How to Tell If a Compressor Is Failing: 3 Real Signals, 4 False Alarms

The short answer

Most compressors I've been called out to inspect weren't broken. The fault was upstream — the air intake, the cooling loop, or the power supply. I coordinate emergency service for industrial compressed-air systems, and I've handled 200+ rush calls in nine years, including same-day turnarounds for plants that were minutes from sending workers home. Roughly 8 in 10 of those 'compressor failure' calls traced back to something other than the compressor block itself.

If a unit trips, overheats, or shuts down under load, bet on a 4-to-1 chance the block is fine. The three signals that do point to real mechanical trouble are metal in the oil sample, sustained current draw above nameplate at stable load, and discharge temperature that keeps climbing when nothing else has changed. Everything else — the tapping, the rattle on startup, the pressure that won't hold — usually isn't the compressor.

Three signals that actually mean something

  1. Metal in the oil sample. On oil-flooded screws, a lab analysis runs $150–250 and catches bearing wear weeks before you'd hear anything (US lab pricing, January 2025 — verify current rates). On oil-free units there's no oil to sample, so you're watching vibration trend data instead. A sudden shift in the 1× running-speed amplitude, not the overall number, is the one I pay attention to.
  2. Current draw above nameplate at stable load. Clamp meter on the incoming line, not the controller display. If the machine is pulling 105–110% of its rated current at a load and ambient temperature you've run at for years, something inside is losing efficiency. Rotor clearance, bearing drag, or a failing motor — the current tells you there's a problem, the vibration analysis tells you where.
  3. Discharge temperature creeping at constant load. If exhaust temp goes from 78°C to 88°C over two weeks at the same 70% load and the same ambient, that's a real trend. Fouled cooler, worn clearances, or a bearing on its way out. High temp on its own at peak summer ambient is a different story — see the next section.

Four alarms that don't mean what you think

  • The click on startup. It's a contactor engaging, or a check valve seating. It's the most benign sound in the room and the one people panic about most.
  • Pressure won't hold overnight. Leaks, a clogged filter, or an inlet valve that's sticking open. I've found $40 of leaking fittings on systems where the owner was ready to replace the compressor.
  • High-temperature trip in July. Cooler fouling plus a room that's 8°C hotter than design. Clean the cooler, check the room ventilation, then look at the machine.
  • Unit won't load. Pressure band settings, insufficient storage, or demand that grew past the system's capacity. Not a mechanical fault until you've ruled those out.

Where I got this wrong

In March 2023, a long-time client called about a 75 kW oil-free screw unit making what he described as 'a slight tapping.' I'd serviced that machine for six years. I told him over the phone to keep running it until Monday. I knew I should pull an oil sample first — I even said the words 'what are the odds' out loud. Monday morning the rotor contacted the housing. The repair quote came in at around 60% of a new unit's price. That tap was a bearing starting to go, and a $180 oil analysis would have flagged it three weeks earlier. We now require an oil sample on every tapping complaint, no exceptions, even for clients I've known for a decade.

Dodged one the other way in 2021. Another client was two signatures away from ordering a replacement centrifugal unit — the machine had lost about 15% of its rated flow over four months, and everyone assumed the impeller was fouled beyond cleaning. Turned out the inlet guide vane actuator had a corroded linkage that wasn't opening fully. A $400 part and two hours of technician time. If I'd been less stubborn about checking the cheap thing first, that client would have spent six figures on a machine they didn't need.

Screw, turbo, blower, and the AC question

Different machines fail in different ways, and the diagnostic logic doesn't transfer cleanly.

On an Atlas Copco turbo compressor — the centrifugal range — the early warning isn't noise. It's performance decay showing up in the vibration spectrum as sidebands around the running-speed peak, plus a slow drop in flow at the same inlet conditions. Impeller fouling looks identical to bearing wear from the operator's chair. Only the spectrum separates them. By the time a centrifugal unit makes an audible noise, you're usually past the point of a cheap fix.

The same applies to an Atlas Copco blower in a process line. It's a simpler machine, but the failure path is gearbox oil condition and drive coupling wear, not the lobes themselves. Check the oil, check the coupling alignment, then look at the rotating assembly.

On the 'how to tell if ac compressor is bad' question — the HVAC logic doesn't transfer to industrial air at all, and it's worth saying that plainly because the search results blur the two together. An AC compressor lives inside a sealed refrigerant loop. The most common reason it 'isn't working' is a failed run capacitor, a burned contactor, or a low-pressure cutout triggered by a slow refrigerant leak. Actual mechanical seizure is rare. Measure the capacitor, check the high and low side pressures, and compare running amps to the nameplate RLA — those three steps rule out the majority of misdiagnosed 'bad compressors.'

Refrigerated dryers and dehumidifiers get the same treatment. When someone tells me the compressor is dead on a refrigerated dryer, I ask about condenser cleanliness and refrigerant charge before I ask about the compressor. In my experience a dirty condenser is the answer more often than a failed block.

Deep freezer and low-temperature systems are a separate discipline entirely. There, the compressor killer is oil return and liquid slugging, not discharge temperature. You diagnose those on superheat and oil return, not on the metrics that work for a plant air system.

When you shouldn't fix it

I have mixed feelings about telling clients to repair rather than replace. On one hand, a rebuild keeps capital spending flat and gets the machine back in days instead of months. On the other, I've watched people sink 60% of a new unit's price into a machine that failed again 14 months later because the root cause — undersized piping, a hot room, dirty intake air — never got addressed.

My rule of thumb: if the repair exceeds 50% of replacement cost and the machine is past 12 years old and the model is discontinued with long lead times on major components, recommend replacement. All three, not any one. I've seen plenty of 15-year-old machines that were worth a rebuild because the parts were on the shelf and the rest of the system was clean.

Two boundary conditions worth stating honestly. First, don't diagnose an oil-free rotor assembly by ear — the clearances are tight enough that you need vibration data or a teardown, and guessing wrong is expensive. Second, none of this applies cleanly across categories. Plant air compressors, turbo compressors, blowers, HVAC compressors, dehumidifiers, and deep freezer systems are six different diagnostic problems that happen to share a word. The search query that brought you here probably mixed them together. Pick the one that matches your equipment before you act on anything above.

Leave a Reply