Article content will be supplied from the Insights CMS collection.
Partial discharge testing is one of the more information-dense diagnostic tools available for stator winding condition assessment — and one of the more consistently misinterpreted. The fundamental challenge is that PD data is inherently comparative: a PD magnitude reading in isolation tells you significantly less than the same reading placed in the context of the machine's history, its operating conditions, and the pattern characteristics of the discharge activity.
This article covers what partial discharge testing measures, what the results indicate about insulation condition, and what the data actually supports in terms of maintenance decisions.
What Partial Discharge Is
Partial discharge is localized electrical discharge within or on the surface of insulation that does not constitute a complete breakdown of the insulation system. In generator stator windings, PD occurs at locations where the electric field exceeds the dielectric strength of a small void, delamination, or contamination site within the insulation — but the surrounding insulation remains intact.
The physical mechanism is straightforward: the insulation contains gas-filled voids (inherent in manufacturing or created by degradation), and the gas in these voids has a lower dielectric strength than the solid insulation. When the local electric field exceeds the gas's breakdown voltage, a small discharge occurs within the void. This discharge produces electromagnetic emission in the radio frequency range, a small charge transfer, and trace chemical byproducts (ozone, NOx in air-cooled machines).
Why PD matters. Each discharge event is small — measured in picocoulombs of charge. But the cumulative effect of repeated discharge activity is progressive insulation erosion. The discharge erodes the void surfaces, creating a larger void with more surface area for subsequent discharges. Over time, the discharge activity increases and the insulation thickness at the affected location decreases. If unaddressed, PD activity progresses until the insulation is sufficiently degraded to produce a complete ground fault.
Timescale. The progression from first detectable PD activity to insulation failure is not predictable from a single measurement. Some generators have operated for decades with elevated PD activity without failing; others have failed relatively quickly after PD was first detected. The rate of progression — assessed through trend analysis across multiple tests — is what drives maintenance decisions, not the absolute magnitude at any single test point.
Types of Partial Discharge in Stator Windings Stator winding PD is classified by location and mechanism. The pattern characteristics of each type are distinct in phase-resolved PD (PRPD) analysis, which makes pattern recognition a useful tool for identifying which type of degradation is occurring.
Slot Partial Discharge
Slot PD occurs between the surface of the stator bar's groundwall insulation and the stator core bore. In a properly installed stator winding, the outer surface of the groundwall insulation makes good electrical contact with the core bore through a semiconducting slot coating (the "semi-con" layer), which prevents large electric field gradients at the bar-to-core interface.
When the bar becomes loose in the slot — due to insulation shrinkage, vibration, or wear — the contact between the bar surface and the slot becomes intermittent. Air gaps at the interface create high-field regions, and slot PD develops at these locations. Slot PD is characteristically associated with bar looseness and is often accompanied by evidence of slot discharge contamination (gray powder composed of eroded insulation and core material) visible on the bars and wedges during outage inspection.
End-Arm and End-Turn Partial Discharge
End-arm PD occurs in the region where the bar exits the core slot and enters the end-winding space. This region has a specific concern: the grading of the electric field between the grounded core and the high-voltage conductor. Stress grading systems (resistive or non-linear materials applied to the bar surface in the transition zone) control this field gradient. Degradation of the stress grading system or mechanical damage to the end-arm region creates elevated field gradients and PD activity.
End-turn PD involves discharges between adjacent conductors in the end-winding region, typically at phase connections or where bars from different phases are in proximity.
Internal Delamination PD
Internal PD occurs within the body of the groundwall insulation at void or delamination sites. This is the most insidious PD type because it occurs within the insulation and is less visually apparent at inspection than slot or end-arm discharge. Internal delamination PD is associated with thermal cycling fatigue, where differential thermal expansion between the copper conductor and the insulation system creates interfacial separation.
Online vs. Offline PD Testing Generator stator PD testing can be conducted either online (while the machine is operating at voltage) or offline (with the machine de-energized and a high-voltage AC test source applied to individual phases).
Online PD Testing
Online PD testing is conducted while the machine is at operating voltage and connected to the system. Its primary advantage is that it captures PD activity under the actual thermal and mechanical conditions of service — slot discharge, which depends on bar looseness under centrifugal and electromagnetic forces, may only be detectable online.
Online PD sensors are permanently installed coupling capacitors (typically 80 pF) connected to the machine terminals. The sensors are passive and do not affect machine operation. Testing requires a portable PD analyzer; some installations have permanent online monitoring systems that provide continuous PD trending without scheduled testing.
Limitation. Online testing is conducted at the plant electrical environment, which includes electrical noise from the switchyard, the transformer, and adjacent equipment. Separating the machine's PD signal from this background noise is a critical part of online PD data interpretation — a reading that is dominated by external noise is not a reliable indicator of machine condition. Advanced PD analyzers use pulse polarity and arrival-time analysis to distinguish machine-generated PD from external interference.
Offline PD Testing Offline PD testing is conducted during an outage with the winding energized by a dedicated test source, typically at rated phase-to-ground voltage. Each phase is tested independently. The test environment is free from the electrical noise of online conditions.
Offline testing permits testing at multiple voltage levels — typically 25%, 50%, 75%, and 100% of rated voltage — which generates information about the voltage threshold at which PD initiates (inception voltage) and the rate at which PD activity increases with voltage. Both parameters are informative about the nature of the discharge.
Limitation. Offline testing does not capture thermally or mechanically dependent PD (such as slot discharge under centrifugal load). A machine may test acceptably offline and still have active PD when operating.
Interpreting PD Results Absolute Magnitude vs. Trend
PD magnitude (expressed in picocoulombs, or as normalized quantities Qm, NQN, or similar) does not have universal acceptance thresholds in the way that EL CID results have the 100mA guideline. IEEE 1434 (the standard guide for partial discharge testing of rotating electrical machinery) provides a framework for measurement and data acquisition but explicitly notes that interpretation is machine-specific.
The most meaningful comparison is to the machine's own history. A machine with a Qm of 5,000 pC at this outage and 800 pC at the last outage has a different risk profile than a machine with Qm of 8,000 pC at this outage and 7,500 pC at the last — regardless of which absolute reading is higher.
Phase-Resolved Partial Discharge (PRPD) Patterns
PRPD analysis plots PD events as a function of the phase angle of the applied voltage. Different PD mechanisms produce characteristic patterns:
● Internal void PD: Symmetric clusters centered at the positive and negative voltage peaks ● Slot discharge: Asymmetric patterns with predominant activity at one polarity, often with characteristic pulse shapes ● Surface tracking: Patterns spread across the voltage cycle with large magnitude pulses
Pattern recognition is a tool for identifying which type of degradation is producing the PD signal — which matters for both risk assessment and for directing maintenance. Slot discharge indicates mechanical looseness; internal delamination indicates insulation aging; stress grading degradation indicates a specific end-arm issue. Each has different remediation requirements. Context with Other Stator Tests
PD results are most useful when interpreted alongside polarization index (PI), insulation resistance (IR), and EL CID results. A machine with elevated PD and declining PI trend has a different risk profile than a machine with elevated PD and stable PI. Elevated PD in a machine with EL CID findings concentrated at the end region suggests a convergent degradation mode that warrants more urgent attention than the same PD result in a machine with a clean EL CID test.
What the Data Supports: Follow-Up Actions Stable trend, no pattern concern: Monitor at next outage interval. No current-window action unless online monitoring or IR/PI trends indicate acceleration.
Increasing trend, pattern change: Engineering review warranted to characterize the discharge type, correlate with other condition indicators, and assess the rate of progression.
Slot discharge pattern, increasing trend: Inspection priority during the next outage — slot tightness assessment, wedge condition, evidence of slot discharge contamination. The mechanical condition driving the discharge should be addressed, not only the discharge itself.
High inception voltage increase or decreasing PQ inception: Indicates advancing degradation at the void or fault site. Combined with trend data, this warrants evaluation of whether the current operating interval is supportable.
Any significant change between consecutive outages: Requires documented engineering basis for the return-to-service decision — not simply that the absolute readings are within some informal threshold.