of your assumption.

What I now check before an inductive sensor swap

Since that June night, the checklist I use for every position sensor retrofit includes a few steps I used to skip:

• Read the full datasheet, especially the part about target material and rated sensing distance. Do not assume the rated distance applies to every metal.
• Bench-test with the same metal as the actual target, or something close to it, and measure where the real switching point is.
• Measure and record the installed gap with calipers, not by eye.
• Re-check the gap after any stroke limiter or actuator adjustment. Inductive sensors are not as forgiving as mechanical cam followers.
• Use a handheld multimeter to prove the electrical side, but do not stop there when the mechanical side has not been verified.

The transmitter part of the project went smoothly because we respected its datasheet and its support process. The sensor part burned us because I treated a precision component like a commodity. The industry has changed — a lot of what we used to do with mechanical switches is now done with proximity sensors, digital inputs, and configurable logic. But the fundamentals have not changed. Read the specification. Measure what is actually in front of you. Document what you changed. That lesson cost me about $7,200, and now it is on the checklist where it belongs.