Refrigeration

Phantom Shutdowns: The Level Probe That Read 38% on a Column Well Past Half Full

By the operator · November 2, 2026

The worst kind of plant shutdown is the one you can’t explain. Everything trips, you get it running again, and when you go looking for the cause, the graphs look normal.

I run an underground freezer, a two-stage ammonia plant feeding a calcium chloride brine loop, with rooms held anywhere from about -15 to 8 °F. In the spring of 2025 the compressor room shut down on us and nobody could say why. This is how that one got tracked down, and why the first place I look now isn’t the compressors.

What we had to go on

After the shutdown, I went back through the cameras and the trend graphs with our refrigeration tech. Nothing pointed to a power problem. Nothing pointed to a compressor problem either. We could see that the plant had stopped, but not what told it to stop.

That’s the signature of a safety doing its job without telling anyone. Something tripped, the plant did exactly what it was wired to do, and the controls never put a clear alarm on the screen.

So we stopped looking at the machines and started looking at the safeties.

The column that lied

On one of our vessels, the liquid level is read two ways. A level probe sits in a column beside the vessel and sends a signal to the controls. At the top of that same column there’s a mechanical high-level float, a hard safety that shuts the system down if the liquid gets too high.

The probe was sending a signal in its normal range. It was reading about 38%. When I looked at the physical column itself, the real level was far above 50%.

So the controls thought the level was comfortable, while the actual level was much higher and unknown. Every so often it rose far enough to trip the float at the top of the column, and the float shut the plant down. From the screen it looked like a phantom. From the column it was obvious.

What we tried

Clean the column. We valved the column off and opened the bottom. Nothing came out, which told us something was blocking it. The plan was to have the column unclogged and the probe cleaned.
Buy some margin in the meantime. We lowered the probe’s setpoint and lengthened the time delay on the high-level cutoff. Our refrigeration contractors agreed with both changes before we made them.
Fix the alarm itself. This turned out to matter as much as anything. The controls weren’t set up for how the high level actually works here. There are two separate high-level points, the probe and the float, and the system wasn’t showing them that way. Once the controls contractor reconfigured it, both points showed correctly on the screen and the alarms were addressed correctly.

The part I didn’t expect

When the column was finally cleaned out, it was pretty clean. That was good news for the vessel, but it didn’t explain the high level.

What did seem to settle it was the change to the alarm delay. My honest read is that the delay change may be what actually solved the high-level trips. I can’t prove the original cause, and I’d rather say that than tell you a tidier story than the one we had.

What I’d check first next time

Compare the probe to the real level. If you have a sight column or gauge glass, look at it. A probe can send a perfectly normal-looking signal and still be wrong.
List every safety that can stop the plant. For each one, know whether it shows up on the screen as its own alarm, or whether it just stops things quietly.
Make sure each point alarms on its own. If two devices protect the same level, the controls should show two points, not one.
Check the timers on your safeties. Delays, setpoints and the units they’re set in. A delay that’s too short can trip on a brief swing. A delay that’s too long can let a real problem run. Change them only with your refrigeration contractor or engineer in the loop, and write down what you changed.
Write it in the log. Date, time, what tripped, what the probe said, what the column said. The next phantom is easier to catch when the last one is on paper. Our daily, weekly and monthly checklist has a place for it.

The bigger lesson is the one this plant keeps teaching me: when the machines look fine and the plant still stops, check the controls and the safeties before you start tearing into compressors.

Related: Restarting a Two-Stage Ammonia Plant After a Shutdown, on bringing it back up once you know why it stopped, 1:20 Isn’t 120, another lesson about what the controls think a setting means, and Retro-Commissioning an Ammonia Plant, which has the minutes-versus-seconds unloader timer story.

← All posts