Most stator failures are not sudden. They are the end point of a process that has been running for months or years — visible if you know what to look for, detectable by testing before the insulation finally fails in service.
Understanding the failure mechanism behind a bad test result is the difference between cleaning a winding and rewinding it. Between scheduling an outage next month and pulling the machine this weekend. This article maps the main stator failure mechanisms to their root causes, visual symptoms, and test signatures, so that a low IR reading or an unusual PI can be traced to something specific rather than written off as “the winding is deteriorating.”
The mechanisms covered here apply to gradual aging processes — the kind that develop in service over time. Catastrophic events like out-of-phase synchronisation, foreign objects ingested into the machine, or incorrect winding connections are outside this scope.
Table of Contents
1. Thermal Deterioration
Who it affects: Both form-wound and random-wound stators. Most common in air-cooled machines. Unlikely in direct water-cooled or hydrogen-cooled windings where cooling is maintained.
What happens: Insulation is a polymer. At elevated temperatures, the chemical bonds within the organic components of the insulation undergo scission — they break under thermal stress. Oxygen from the surrounding air attacks the broken bonds. The result is a shorter, weaker polymer chain. Macroscopically, the insulation becomes more brittle, loses mechanical strength, and loses its ability to bond the tape layers together.
In random-wound stators, brittle insulation cracks when the copper conductors move under magnetic forces during starting or operation. Cracked or peeled magnet wire insulation leads to turn shorts. A turn short creates a local hotspot, melts the copper, and destroys the groundwall within seconds to minutes of occurring.
In form-wound stators, the bond between the mica tape layers begins to fail. The result is delamination — the groundwall insulation separates into distinct layers with air gaps between them. Once delamination occurs, two failure paths open: the conductors vibrate relative to each other and abrade the strand or turn insulation; or in machines rated above 3 kV, partial discharge (PD) initiates in the air gaps within the delaminated insulation and erodes a path through to ground.
In older thermoplastic (asphaltic) stators, thermal aging causes the asphalt binder to soften and flow out of the groundwall once temperatures exceed approximately 70–100°C. This leaves the tape layers insufficiently bonded and the insulation physically weakened.
The Montsinger rule applies: Every 10°C rise in operating temperature approximately halves the thermal life of the insulation. A Class F (155°C) winding running continuously at its rated temperature has a predicted average life of about 20,000 hours at that temperature — just over two years. Running it at 120°C instead of 155°C dramatically extends that life. Running it at 165°C halves it again.
Root causes: Overloading; poor ventilation or blocked cooling passages; dirty heat exchangers; inadequate winding design; strand shorts causing circulating currents; too many thermal cycling events from frequent starts; operating synchronous machines under-excited (causing axial flux in the core ends).
Visual symptoms: Random-wound stators show cracked or peeling magnet wire insulation, discoloured or brittle slot liners. Thermoplastic form-wound stators have puffy insulation that sounds hollow when tapped; asphalt may be visible oozing from the groundwall. Thermoset form-wound stators also sound hollow when tapped, but only after severe deterioration. Surface scorching on indirectly cooled coils appears only at very advanced stages.
Test signatures: In random-wound stators, IR may be low if magnet wire insulation has cracked or peeled. In form-wound stators, thermal deterioration is largely invisible to IR and PI testing in modern epoxy-mica windings until cracking is severe — delamination within the groundwall does not create a conduction path through intact mica layers. Power factor and capacitance tip-up testing (see Article 05) is more sensitive to bulk thermal deterioration. Elevated PD activity may be detectable once delamination has progressed sufficiently.
Remedies: Thermal deterioration is not reversible. The only actions available are to slow the rate of progression — cleaning to restore airflow, fixing cooling system deficiencies, reducing load, improving ventilation — and planning for a rewind when degradation is widespread.
2. Thermal Cycling
Who it affects: Form-wound stators in machines subject to rapid load changes — peaking generators, pump storage machines, large air-cooled gas turbine generators. Random-wound stators are very unlikely to experience this failure mode.
What happens: When load increases rapidly, the copper temperature rises quickly due to I²R losses. The groundwall insulation and stator core respond more slowly. This temperature differential creates axial shear stress between the copper conductors and the groundwall.
In thermoplastic (asphaltic) windings, repeated thermal cycles cause the tapes to separate in the regions just outside the slots — a phenomenon called girth cracking. The crack eventually penetrates the full depth of the groundwall. With enough cycles, the bond between copper and groundwall fails and the copper is free to move.
In modern thermoset (epoxy-mica) windings, the same differential expansion creates an axial shear stress at the copper-to-groundwall bond. Epoxy is more rigid than asphalt, which means it resists shear well but fatigue-cracks rather than flowing under repeated stress. After many thermal cycles, the bond between copper and groundwall fails, creating gaps in which PD can occur at operating voltages above about 3 kV.
In global VPI (GVPI) stators, where the coil is bonded to the core as well as to the copper, the same mechanism can also shear the groundwall away from the core surface after many cycles.
Root causes: Too-rapid load changes relative to what the insulation system was designed to handle; operation at high stator winding temperature, which reduces bonding strength; inadequate insulation system design for cyclic operation.
Visual symptoms: In asphaltic windings, circumferential cracks in the groundwall just outside the slots and puffy insulation at ventilation ducts. In thermoset windings, the groundwall may sound hollow when tapped at the slot exits. Cross-section examination shows gaps between copper and groundwall.
Test signatures: Elevated PD in form-wound windings at high voltage ratings. PI may remain apparently normal until cracking is advanced enough to create surface paths. Power factor tip-up and offline PD testing are more sensitive to this mechanism than IR/PI.
Remedies: Reduce the rate of load change where the operating regime permits. Reduce maximum operating temperature. Rewinding with an insulation system specifically designed for cyclic duty may be necessary for machines in peaking service.
3. Poor Impregnation
Who it affects: Both random-wound and form-wound stators. In random-wound stators, the problem is visible and often causes early failures. In global VPI form-wound stators, it is difficult to detect during manufacture and can cause failure in as little as 2 years for windings rated 6 kV and above.
What happens: The purpose of impregnation — whether dip-and-bake, resin-rich tape, or vacuum pressure impregnation (VPI) — is to fill all voids within the winding with a solid dielectric material. Air pockets within the groundwall or between conductors support PD at operating voltages. PD in air erodes organic insulation materials. In random-wound stators fed from PWM inverters, this failure can progress to ground fault within months.
In form-wound stators with mica-based groundwall insulation, the mica is highly resistant to PD, so the progression is slower — but inadequate impregnation can still cause failure in 2 to 5 years in machines rated above 6 kV. In the slot, where PD in voids has nowhere to discharge along the surface, the attack is concentrated and can penetrate the groundwall relatively quickly.
Root causes (GVPI stators): Inadequate impregnating pressure or time; incorrect resin viscosity; stator not rotated during cure; moving stator from impregnation tank to curing oven too slowly, allowing resin to drain.
Root causes (resin-rich coils): Old or improperly stored tapes that have partially pre-cured; poor taping technique leaving air wrinkles; inadequate pressure during cure.
Visual symptoms: In random-wound stators, poor impregnation is visible — the interstices between conductors are not filled, there is no resin sheen. In form-wound stators, the problem is typically adjacent to the copper conductors and invisible from the surface. Tapping the coil may sound hollow in severe cases. The PD test is the most reliable detection method for form-wound stators.
Test signatures: IR may be low in random-wound stators. In form-wound stators, IR and PI are not reliable indicators of poor impregnation unless cracking has progressed to a surface fault. Offline PD testing at rated voltage is the most sensitive test. Capacitance tip-up and power factor testing can also detect widespread void content.
Remedies: Poor impregnation in form-wound coils cannot be repaired by injection alone — the voids adjacent to the copper conductors cannot be reached from the outside. The only permanent fix is rewind with a better-controlled impregnation process. Prevention requires manufacturer process controls: monitoring temperature, pressure, time, and winding capacitance during the VPI process.
4. Loose Coils in the Slot
Who it affects: Form-wound stators using conventionally manufactured (non-GVPI) thermoset coils and bars. Most likely in large gas and steam turbine generators and hydrogenerators. The GVPI process normally prevents this, but poorly impregnated GVPI stators or those subject to severe thermal cycling can also experience loose coils.
What happens: Stator bars and coils experience magnetic forces at twice the power frequency — 100 or 120 Hz — proportional to the square of the current. A bar that fits tightly in the slot resists these forces. A bar that is slightly loose vibrates. Vibration causes two failure processes in sequence.
In Stage 1, the semiconductive coating on the bar surface (which exists on stators rated 6 kV and above to control the electric field at the slot exit) abrades against the grounded stator core. The coating makes and breaks contact with the core 100 times per second, causing small contact sparks — Stage 1 slot discharge. This is relatively benign and can continue for years without groundwall failure.
In Stage 2, the semiconductive coating has been abraded away at the contact points. The groundwall insulation surface is now exposed and not grounded, even through the coating. Classic PD occurs across the air gap between the core and the bare groundwall surface. PD in air-cooled machines generates ozone, which combines with moisture to form nitric acid. The acid attacks the organic components of the insulation, the slot wedges, and adjacent materials — accelerating the coating deterioration and potentially loosening the coil further. Stage 2 can progress to groundwall failure in as little as 2 years in machines with high bar forces.
Root causes: Initial loose fit in the slot due to manufacturing tolerances, shrinkage of packing materials over time, or oil contamination that lubricates the coil-core interface and allows movement to occur. Loose or degraded slot wedges. Use of magnetic wedges that shrink over time and eventually crumble.
Visual symptoms: White deposits (eroded coating) on the bar surface in the slot, visible through radial vent ducts. Loose wedges detectable by tapping — a ringing sound versus the solid thud of a tight wedge. Carbon or white powder deposits around slot exits. Abrasion marks on bar sides from contact with slot walls or ripple springs.
Test signatures: Elevated PD, especially if Stage 2 has begun. IR and PI tests are not sensitive to this mechanism in early stages. Wedge tightness assessment (physical tapping) and side clearance measurement are the primary diagnostic tools. Offline and online PD testing provide the most reliable electrical indication.
Remedies (Stage 1): Re-wedging with replacement side packing and wedges; replacement of flat packing with ripple springs; injection of graphite-loaded paint or epoxy into the slots to restore contact between the bar and core. Stage 1 detection and remediation can restore the winding close to new condition.
Remedies (Stage 2): The same mechanical remedies apply, but the semiconductive coating cannot be fully restored by injection — there will always be gaps in coverage. Slot injection slows the failure process significantly but does not stop it. Periodic monitoring is essential to track progression rate.
5. Semiconductive Coating Failure
Who it affects: Form-wound stators rated 6 kV and above energised from the power frequency supply. Also inverter-fed (IFD) motors rated 3 kV and above. More likely in air-cooled machines and those operating at altitude (above 1000 m), where air breakdown strength is reduced.
What happens: The semiconductive coating on the stator bar in the slot gradually loses conductivity through oxidation of the graphite particles. The coating becomes resistive or non-conductive in localised areas. In these areas, the bar surface is no longer grounded, and PD occurs between the bare groundwall insulation surface and the stator core — even if the bar is not vibrating and is still firmly held in the slot.
The PD generates ozone in air-cooled machines. Ozone combines with nitrogen and moisture to produce nitric acid, which attacks the organic materials in the slot — further degrading the coating, attacking the slot wedges, and potentially loosening a winding that was initially tight.
In IFD motors, the switching frequency (typically 1 kHz or more) drives much higher capacitive currents through the coating than at power frequency. This increases resistive heating (I²R losses) in the coating by a factor of 20 or more, dramatically accelerating the oxidation process.
Root causes: Insufficient graphite particle density in the original coating — a manufacturing quality issue. Air pockets just below the coating surface, allowing PD to attack the coating from underneath. In IFD motors, the combination of high capacitive currents and the elevated frequency creates thermal conditions the coating was not designed to withstand.
Visual symptoms: The coating turns white and may disappear entirely in affected areas. This is most visible on coils connected to the phase terminal — neutral-end coils see much lower voltage and are unlikely to be affected. The change in appearance is typically visible in radial vent ducts with the rotor removed, or by direct inspection if the wedges are removed. Elevated ozone levels in enclosed air-cooled machines.
Test signatures: Elevated PD in the affected area. Offline PD testing can detect the problem before failure. Online PD monitoring (Section 16) may also show increasing activity over time.
Remedies: If the coils can be removed without damage, stripping and reapplying the coating is the definitive repair. In-situ repair by injecting graphite-loaded varnish, silicone rubber, or epoxy compounds into the slot is possible but cannot reach all affected areas — particularly the bottom edge of the coil. Re-wedging to ensure the coil remains tight in the slot should accompany any injection repair. Reversing the winding connections (placing previously phase-end coils at the neutral) can extend life by reducing the electrical stress on the most severely affected coils.
6. Grading Coating Overlap Failure
Who it affects: Form-wound stators rated 6 kV and above that have both a semiconductive slot coating and a silicon carbide (SiC) stress control coating on the endwinding. The two coatings must overlap at the slot exit to maintain electrical continuity. When that overlap degrades, discharging begins at the overlap region.
What happens: The silicon carbide coating is grounded through its overlap with the semiconductive coating at the slot exit. Capacitive currents from the high-voltage copper flow axially through the SiC coating and discharge to ground through the overlap. If the overlap resistance is too high, I²R losses at the overlap region heat the material, causing oxidation. As the resistance increases, the heating increases — a self-reinforcing process. Eventually the overlap becomes non-conductive and large discharges occur parallel to the insulation surface at that point.
The rate of deterioration of the groundwall from this process is relatively slow — discharges parallel to the insulation surface are less aggressive than discharges perpendicular to it. On its own, this mechanism is unlikely to cause groundwall failure within 20 years. However, it generates ozone, is a reliable indicator that the semiconductive coating within the slot may also be failing, and creates PD signatures that can mask other more serious failure modes.
Visual symptoms: A white band around the coil circumference a few centimetres outside of the slot exit. Appears only on phase-end coils, not neutral-end coils. Easily identified in visual inspection.
Test signatures: PD testing shows activity associated with the overlap region. The characteristic PD signature is different from groundwall void discharge or loose coil PD. Blackout testing (UV inspection with lights off and machine energised) can visualise the discharging.
Remedies: Permanent repair requires removing the coils and reapplying the stress control coatings. In-situ repair by applying new partly conductive paint over the overlap region can restore the ground connection — but only if the endwinding has not been overcoated with insulating varnish. Overcoating with insulating varnish (a common aesthetic choice) will cause the white band to reappear within months and does not address the underlying problem.
7. Transient Voltage Surges
Who it affects: All stator winding types. The groundwall insulation in all machines; the turn insulation specifically in multiturn coil windings. Random-wound stators are most vulnerable to turn insulation damage from short-rise-time surges.
What happens: Voltage surges from lightning, power system ground faults, or motor switching create transient overvoltages that may exceed the breakdown strength of the insulation. At power frequency, voltage distributes evenly across turns in proportion to the number of turns. At high frequencies — created by fast-rising transients with rise times of 100 ns or less — the distribution becomes highly non-linear, with as much as 40% of the applied surge voltage appearing across the first turn in the coil connected to the phase terminal.
For random-wound stators, this means that every motor switching event creates a brief but very high voltage across the turn insulation of the first few turns at the phase terminal. If the turn insulation has been weakened by prior aging (thermal, abrasion, PD), or if the magnet wire insulation has pinholes from manufacturing, the surge may puncture the turn insulation. A turn short then develops into a ground fault within seconds.
Stone et al. note that a single surge event either causes immediate failure or has no lasting effect — transient voltage surges do not gradually deteriorate insulation that is otherwise intact. A winding that is in good condition will survive switching surges. A winding that has been aged by other mechanisms may fail when a surge that would previously have been withstood exceeds the reduced breakdown strength.
Root causes: Power system disturbances (unavoidable in most applications). Motor switching events, especially with vacuum or SF6 breakers that can produce multiple transients per switching operation. Cable lengths above approximately 30 m create travelling wave effects that can double the surge voltage at the motor terminals.
Visual symptoms: Ground faults from surge damage typically show melted copper at the fault location — the result of the high currents that flow after a turn short develops into a ground fault. Phase-to-phase faults from surges produce extensive damage and usually leave visible destruction of coils across multiple phases.
Test signatures: A surge comparison test (Section 15.16 in Stone et al.) is the most effective offline test for detecting weakened turn insulation before a fault. Standard IR and PI tests do not detect turn insulation problems unless the fault has already developed to a ground path.
Remedies: Surge arrestors at the switchgear or transformer protect against lightning and slow-rise-time power system transients. Surge capacitors (approximately 0.2 µF) at the motor terminals limit the voltage rise across turn insulation by lengthening the surge rise time. Both should be used for critical motors or generators where turn insulation damage from switching is a known risk. For motors that must be switched frequently, designing the winding with heavier turn insulation is the most reliable approach.
8. Repetitive Voltage Surges from Inverter Drives (VFD/IFD)
Who it affects: Random-wound motors rated 400 V to 1000 V driven by PWM voltage source inverters with IGBT switching devices. Also form-wound motors rated 3 kV and above driven by the same type of drive, where the stress control coatings are the primary vulnerability.
What happens: PWM inverters switch a DC bus voltage at frequencies typically between 1 and 20 kHz. Each switching event sends a rectangular voltage pulse down the power cable to the motor. If the cable is longer than approximately 30 m, the mismatch between the cable surge impedance and the motor surge impedance creates a reflection that can double the pulse voltage at the motor terminals.
The resulting fast-rise-time pulses (100 ns or less) create a non-uniform voltage distribution across the turn insulation — the same mechanism as single switching surges, but repeated thousands of times per second. In random-wound motors with conventional organic magnet wire insulation, PD occurs in air pockets adjacent to the conductors. Each discharge slightly erodes the magnet wire insulation. The cumulative damage from thousands of discharge events per second can cause turn-to-turn failure within weeks or months of commissioning.
Modern epoxy-mica turn insulation in form-wound windings is highly PD-resistant, making turn insulation failure from this mechanism less likely. However, the high-frequency capacitive currents accelerate deterioration of the semiconductive and SiC grading coatings — the failure modes described in Mechanisms 5 and 6 above occur much faster on IFD motors.
Root causes: PWM inverter with IGBT devices producing surges with rise times below 200 ns; cable lengths creating travelling wave voltage doubling; inadequate motor insulation system for inverter service; absence of output filters on the drive.
Visual symptoms: In random-wound stators, whitish powder on magnet wire in the phase-end coils where PD is occurring. In form-wound stators, white discolouration of the stress control coatings at slot exits, and elevated ozone in enclosed machines.
Test signatures: Offline PD measurement during normal inverter operation can detect the problem. Online PD monitoring methods designed to reject the switching transients from the inverter are available. Surge voltage measurement at the motor terminals using a high-bandwidth oscilloscope is the definitive way to assess the severity of the surge environment before failures occur.
Remedies: Measure the actual surge environment at the motor terminals before assuming the worst. Many installations do not produce severe surges even with IGBT drives. If surges exceed the limits in IEC 60034-18-41, the options include: output filter on the drive (most effective); change cable length or type; PD-resistant magnet wire in random-wound motors; full VPI impregnation to eliminate air pockets; increase separation between phase-end turns. For form-wound motors in IFD service, the winding should be qualified to IEC 60034-18-42, or the insulation voltage class stepped up by one level.
9. Contamination and Electrical Tracking
Who it affects: Any stator winding in any machine exposed to contaminated cooling air. Higher-voltage machines are more susceptible, but tracking can cause failures even at 120 V in dirty environments.
What happens: Contamination in the cooling air — oil mist, carbon dust from brushes, fly ash, cement dust, salt, chemical byproducts from industrial processes — deposits on the winding surfaces. Most contaminants are insulating when dry but become partially conductive when combined with moisture.
In form-wound stators, the contaminated end winding surfaces create a leakage path between coils at different phase voltages. The current is initially very small — nanoampere range. But contaminated surfaces are not uniform. Dry bands have much higher resistance than the surrounding contamination. The full voltage that would otherwise be distributed across the entire contaminated path appears instead across the small dry band, causing localised breakdown of the air or hydrogen. The discharge carbonises the underlying insulation material in that spot. That spot is now permanently conductive. The electrical stress transfers to the next dry band. Over time, a carbonised tracking path grows between adjacent coils in different phases.
In random-wound stators, contamination interacts with any pre-existing pinholes or cracks in the magnet wire insulation. Current flows between turns at different voltages through the conductive contamination, carbonising the surrounding material, lowering the resistance further, and eventually creating a turn short.
Root causes: Contaminated cooling air entering an open-ventilated machine. Condensation forming on a cold winding when a machine is shut down without space heaters. Oil ingress from bearing seals. Carbon dust from brush gear. The fundamental problem is a combination of surface contamination and moisture.
Visual symptoms: Dark film on winding surfaces. Carbonised tracks between coils in the end winding — black lines following the surface of the insulation between adjacent high-voltage coils. In random-wound machines, evidence of dried oil or dirt deposits between conductors. IR and PI tests are a reliable way to detect this problem electrically before tracking has progressed to failure.
Test signatures: Low IR₁ and low PI (close to 1.0) are the primary electrical indicators. The IR test is specifically sensitive to surface contamination because surface leakage current is constant with time, which flattens the IR curve and reduces the PI. This is one of the failure modes the IR/PI test was designed to detect.
Remedies: This is one of the most reversible failure mechanisms. Even at an advanced stage, cleaning the winding thoroughly, followed by a dip-and-bake, varnish application, or GVPI treatment, can restore the winding to near-new condition. The varnish or impregnating resin must be rated for tracking resistance — some varnishes accelerate tracking rather than prevent it. Existing carbonised tracks should be removed by abrading before re-treatment, as they are permanently conductive.
Prevention: space heaters to prevent condensation during shutdown; improved air filtration on open-ventilated machines in dirty environments; brush gear maintenance to minimise carbon dust; regular cleaning intervals in machines known to operate in contaminated conditions.
Test Result Cross-Reference
When a test result is poor, the question is not just “is the winding bad?” but “what is causing the low reading and what does that imply for the machine?
Low IR₁ and low PI (PI close to 1.0): Most likely: surface contamination or moisture (Mechanism 9). Possibly: moisture absorbed into older thermoplastic insulation bulk. Action: clean and dry before any further testing or energisation.
Low IR₁ with acceptable PI (PI above 2.0): The surface or bulk conduction is driving the low reading, but the absorption current is still relatively dominant. Surface contamination with less severe coverage than the PI=1.0 case. The insulation may be in better condition than the IR reading alone suggests.
Normal or high IR₁, normal PI, but high PD on offline PD test: Most likely: loose coils in the slot (Mechanism 4); poor impregnation (Mechanism 3); semiconductive coating failure (Mechanism 5); or thermal deterioration (Mechanism 1) in a modern epoxy-mica winding that has delaminated without creating surface conduction paths.
Very high PI (above 6) on an old thermoplastic winding: Possible thermal deterioration of the asphaltic bonding material (Mechanism 1). Physical inspection — is the insulation brittle and dry? Does it sound hollow when tapped? High PI on old asphaltic windings is not always good news.
Declining trend in IR₁ over successive tests (corrected to 40°C): Any progressive deterioration mechanism. Contamination (9), thermal deterioration (1), or loose coils creating a growing tracking path. The direction of the trend is the signal — absolute values matter less than the rate of change.
Normal offline tests, failure in service: Turn insulation failure from surges (Mechanisms 7, 8), or loose coil failure that has not yet produced a visible PD signature. These mechanisms are the hardest to detect with standard IR/PI testing and require surge testing, PD testing at rated voltage, or wedge tightness assessment to identify.
One Rule That Covers All of Them
Every mechanism in this list is either accelerated by heat, moisture, contamination, vibration, or electrical stress — usually some combination. The machines that last are the ones that run clean, run cool, are properly impregnated, and are held tight in the slot. The machines that fail prematurely are the ones where one or more of those conditions was allowed to deteriorate — often for years — before anyone ran a diagnostic test.
The diagnostic tests exist to find these problems while they are still reversible. Most of the failure mechanisms above are repairable in Stage 1. Most require a rewind by the time Stage 2 or Stage 3 is reached.