A generator rarely fails without warning. It fails without anyone reading the warning. Partial discharge climbs for months, one stator slot runs a few degrees hotter than its neighbours, rotor vibration changes with field current, and each signal sits in a different system with a different owner. Then a winding fails in service, the unit is out for weeks, and the cost runs into millions. Generator condition monitoring puts those signals together and reads them as trends. This guide shows what to monitor on the stator and the rotor, how to read partial discharge, temperature, vibration and flux data, how to set baselines and alert rules and how to turn a drifting signal into a planned outage instead of a forced one. To see how it looks on your own unit data, book a generator monitoring walkthrough.
Generator Stator and Rotor Condition Monitoring
A practical guide for maintenance teams: how partial discharge, winding temperature, vibration and rotor monitoring show a developing problem early, and how to turn those trends into inspections and planned outages.
- What to monitor on the stator, the rotor and the cooling system
- How to read partial discharge and temperature as trends
- How to set baselines and alerts that name a maintenance action
Four causes explain almost nine tenths of the lost days, and the dashed line marks where the cumulative share passes 80%. Stator insulation and rotor winding alone account for three fifths. These are also the two areas where trending data gives the earliest warning, which is why monitoring starts there.
What Generator Condition Monitoring Covers
The stator, rotor, cooling system and bearings each fail differently, so each needs its own signal.
Online condition monitoring tracks partial discharge, winding temperatures, end-winding and shaft vibration, rotor flux, excitation and the cooling system while the unit runs, and trends them against load and cooling conditions. Offline tests during outages, such as insulation resistance and polarization index, winding resistance and core and rotor tests, confirm what the online data suggests. Together they show which winding or subsystem is changing and how fast. Our generator analytics team can show how this maps to your own units.
Insulation and core
Partial discharge, winding temperature, end-winding vibration and core condition.
Winding and shaft
Shorted turns, field ground, rotor temperature and shaft vibration.
Gas and water
Hydrogen purity and pressure, stator water flow and conductivity, temperatures.
Bearings and collector
Bearing temperatures, seal oil, brushes, slip rings and excitation.
Online data shows how the machine changes while it runs under load and heat. Offline tests measure condition under controlled conditions with the unit stopped. Use online trends to decide where to look, and offline tests to confirm what you found.
What to Monitor on the Stator
Stator insulation is the largest single outage risk, so it gets the most signals.
The stator winding ages through heat, electrical stress, vibration and contamination. Each mechanism leaves its own mark, and no single sensor shows all of them. To see which signals are available on your units, book a stator monitoring review.
Partial discharge level depends on load, temperature, humidity and the sensor, so one reading proves little. Compare like with like, watch the rate of change and the phase pattern, and confirm with offline tests before drawing a conclusion.
What to Monitor on the Rotor
Rotor problems show up in vibration, flux and field data, not in a single sensor.
The rotor is harder to see than the stator because it spins. Its winding can develop turn-to-turn shorts or grounds, its shaft can bow or rub and its collector can wear or spark. Monitoring has to read the effects of these on field current, vibration and temperature.
Some rotor shorts only show at higher field current or once the winding is hot. Compare data across the load range and at different temperatures, not at a single operating point.
Set Baselines and Alert Rules
An alert is only useful if it names what the maintenance team should do next.
Every generator is slightly different, so signals mean little without a baseline for that unit. Capture it when the machine is in known good condition, normalise for operating conditions and then alert on drift. Agree the alert tiers with the maintenance team before the data starts to flow.
A drifting signal should end in one of three actions: keep watching, inspect at the next opportunity or plan an outage. An alert that does not lead to one of these gets ignored within a month.
What One Avoided Generator Failure Is Worth
Catching a developing winding problem does not make the repair free. It moves the repair from an unplanned outage, with damage and a scramble for parts, to a planned one with spares ready and a date everyone agreed on.
From Drifting Signal to Planned Outage
The value comes from acting on the trend, not from collecting it.
Monitoring data tells you which unit and which winding to look at. Inspection and offline tests tell you what is wrong. Together they give the maintenance team a short list of actions with a date, which is what the plant needs to schedule an outage around demand and spares.
Connect
PD, temperature, vibration and excitation data.
Baseline
Known good condition for each unit.
Trend
Signals normalised for load and cooling.
Alert
Watch, inspect or plan an outage.
Inspect
Offline tests and visual checks.
Confirm
Result recorded next to the trend.
Common Generator Monitoring Mistakes
Most missed warnings were in the data and were read the wrong way.
Generator monitoring is easy to set up and easy to misread. A few simple rules keep the signals honest.
Good practice
- Set a baseline for each unit, not for the fleet
- Normalise for load, cooling and ambient before reading a trend
- Confirm online trends with offline tests at outages
Common mistakes
- Alarming on a single partial discharge number
- Reading signals without the operating conditions
- Treating monitoring as a replacement for outage inspections
Online data narrows the search and gives you time to plan. It does not replace visual inspection, offline testing or the OEM's guidance, so keep outage inspections in the plan and use the trend to set their scope.
How iFactory Generator Analytics Works
Generator signals in, ranked unit health out.
iFactory collects data from your partial discharge monitors, generator protection, DCS, historian, vibration monitors and hydrogen and water cooling systems. It builds a baseline for each unit, normalises the signals for load and cooling, scores unit health and compares units of the same design. It raises tiered alerts and links each one to an inspection or work order. It runs on an on-prem server inside your network, with one view across all units and plants. It brings your sensor data together and does not replace specialist monitors. Questions on fit go to our support desk.
All generator signals
PD, temperatures, vibration, excitation and cooling system data.
Unit-level normal
Signals normalised for load, cooling and ambient conditions.
Sister units
Units of the same design ranked on the same signals.
Tiered alerts
Watch, inspect or plan an outage, linked to a work order.
Results depend on your units, the sensors fitted and how quickly alerts are acted on. We measure baselines and the effect of each alert on your own data during the pilot, rather than promising a general figure.
Turnkey AI: Delivered, Connected and Live in 6–12 Weeks
You do not build this. It arrives ready.
iFactory ships as a pre-configured NVIDIA AI server with the software pre-loaded. Rack it, plug in power and Ethernet, and the AI is live on your network. Our team handles cabling, network setup, PLC and SCADA integration, team training and 24×7 remote monitoring. Data stays on your own network. For a scope matched to your units, request a turnkey quote.
Ship, network and generator data
Server installed. DCS, historian and monitor feeds connected for the pilot units.
Baselines and alert tiers
Baselines captured and normalisation checked. Alert tiers agreed with your maintenance team.
Go-live and training
Dashboards and alerts live. Teams trained. 24×7 remote monitoring begins.
Frequently Asked Questions
What is partial discharge monitoring on a generator?
It measures small electrical discharges inside or on the surface of the stator winding insulation, using couplers or sensors on the machine. A rising level over time can point to ageing or damaged insulation, so the trend matters more than any single reading.
How early can monitoring catch a problem?
It depends on the failure mode. Slow insulation ageing can show in trends over months. A sudden event, such as a foreign object or a fast-developing turn short, can progress much faster, so monitoring lowers risk without removing it.
What should we monitor on the rotor?
Shaft vibration, field current and voltage, rotor temperature derived from resistance, field ground detection, brush and collector condition and, where fitted, air-gap flux to detect shorted turns. Read them across the load range, not at one point.
Do we need new sensors to start?
Not always. Many plants start with signals they already have, such as winding and bearing temperatures, vibration, excitation and cooling data. Partial discharge couplers and flux probes add visibility, and the case for them depends on unit age and criticality.
Which standards and guidance apply?
IEEE 1434 and IEC 60034-27-2 cover partial discharge measurement on rotating machines, and IEEE 43 covers insulation resistance testing. Always follow your generator OEM's guidance too, and confirm the current editions with your reliability team.
How do we start?
With one or two critical units and the signals you already collect. A 6-week pilot connects the data, sets baselines and shows the first health ranking and alert tiers. To plan it, contact our team.
Know Which Generator Is Drifting Before It Trips
In thirty minutes we look at your units, your monitoring data and how drifting signals are followed up today. You keep the notes whether or not you go further with iFactory.
- 1The generator OEM manual and monitoring list
- 2Historian tags for temperatures, vibration and excitation
- 3Any partial discharge data and the last test reports
- 4Outage and inspection history by unit
- 5A list of sister units with ratings
