Compressor Failures Start Weeks Early | What to WatchNew Post

Compressor failures don't start at breakdown.

By the time a compressor fails, the problem's been building for weeks—sometimes months.

I've responded to enough unplanned shutdowns to see the pattern. The bearing didn't just decide to fail on Tuesday. The seal didn't randomly start leaking last night. Something changed earlier, and nobody caught it.

The failure you see is the end of a sequence. And if you know what to look for, you can catch it before the equipment shuts down.

Most compressor failures follow the same path:

1. Something changes in the operating conditions

Suction pressure drops. Discharge temperature creeps up. Flow rate decreases slightly. These aren't alarm conditions yet—they're just trending in the wrong direction.

2. The compressor starts working harder

It's trying to compensate for whatever changed. Bearing temperatures start rising. Vibration increases—still within spec, but higher than normal.

3. Secondary systems start showing stress

Lube oil temperature goes up. Seal leakage increases. Cooling water outlet temperature rises. These are symptoms, not causes.

4. The primary failure occurs

Bearing fails. Seal blows out. Rotor rubs. By this point, you're in shutdown mode and trying to figure out what happened.

Most condition monitoring focuses on thresholds. Is the vibration above 0.3 in/s (typical for some centrifugal compressors, though specs vary by equipment type)? Is the bearing temp above 180°F? Is the seal leaking more than five drops/minute?

You check your readings. Everything's in spec. You check the box.

But here's the problem: Your vibration went from 0.10 in/s to 0.18 in/s in two months. Your bearing temp climbed from 145°F to 165°F. Your seal went from 1 drop/minute to 4 drops/minute.

All still "in range." None triggering alarms. But every single one trending the wrong direction.

Thresholds tell you when something's already wrong. Trends tell you when something's starting to go wrong.

I don't wait for alarms. I look at trends over time:

Suction and discharge behavior — If suction pressure is dropping or discharge pressure is rising, something upstream or downstream changed. Find out what.

Temperature trends — I record actual bearing temperatures every PM, not just pass/fail. If the inboard bearing was 145°F in January, 150°F in February, and 155°F in March, that's a 10-degree climb in 60 days. Spec says 180°F max, so it's still "in range"—but something's changing. Same with lube oil temps. I track the numbers, not just the alarm status.

Vibration direction — If vibration is increasing—even if it's still low—I want to know why. Went from 0.08 in/s to 0.12 in/s to 0.16 in/s over three months? That's a pattern. Alignment shift? Bearing wear? Imbalance?

Performance data — Flow, pressure ratio, power consumption. If efficiency is dropping, the machine is working harder for the same output.

The key is comparing current data to baseline data. What's normal for this machine? What's changed?

I don't need fancy software for this. A spreadsheet with monthly readings is enough. Date, bearing temps (all positions), vibration (all bearings), suction/discharge pressures, flow rate, power draw. Compare this month to last month. Look for the climb.

I was on a commissioning project where we had a centrifugal compressor running smoothly—vibration low, temps normal, everything looked good.

But I was tracking discharge temperature against the calculated design value. The calculated temperature was 285°F. We were running 290°F consistently. Not enough to trip an alarm (setpoint was 310°F), but 5°F above what the thermodynamics said we should be.

We checked the intercooler. Turned out there was partial fouling on the tube side—cooling water flow was restricted. The compressor was compensating by running hotter.

If we'd waited for an alarm, that fouling would've gotten worse, discharge temps would've spiked, and we'd be looking at a much bigger problem.

Instead, we cleaned the intercooler, temps came back to normal, and we kept running.

Compressor failures don't happen suddenly. They give you warnings—you just have to know where to look.

Don't wait for alarms. Watch the trends. Compared to baseline. When something changes, figure out why.

The data's already there. Record the actual numbers. Track them over time. A 10-degree climb over 60 days tells you more than a single reading ever will.

What early warning signs have you learned to watch for? What tells you a compressor's heading for trouble before the alarms go off?

See you next week.