How to Diagnose a Failing Chiller Compressor Before It Seizes: Oil, Amps, Superheat and Sound

An air conditioning compressor seizure feels like an event. It almost never is. It is the last five seconds of a process that has usually been running for months, and that process leaves evidence in four places: the oil, the motor current, the refrigerant temperatures, and the noise the machine makes.

The other thing worth saying at the outset is that most compressor failures are not compressor failures. The compressor is where the damage shows up, but the cause is usually somewhere else in the system — a flooding expansion valve, a fouled condenser, a stuck crankcase heater, moisture that got in during a repair two years ago. Replace the compressor without finding that cause and you will be doing it again.

This is a guide to reading the four indicators, what thresholds actually mean something, and how to combine them into a diagnosis rather than a guess.

A note on safety. Everything described here involves pressurised refrigerant and live three-phase equipment. Insulation testing in particular requires the compressor to be isolated and all leads disconnected. This work belongs to trained, competent personnel with the right instruments. If you are reading this to decide whether to call someone, the answer is yes.

How compressors actually die

Three mechanisms account for the overwhelming majority of failures, and each one leaves a different signature.

Lubrication failure. The oil stops doing its job, metal contacts metal, and bearings or scroll surfaces destroy themselves. The usual route is liquid refrigerant returning to the compressor and diluting the oil, washing it off the load-bearing surfaces. Refrigeration training material describes this precisely: when refrigerant cannot boil off in the evaporator it works its way back to the compressor, the oil becomes diluted, and it gets washed off surfaces until there is metal on metal at the drive bearing and bushing in a scroll, or the main bearing in a semi-hermetic. The other route is overheating — oil that gets hot enough stops lubricating regardless of how much of it there is.

Electrical breakdown. Acid and moisture attack the motor winding insulation until it fails to ground. This is slow, measurable, and almost entirely preventable if anyone is looking.

Mechanical wear. Valve plates, rings, bearings and scroll tips wearing to the point where the machine cannot compress. This tends to show up as lost capacity long before it shows up as a bang.

All three announce themselves. Here is where.

Oil: the slowest signal and the most informative

Oil analysis is the closest thing to a blood test for a compressor. It is also the indicator most Malaysian plants never use, which is a shame, because it is cheap relative to a compressor.

What to measure

Total Acid Number (TAN). Measured in mg KOH per gram of oil, by ASTM D664. Acid forms from thermal breakdown of refrigerant and from moisture reacting with the oil. It is the single most direct measure of chemical degradation, and rising acid is what eventually eats the motor windings.

Thresholds depend on the oil type, which trips people up. Research on refrigeration lubricants established a critical value of 0.05 mg KOH/g for older CFC and HCFC systems running mineral oil. Polyolester (POE) oils used with HFC refrigerants naturally sit higher, typically in the 0.1 to 0.3 range, so applying the mineral-oil threshold to a POE machine will have you changing oil that is perfectly healthy. General industry guidance is to change refrigeration oil when TAN reaches roughly 0.1 to 0.2 mg KOH/g, but the number that matters is the one relative to your own baseline.

Moisture. Measured by Karl Fischer titration, ASTM D6304. POE oil is hygroscopic — it pulls water out of the air aggressively — and tolerates only around 200 to 300 ppm. Guidance commonly puts the change point at 100 to 200 ppm. Moisture matters because it drives acid formation, so a moisture result is an early warning for a TAN problem that has not arrived yet.

Wear metals. Spectrographic analysis per ASTM D6595 reports metals in parts per million, and the metal tells you which part is wearing. Iron points at bearings and valve plates. Copper points at bearings, and at copper plating on steel surfaces caused by acid attack. Aluminium points at pistons or scroll components. A sudden rise in any of them is more meaningful than any absolute number.

Viscosity. Tested at 40°C and 100°C. A change beyond about ±10% from baseline warrants investigation. Falling viscosity in a POE system usually means refrigerant dilution, which loops straight back to the floodback problem.

Colour. The cheapest test there is. Good refrigeration oil is clear or nearly so, and goes pale yellow with normal service. Dark brown or black means it has been overheated. If you draw a sample and it looks like used engine oil, you have already lost time.

How to sample

Take the sample from the same point every time, with the machine at normal operating temperature and having run for at least half an hour. Purge a little oil through before catching the sample so you are not testing what has been sitting in the valve. Cap the container immediately — POE will start absorbing atmospheric moisture within minutes and skew your own result.

Sample quarterly on critical machines, at minimum annually on everything else, and always after any event that opened the system.

Trend it. One oil analysis tells you very little. Four of them tell you where the machine is going.

Amps: useful, but only if you go past the clamp meter

Reading running current against RLA is the standard first check, and it is worth doing, but it is a blunt instrument. Current tells you what the motor is drawing, not what the machine is doing.

Two refinements make it far more useful.

Compare phases. Significant current imbalance across the three phases points at supply problems or developing winding faults. Check the supply voltage balance first — an unbalanced supply will produce unbalanced current in a perfectly healthy motor, and fixing the compressor will not fix the incoming supply.

Watch for low amps, not just high. A compressor drawing noticeably less than expected while failing to hold suction pressure is usually losing compression — worn rings, leaking valve plates, worn scroll tips. It is doing less work because it has stopped pumping properly. High amps get investigated because they trip things. Low amps quietly cost you capacity and get blamed on the chiller being “undersized.”

Winding resistance

Measure DC resistance across the three winding pairs. On a healthy motor they should be near identical. AHRI guidance flags an imbalance above 5% as a potential problem, and Trane’s guidance for their machines is that DC winding resistances should be within 5 to 10% of one another, typically ranging from around 0.2 to 2.5 ohms. Experienced rebuilders work tighter, flagging anything more than 2 to 3% apart for investigation.

Compare against the manufacturer’s specification sheet for the model, not against a general rule.

Insulation resistance, and the threshold everyone gets wrong

Megohm testing is the best early warning of electrical failure available, and the thresholds published for ordinary motors do not apply to compressors.

IEEE 43 gives a minimum of 1 megohm plus 1 megohm per kilovolt of operating voltage — for a 415V Malaysian supply, about 1.4 megohms. But that standard is explicitly for motors that are not hermetically sealed in oil and refrigerant. A motor sitting in a refrigerant and oil environment behaves differently, and manufacturers publish their own figures. Trane’s guidance for hermetic compressor motors is at least 1,000 ohms per applied volt, which on 415V works out to roughly 0.42 megohms as an absolute floor.

That floor is not the number to manage to. Insulation systems made since the mid-1970s should read far higher — commonly in the thousands of megohms — and guidance from test equipment manufacturers treats readings below 100 megohms on modern insulation as potentially unacceptable regardless of surface condition. Good windings should read from 100 megohms up to effectively infinite.

The signal is the trend, not the number. A compressor that read off-scale for six years and now reads 40 megohms is telling you something urgent, even though 40 is comfortably above any published floor. Log every reading with the date and the winding temperature.

Two practical points on method. Test at 500 VDC with all leads disconnected from the compressor terminals, and make sure the probe touches bare metal on the shell — paint will insulate and give you a falsely reassuring reading. And run the test twice: once with the compressor cold after being off for a while, which is more likely to expose contamination in the oil and refrigerant, and again after five to ten minutes of running, which drives refrigerant and oil off the windings and gives a truer picture of the insulation itself.

Superheat: where the damage is usually being caused

If oil analysis tells you the compressor is dying, superheat usually tells you why.

Suction superheat

Measure it at the compressor, not just at the evaporator outlet. Those are two different numbers, and the gap between them tells you how much heat the suction line is picking up.

Low suction superheat is the most destructive condition on this list. It means liquid refrigerant is reaching the compressor. The oil gets diluted, gets washed off the bearings, and the machine grinds itself down over weeks. Causes include an overfeeding or hunting expansion valve, a bad TXV bulb mounting or lost charge, evaporator airflow or water flow problems, overcharge, and sudden load changes such as hot gas defrost or condenser fan cycling.

High suction superheat starves the compressor. Return gas arrives hot, motor cooling suffers on suction-cooled machines, and discharge temperature climbs. Causes include undercharge, a restricted liquid line or filter drier, a starving TXV, and a plugged suction filter.

Watch it dynamically, not once. A TXV that hunts — superheat swinging between 2K and 15K over a few minutes — is doing damage that a single spot reading will never reveal.

Off-cycle migration and flooded starts

Refrigerant migrates to the coldest point in a system when it sits idle, and in many installations that is the compressor. On start-up, the oil foams violently and the machine runs momentarily on almost no lubrication.

Check that crankcase heaters are actually working. Not that they are fitted, and not that they have supply — that they are warm. A failed crankcase heater is a cheap part that destroys expensive compressors, and it is one of the most common things we find on machines that failed “for no reason.”

Discharge temperature and discharge superheat

Discharge superheat is discharge temperature minus saturated condensing temperature, and it is the cleanest indicator of whether the compressor is being cooked.

Measure discharge line temperature about 150 mm from the discharge service valve. Field guidance for semi-hermetic compressors is to stay below 107°C (225°F), with some scroll compressors permitted up to around 127°C (260°F) — check the manufacturer’s figure for your model rather than assuming. Above the limit and the system starts failing from worn rings, acid formation and oil breakdown.

The margins are thinner than the numbers suggest, because the discharge valve itself runs considerably hotter than the line — commonly cited as around 40°C hotter. POE oil starts losing its ability to lubricate above roughly 149°C (300°F) and breaks down around 177°C (350°F). A discharge line at 110°C is not comfortably below a 149°C limit; it is already close to it where it matters.

High discharge temperature has four classic causes: high condensing pressure, low suction pressure, high compression ratio, and high compressor superheat.

Two quick checks that narrow it down fast:

Condensing temperature versus ambient. Convert discharge pressure to saturated condensing temperature and compare against the air entering the condenser. A difference beyond roughly 11 to 17°C (20 to 30°F) points at non-condensable gases in the system or restricted condenser airflow. Design differs between machines, so calibrate this against your own unit when it is known good.

Compression ratio. Absolute discharge pressure divided by absolute suction pressure. Copeland air-conditioning scrolls are cited as having a maximum around 11:1 before the floating seal unloads, with refrigeration scrolls considerably higher. Running persistently near the limit is a slow death sentence.

Sound: the indicator everyone has and nobody records

Every experienced plant technician can hear when a machine is wrong. Almost nobody writes it down, which means the knowledge walks out when they change jobs.

Fix that cheaply. Record thirty seconds of audio on a phone, at the same three or four positions, monthly. Name the files with the date. You will not need them until the day you do, and then a recording from eight months ago is worth more than any opinion.

What to listen for:

Irregular metallic knocking that comes and goes, often worse at start-up or after a load change, is liquid slugging. Cross-check suction superheat immediately.

A low rumble or growl that deepens under load suggests bearing wear. Correlate with iron and copper in the oil analysis.

Rhythmic ticking or clattering on a reciprocating machine points at valve problems. Expect it to come with reduced capacity, lower-than-expected amps and a low discharge temperature — the machine is not compressing properly.

A loud squeal or howl on a scroll at start-up, particularly after electrical work, is often reverse rotation from swapped phases. Scrolls do not pump backwards; they get damaged. Check phase sequence after any switchgear or supply work.

A change in the electrical hum or a new vibration through the frame deserves attention even if you cannot name it. On critical machines, proper vibration monitoring will find bearing degradation far earlier than an ear can.

Putting the four together

No single reading is diagnostic. The combination usually is.

The pattern worth internalising: oil tells you how much damage has accumulated, superheat tells you what is causing it, amps tell you how close the motor is to failing, and sound tells you which of them to check first.

Why this bites harder in Malaysian plants

High ambient eats your discharge temperature headroom. An air-cooled chiller in Malaysia rejects heat into 32 to 35°C air all year, so condensing pressure sits high permanently. A condenser coil that would be a minor efficiency nuisance in a temperate climate pushes discharge temperature toward the limit here. Coil cleaning is not cosmetic maintenance in this country — it is compressor protection.

Humidity punishes any open system. Every time a system is opened for repair, POE oil starts pulling moisture out of some of the most humid air on earth. Minimise open time, use fresh oil from sealed containers, pull a proper vacuum, and change the filter drier. Every time.

The ageing R-22 fleet changes the economics. Malaysia is working toward complete HCFC elimination by 2030 under Stage III of its phase-out plan. On an R-22 machine running mineral oil, a motor burnout is no longer just a compressor bill — it is a system clean-up plus a refrigerant recharge from a shrinking, increasingly expensive supply. Early detection is worth considerably more on these machines than on a modern HFC unit.

Dust, haze and industrial fallout foul condensers faster than the maintenance interval in the manual assumes. That interval was written for somewhere else.

Coastal air corrodes electrical terminations. At Port Klang, Pasir Gudang, Prai and similar locations, terminal box corrosion produces resistance, heat and phase imbalance. Inspect terminal plates for dirt, corrosion and moisture as a routine item, not just when something goes wrong.

A monitoring routine that actually gets done

The best programme is the one your team will still be running in a year.

Monthly: running amps on all three phases, suction and discharge pressures with saturated temperatures, suction superheat at the compressor, discharge line temperature, oil level and colour through the sight glass, thirty seconds of audio.

Quarterly: oil sample for TAN, moisture, wear metals and viscosity on critical machines. Insulation resistance test. Terminal box inspection.

Annually: winding resistance across all three pairs. Crankcase heater verification. Full performance check against the original selection data. Condenser coil clean before it needs it.

After any system opening: filter drier change, oil sample, acid test.

None of that requires exotic equipment. It requires someone writing the numbers down in the same place every month, and someone else looking at the trend.

The one thing worth remembering

If you take a single habit from this: check suction superheat at the compressor, and check that the crankcase heater is warm. Liquid getting into a compressor, whether while running or on start-up, causes more failures than everything else on this page combined, and both checks take minutes.

A compressor that seizes without warning is nearly always a compressor that nobody was measuring.