An alternator that has been standing still for months can be wetter than one that runs every week. The insulation resistance test is the cheapest check that stands between a stored machine and a first start that ends in a flashover.
Why a stopped machine is not a dry machine
Insulation does not age only while the set runs. In a standby plant room the winding cools overnight, warms when the heater cycles and passes through the dew point twice a day. Moisture films form on the end windings, dust and salt bind to them, and the resistance path to earth falls. The machine has not moved, but its margin has.
The causes are ordinary. An anti-condensation heater fails and nobody notices because the panel lamp still works. A louvre is left open through a sandstorm, a roof leak drips onto the enclosure, or a plant room has no dehumidification during a long shutdown. On a coastal site, salt mist settles on everything. Any of these can turn a healthy winding into a marginal one during storage.
Site design often uses 40 °C as the reference ambient for rating and cooling, and hot, humid conditions are covered in the climate article. The insulation test has its own reference: the winding temperature at the moment of measurement. A cool-morning reading and a post-run reading are not comparable until both are corrected to a common temperature.
The first start after storage is the worst moment to discover a moisture problem. The set is called by the ATS, the site is already on battery, and there is no second chance. Take the reading before the machine is needed, not after it fails.
What the meter measures, and what it does not
An insulation resistance test applies a DC voltage between a winding and earth, or between windings, and measures the leakage current. The result is a resistance in megohms. It describes the insulation system as installed, including surface paths across dirty or damp end windings, not only the slot insulation.
That is both the strength and the weakness of the test. It is sensitive enough to catch moisture, contamination and a damaged winding before energisation, and simple enough to repeat every year. It does not prove dielectric strength, cannot locate a fault, and a single acceptable reading does not prove the machine is healthy. The value comes from the trend, the measurement conditions and the comparison with the manufacturer’s data.
Several factors move the reading without any change in the insulation itself. Temperature is the largest: a warm winding reads lower than a cold one. Surface contamination can dominate a dry winding. Winding and cable capacitance make the reading climb after the test voltage is applied, which is why the reading is taken after a fixed interval. A test voltage that is too low, or an uncalibrated meter, produces a number that looks precise and means little.
| Factor | Effect on the reading | How to control it |
|---|---|---|
| Winding temperature | Falls as the winding warms | Record the temperature and correct to a common base using the manufacturer’s data |
| Moisture film | Falls sharply, often with a low polarization index | Dry out and retest; never average a wet reading with the last dry one |
| Dust, salt or oil on the surface | Falls, sometimes erratically | Clean and degrease before judging; repeat the test after cleaning |
| Winding and cable capacitance | Rises during the test, then settles | Read after the agreed interval and keep that interval the same each time |
| Test voltage and duration | Changes the stress and the charging behaviour | Use the voltage and duration in the manufacturer’s procedure |
| Instrument and leads | False high or unstable readings | Check calibration, use guarded leads, clean the terminals |
Isolation is part of the test
An insulation test on a connected machine is not a test of the machine; it is a test of everything connected to the winding. The AVR, the rotating rectifier and its diodes, the permanent-magnet exciter, surge capacitors, the neutral link, the outgoing cables and the control wiring can all be damaged by the test voltage or can distort the result. The machine must be stopped, locked off and proved dead before any lead is moved.
The practical sequence is short but unforgiving. Isolate the prime mover and the electrical connections, prove the terminals dead, discharge the winding, disconnect the electronic devices named in the manufacturer’s procedure, earth the frame, connect the megohmmeter, apply the voltage for the specified time, read and record, then discharge the winding again before reconnecting anything. The discharge step is not optional: a large winding stores energy like a capacitor, and a lead removed too early can bite.
Large machines may have an insulated bearing at one end to interrupt shaft currents. That insulation is part of the machine’s electrical design. A test that accidentally bridges it, or a frame connection that bypasses it, produces a reading that describes the test leads rather than the winding. Mark the connections before removal and restore them exactly.
| Test point | Isolate before testing | What a valid result shows |
|---|---|---|
| Stator winding to frame and earth | AVR, exciter, outgoing cables, neutral link | Condition of the main insulation to earth |
| Phase to phase | Same, with the phase links separated | Insulation between phases |
| Exciter and rotating rectifier circuit | Main winding and AVR | Condition of the excitation insulation |
| Bearing insulation, where fitted | Shaft earthing brush, if present | Whether the bearing isolation is intact |
| Space heater circuit | Heater disconnected from the winding | Heater insulation and continuity of the anti-condensation function |
| Instrument and leads | Not applicable | That the reading is not a lead or meter artefact |
Reading the number: trend, temperature and time
The first question is not whether the reading is above a limit, but whether it is comparable with the last one. Record the winding temperature, the ambient temperature, the humidity, the test voltage, the duration and the instrument’s calibration status. Without those, the number cannot be compared with the commissioning value or with the manufacturer’s curve.
Most specifications quote a minimum for a new or dried winding, often per kilovolt of rated voltage plus a fixed allowance, with a minimum polarization index. The figure that governs a machine is the one in the manufacturer’s data and the project specification, because it depends on the insulation class, winding type and test method. Treat a number quoted without those conditions as a starting point, not a pass mark.
The polarization index, the ten-minute reading divided by the one-minute reading, is a second view of the same insulation. A ratio close to one suggests that leakage is dominated by moisture or contamination rather than by absorption behaviour. A high, stable ratio with a high absolute value is reassuring. A high absolute value with a poor ratio deserves a retest after drying, not a celebration.
Temperature correction is the step most often skipped. A reading after an hour of running will be lower than a cold reading on the same machine, and a team that compares the two without correction will either condemn a healthy winding or accept a wet one. Use the manufacturer’s correction data, or test at a stable temperature, and state which method was used.
| Pattern | Most likely cause | Next action |
|---|---|---|
| High and stable across years | Healthy, dry insulation | Record and continue the normal interval |
| High but clearly below the last corrected value | Early moisture ingress or surface contamination | Inspect heaters and enclosure, clean, retest after drying |
| Low and falling as the winding warms | Moisture in the insulation | Dry out under supervision and retest at intervals |
| Low but stable, improves after cleaning | Surface contamination rather than bulk moisture | Clean, find the source of contamination, retest |
| Unstable or drifting during the test | Leads, guards, instrument or a surface path | Check the set-up before judging the winding |
| Below the project minimum with a low polarization index | Wet or damaged insulation | Do not energise; investigate, then dry or repair |
Dry-out: what works and what only hides the problem
Drying is not a single procedure. The right method depends on where the moisture is, how much there is, what the insulation can tolerate and what the site can support. The first step is to find the source. A winding dried and returned to a plant room with a failed heater and an open louvre will be wet again before the next inspection.
Anti-condensation heaters are the first line of defence for a standby machine, not a drying method for a wet one. They keep the winding a few degrees above the surrounding air so moisture does not condense on it. If the heater has failed, replacing it and waiting is not enough: the winding must be dried and retested. A heater that is energised but not heating is a common finding, and the only reliable check is the winding temperature, not the indicator lamp.
Forced warm air is gentle and even, and suits a machine that is only slightly damp. It needs a controlled temperature, a path for moist air to leave and a limit that the insulation and bearings can accept. Passing current through the winding, from a low-voltage supply or by short-circuiting the machine at reduced excitation, raises the temperature from inside and is faster, but needs a competent person, a temperature limit and continuous monitoring. A load-bank run warms the machine and proves the set, at the cost of fuel, noise and time.
That cost is easy to underestimate. At 1000 kVA the set burns roughly 200 litres per hour, so a 24-hour proving run is about 4800 litres. At 500 kVA the figures are 101 litres per hour at full load, 80 at 75 percent and 60 at 50 percent, with 0.24 to 0.29 litres per delivered kWh. A drying run at the 10 to 20 percent light-load band would use less fuel but is poor for the engine and slow to warm the winding, which is why load-bank drying is planned around a target winding temperature, not the cheapest operating point.
| Method | Best suited to | Main limit | How to monitor |
|---|---|---|---|
| Space heaters | Keeping a dry machine dry | Will not dry a wet winding | Winding and enclosure temperature |
| Forced warm air | Light surface moisture | Uneven if the air path is poor | Air and winding temperature, humidity |
| Current through the winding | Bulk moisture in the stator | Needs a controlled supply and supervision | Winding temperature and resistance |
| Load-bank run | Proving the set while warming it | Fuel, noise and time | Load, temperature and fuel used |
| Workshop oven or vacuum | Severely wet or rewound machines | The machine must be removed | Workshop records and a retest |
| Cleaning and degreasing | Contamination, not moisture | Does not remove moisture | Resistance before and after |
Whatever method is used, the test that ends the dry-out is the same test that started it. Measure at intervals, plot the reading against winding temperature, and stop when the corrected value stabilises at an acceptable level and the polarization index has recovered. A reading still climbing when the heater is switched off is not a pass; it was taken too early.
| Rating and load | Fuel rate | What the run is useful for |
|---|---|---|
| 500 kVA at full load | 101 L/h | A short, hot run that proves the set and raises winding temperature quickly |
| 500 kVA at 75 percent | 80 L/h | A loaded run with less stress on the plant than full load |
| 500 kVA at 50 percent | 60 L/h | Part-load proving while the winding warms more slowly |
| 1000 kVA at full load | About 200 L/h, about 4800 L over 24 hours | A long soak, planned only when the fuel and the site programme allow it |
| 500 kVA in the 10 to 20 percent band | Below the 50 percent figure | Not a drying method: slow warming and poor engine conditions |
When the machine should not be energised
There are conditions in which the correct action is to keep the set offline, tell the site and repair the cause. A winding below the project minimum, a polarization index that stays low after drying, visible moisture or water marks, a burning smell, a winding that has been flooded, a machine stored for a long period without heater power, or a test instrument that cannot be trusted are all hold points.
Energising to see whether the machine dries itself is not a diagnostic. The first start applies full voltage to the insulation, and if the winding is wet the failure can damage the alternator, the AVR and the switchgear, and can start a fire in the plant room. The UPS covers the seconds between mains loss and generator load; it does not cover a generator that fails after the transfer. In a data centre sized with a PUE of 1.4, the alternator may be carrying close to 1.4 times the IT load at the meter, so a failed machine takes the whole facility down, not just part of it.
Every hold point should have a named person who can release it. The decision is technical, but the record should show who made it, on what reading, at what temperature and with what correction. If the machine is released after drying, the record should show the before-and-after values and the method used.
| Condition | Action | Evidence to keep |
|---|---|---|
| Reading below the project minimum | Hold; investigate and dry | Corrected reading, temperature and method |
| Low polarization index after drying | Hold; repeat drying or repair | Index before and after, drying curve |
| Visible moisture, water marks or flooding | Hold; inspect and repair | Photographs in the file, inspection note |
| Failed space heater with condensation | Replace the heater, dry and retest | Heater test and winding temperature |
| Uncalibrated or damaged instrument | Hold until a checked instrument is used | Calibration record |
| Acceptable corrected reading and stable index | Release for service | Signed release and retest values |
Records, intervals and the annual programme
The interval is set by the manufacturer’s data and the project specification, not one universal rule. Most sites tie the test to the annual load test at 50 percent to 100 percent of rating, add a test before the first start after long storage, and retest after water ingress, a fault, a rewind or a long outage. A machine held at 10 to 20 percent load for long periods deserves a closer watch, because low-load running and a cool winding suit moisture.
The record should be usable by someone who was not present: machine identification, date, ambient temperature and humidity, winding temperature and how it was measured, test voltage, duration, one-minute and ten-minute readings, polarization index, corrected value, instrument identification, isolation steps, drying method if any, and the release decision. A photograph of the terminal box is useful; a number without conditions is not.
Keep this test separate from the earthing test. An earth electrode measurement proves the path to earth; an insulation resistance measurement proves the winding is isolated from earth. Both are needed and neither substitutes for the other, as the electrode verification article explains. The same separation applies to the load test: it proves the machine can carry load, not that its insulation is dry.
For a new or newly installed machine, the commissioning record is the baseline. The first reading after storage should be compared with that baseline, not with a generic table. If the machine has been moved, stored outdoors, shipped in a container or left without heater power, treat the first reading as a survey, not as an acceptance test.
A final point about the heater circuit. The heater is part of the protection scheme, and its failure is a finding in its own right. Record the heater current or the winding temperature rise above ambient when the heater is on, and check it at the same visit as the insulation test. A machine with a good reading and a dead heater is a machine that is already on its way back to a wet winding.
Check the winding before the standby period ends
A stored alternator is only as reliable as its last insulation reading. Send the machine rating, the storage conditions, the last test values and the heater arrangement. The review will set out the isolation steps, the readings to take, the drying options and the hold points before the set returns to service.
What insulation resistance should a standby alternator have?
No single figure applies to every machine. The minimum is set by the manufacturer’s data and the project specification and depends on the insulation class, winding type and test method. Judge the corrected reading against that figure and the machine’s own history, not a table copied from another site.
Can the insulation test be done with the AVR connected?
No. The test voltage can damage the AVR, rotating rectifier, exciter and other electronic devices, and the connected circuit distorts the reading. Isolate the devices named in the manufacturer’s procedure, test the winding, discharge it and restore every connection as marked.
How long does dry-out take?
It depends on how much moisture is present, where it sits and which method is used. Light surface dampness may clear in hours with warm air; a flooded or heavily contaminated winding may need days in a workshop oven or controlled current drying. The end point is a stable corrected reading and a recovered polarization index, not a fixed clock time.
Is a low reading always caused by moisture?
No. Surface dust, salt, oil, a damp terminal box, long or unguarded leads, a cold winding and an uncalibrated instrument can all produce a low or unstable result. Clean the surfaces, check the set-up and repeat before concluding that the insulation is wet.
How often should the test be repeated?
The interval comes from the manufacturer’s data and the project specification. Most sites tie it to the annual load test at 50 percent to 100 percent of rating and add a test before the first start after long storage or after water ingress, a fault, a rewind or a long outage. A failed heater is also a reason to retest.
Does a space heater dry a wet winding?
No. A space heater keeps a dry winding a few degrees above the surrounding air so moisture does not condense on it. Once the winding is wet, the heater is not enough: moisture must be driven out by a controlled drying method and proved by a new insulation test. A heater that is on but not heating is a common cause of repeat moisture problems.
KVA Grid sales engineering desk. Figures are published elsewhere on this site: the annual load test at 50 % to 100 % of rating and the 10 % to 20 % light-load band from the maintenance schedule; 0.24 to 0.29 litres per delivered kWh and the 500 kVA figures from fuel consumption; the 1000 kVA example, UPS role and PUE 1.4 from data center backup power; 40 °C ambient from hot and humid climates; load-bank practice from load-bank testing; earthing from electrode verification; and handover records from the commissioning checklist. Insulation-resistance minima, temperature correction, dry-out temperatures and test intervals are set by the manufacturer’s data and the project specification. Reviewed 7 November 2026.
