On-Line Partial Discharge Monitoring and Testing

Turbo Generators

Presented by: Thorsten Hübers

Phone: +49 1520 8814808
Email: thuebers@qualitrolcorp.com
Website: www.irispower.com
 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
Fortive Platforms
 Qualitrol Overview
Iris Power 2010 DMS 2009 Neoptix Hathaway 2002 AKM
PD Tech 2009 Serveron 2013 2010 LEM/ 2005 1998

Qualitrol was founded in 1946 in East Rochester, NY as a supplier to OEM

for transformer protection devices.
Acquired by Danaher in 1986
AKM PD Tech Iris Power
Operating locations in Belfast, Glasgow, Quebec City, Toronto, Rochester
(NY), Beaverton (OR)
Founded in 1990 by 4 individuals from Ontario Hydro's Research
Division, one of the largest electrical utilities in North America
Core Business: Leader in Partial Discharge products for testing and
monitoring for rotating machines (>3.3kV)
PD Testing was funded by the Electric Power Utilities, not by the
generator manufacturer.
ISO9001 Certified
Size: > 100 Employees
Head Office: Toronto, Canada
Satellite Offices:
Houston, Texas San Paulo, Brazil
Beijing, China Mumbai, India
Brugg, Switzerland
+ Agents located worldwide
 IRIS Portfolio
Online Measurement (periodic &continuous)
- PD
- Endwinding Vibration
- Flux
- Current Signature Analysis
- Air Gap
- Medium Voltage Switchgear /dry type transformer
Monitoring
Offline Measurement
- PD
- Stator Core with ELCID
- Stator Wedge Analyser
Rotating machines
Catastrophic failure
 History of on-line PD Testing

1949: On-line PD testing first developed by Westinghouse. Required experts to analyze results.
1975: Ontario Hydro (with EPRI & CEA) develops the first non-expert, on-line PD test method
for large hydrogenerators.
1988: Development of On-line PD test method for motors and turbo generators.
1990: Key personnel from Ontario Hydro leave to form Iris Power LP to market on-line PD
testing.
2002: Over 50% of utility generators (>20 MW) in North America use Iris
technology. Rapidly expanding use in Europe and Asia.
2012: Meeting the requirements of IEC 60034-27-2 (issued in 2012)

2013: Over 15,000 Machines worldwide using IRIS technology.

 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 What is Partial Discharge (PD)?
A partial discharge is in the form of a small electrical spark that is
caused by the release of energy across a gas filled void (air pockets)
within the insulation or at the surface of the insulation.
These small electrical sparks create small current pulses which can be
detected and monitored.
Breakdown strength:
- Air ~ 3kV/mm
- Epoxy Mica Insulation ~ 300kV/mm
IEC defines a PD as an incomplete (partial) electrical discharge that
occurs within insulation or between insulation and a conductor.
Testing, monitoring and analysis of partial discharges – is a method to
detect most major stator winding failure mechanisms
 PD Pulse Characteristics
▪ Very fast rise-time ~ (1-5 ns)
▪ Using f=1/T=1/(4xRise-time)
▪ PD creates High-Frequency current pulses that is coupled to the line
▪ Goal: Detect the PD in high frequency spectrum

Assuming 5ns Rise Time

T = 20ns

Frequency = 1/T
= 1 / 20e-9
= 50MHz
 PD Theory: Capacitive Model
1) A Potential difference (V) builds up
across an gas-filled void

2) PD occurs when:
Ev=Vv/dv > 3 kV/mm Vcopper
(Electrical stress exceeds electrical
breakdown point of the gas in the Cins
void)
VCopper Vvoid Cvoid
3) Each discharge breaks the
molecular bonds in the insulation Cins

4) The larger the void, the larger the

discharge
0V
5) Monitor PD by detecting and
measuring the resulting current
pulses
 PD Theory: Capacitive Model
Simple plate capacitor:
C = ∈A/d V0
+ Cins
If we assume: V0
- +
dair = 0.5mm Vair Cair
dins = 4mm -

And our dielectric constants are:

∈air = 1 ∈0
∈ins = 4 ∈0
0V
Cair = 1 ∈0 A / dair Cins = 4 ∈0 A / dins
= 1 ∈0 A / 0.5mm = 4 ∈0 A / 4mm
= 2 ∈0 A / mm = 1 ∈0 A / mm

Vair = Vo {Cins / (Cins + Cair)} -> Voltage divider

Vair = Vo {1 ∈0 A / mm / (1 ∈0 A / mm + 2 ∈0 A / mm)}

Vair = Vo / 3
 PD Theory: Capacitive Model
Vair = Vo / 3
V0
Assume 11kV machine + Cins
V0
- +
-> Phase to ground voltage is 6,35kV. Vair Cair
-

Vair = 6,35kV / 3

Vair = ~2,12kV

Eair =Vair/dair
0V
dair = 0.5mm

Therefore, electrical stress across gap = 2,12kV/0.5mm = 4,24KV/mm

This exceeds 3KV/mm -> Electrical breakdown would occur

-> Partial Discharge

 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 Rotating Machine Failure
According to IEEE & EPRI (Electrical Power Research Institute) studies:

Mechanical(50%)
50% due to bearing/vibration

Electrical (50%)
40% due to failure of the Stator Winding Insulation
Solution: ON-LINE PARTIAL DISCHARGE TESTING

10% due to Rotor

Solution: FLUX MONITORING (SYNCRONOUS)
CURRENT SIGNATURE ANALYSIS (INDUCTION)
Failure Mechanisms in Stator Windings
Load Cycling
Insulation voids next to copper conductors

Poor impregnation of coils

Internal voids
Loose windings
Slot discharge
Overheating of coils
Insulation delamination
Endwinding Vibrations
Deterioration of semi-conductive or grading coatings
Slot or endwinding discharge
Contaminated windings
Electrical tracking
 Failure Mechanisms
VOIDS WITHIN THE INSULATION CAUSED BY POOR IMPREGNATION

6.6 kV Motor
Poor impregnation using global VPI technique leads to inadequate bonding of
groundwall – produces VOIDS
 Failure Mechanisms
VOIDS NEAR THE CORE IRON CAUSED BY LOOSE COILS

Abrasion of bar/semicon
coating

200 MW
Turbogenerator
 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 Online detection of PD
Challenge when measuring PD
- Differentiate between electrical noise and
machine PD
Noise sources:
- Sparking electrical connections
- Slip ring sparking
- Transmission line corona
- Electrical Equipment, sparking power tools
- Etc..
 Online detection of PD
Accuracy of PD data is key!
- False readings lead to incorrect diagnoses and
unnecessary shutdown & cost.
- IRIS Power has 1.5% false indication rate

If electrical noise is NOT separated, testing is

inaccurate and is SUBJECTIVE.
If electrical noise IS separated, testing is
accurate and is OBJECTIVE.
 4 Methods of noise separation
Simultaneously use 4 Methods to Separate PD
from Noise:
High pass filter (noise predominantly <20
MHz)
Time of pulse arrival between a pair of
couplers
Pulse shape analysis (rise time and degree of
pulse oscillation)
Surge impedance mismatch
 Noise Separation Technique
Test Frequencies
Noise
Partial Discharge

5 20 40 100 350

80 pf Sensor
Test Instrument
Frequency (MHz)

Remove Noise to MAXIMIZE Signal to Noise Ratio

Iris’ test equipments are capable of detecting the fast

rise-time initial pulses and high frequencies between
50-250Mhz – typical of PD originating in the winding.
 Capacitive Couplers
Epoxy Mica Capacitor (EMC)
▪ 80pF connected directly to high
voltage line
▪ Creates a High Pass Filter (HPF)
Large for < 25kV
with cutoff frequency at 40Mhz
▪ Block the 50Hz High Voltage and
pass the high frequency PD Medium for < 16kV
signal
▪ Over 60,000 installed
▪ Meets reliability requirements Small for < 6.9kV
of IEC 60034-27-2
▪ Permanently install 1 or 2 per
phase, depending on machine
application
 Time of flight
Double-Ended Install - “Time of pulse arrival between a pair of couplers”

DISTANCE X DISTANCE X + Delay

TGA-B
Classification Machine sensor System Sensor
Machine PD X X + Delay
System Noise* X + Delay X
Bus Noise* X+ < Delay X+ < Delay
 Time of flight
Double-Ended Install - “Time of pulse arrival between a pair of couplers”

Speed of light:
300.000m/s=0,3m/ns
delay= 10 m
DISTANCE X DISTANCE X
+ Delay V=d/t
T=d/V=10m/0,3m/ns TGA-B
T=33ns
Classification Machine sensor System Sensor
Machine PD X X + Delay
System Noise* X + Delay X
Bus Noise* X+ < Delay X+ < Delay
 Long cable effects
Single-Ended Install - “Pulse shape analysis (rise time and degree of pulse
oscillation)”

Power Cable
>30m
NOISE SIGNAL
ATTENUATES

PD SIGNAL TGA-B
PASSES
Pulse Shape versus Cable Length
Couplers on Machine Terminals
25 kV Couplers on Large Turbo
 Stator Slot Couplers

PD Sensors for Large Hydrogen-Cooled

Turbogenerators:

▪ Capacitive couplers on stator terminals may give

false indications if
▪ sparking from stator core and /or
▪ flexible leads

▪ Stator Slot Couplers (SSCs) will not give such

false indications
 Stator Slot Couplers

▪ Serves as an antennae (1-

1000Mhz)
▪ Comprised of an electrode structure
printed on an epoxy-glass laminate
▪ No high voltage necessary
▪ 1 coupler installed per parallel per
phase
▪ Custom made to fit your unique
stator slot
▪ Over 1500 installations worldwide!!!
Location of SSC
Location of SSC
 Pulse Shape Analysis
PD pulses are unipolar and have a width < 6 ns
 Pulse Shape Analysis
Noise pulses are oscillatory and have a width >6 ns

DM 2000DM 2000
44
SSC Termination
 Iris Termination Box
Most Motors & Small Turbo-Generators Large Turbo-Generators (>200MW)

2 sensors installed per phase 1 sensor installed per parallel per phase
 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 Periodic Monitoring
Portable Instrument:
Turbine Generator Analyzer (TGA-SB)

Used on most motors & turbine generators ( > 200MW)

Results: Complete analysis of PD data: 2D and 3D plots
 Early warning

On-Line PD
Provides early warning: 24kV ~ 4-6 years

Testing:
11kV ~ 3-4 years
6kV ~ 1-2 years
3,3kV ~ several weeks or months

With regular PD testing, failure

mechanisms can be detected years
in advance.
 Continuous Monitoring
With TracII & GuardII Systems
▪ When PD levels exceed predetermined levels:
➢ System triggers Alert
▪ Detect rapidly evolving failure mechanisms: maximize
warning time in case of problem.
▪ Utilized to monitor machines
in remote or hazardous locations.
▪ ‘Set and Forget’ type of
technology.
▪ Can be part of plant data
automation process.
 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 How to Analyze Results?
2-step process:

1. Which machines are at risk?

2. What is the cause of the problem?

Iris uses 3 ways to determine which machine needs attention:

1. Compare with vast database.

2. Compare with the similar machines.

3. Compare machine based on trending.

 Iris’ PD Database
▪ Over 550,000 test results from thousands of machines around the world
▪ Every 2 years, Iris publishes the statistical range of Qm (Peak PD) for each type
of stator (voltage rating, air/gas cooling, type of sensor, etc.)
▪ If a stator has a Qm that exceeds 90% of readings then stator winding insulation
is deteriorated and considered in “Bad Shape”
▪ The definition of “Bad Shape” is based on our extensive history of comparing
test results with the subsequent machine inspections and rewinds

kPa No H2 75-205 210-345 >345

25% 1mV 0mV 0mV 0mV
50% 4mV 2mV 3mV 0mV
75% 14mV 15mV 14mV 6mV
90% 68mV 45mV 37mV 15mV
95% 105mV 78mV 53mV 22mV
Table – Qm values (mV) for H2-cooled machines with SSC sensors [16-18kV]
 PD Trend Plot
PD Trend Plot:
Effect of maintenance on Partial Discharge Level
1000

900

800

PD+
700 DIAGNOSES?
PD-
600

Qmax 500
(mV)
400

300

200

100

0
 Phase Resolved Partial Discharge Plot

Importance of Noise Separation

Unit PD Noise
Pu ls e D en sity L in ear Plo t Pu ls e D en sity L in ear Plo t
B ipola r M a c hine P D B ipola r T ota l S ys te m N ois e

0 to 3.16 pps 3.16 to 10 pps 10 to 31.6 pps 31.6 to 100 pps 0 to 3.16 pps 3.16 to 10 pps 10 to 31.6 pps 31.6 to 100 pps

100 to 316 pps 316 to 1000 pps > 1000 pps 100 to 316 pps 316 to 1000 pps > 1000 pps
150 150 150 150

100 100 100 100

P u lse M a g n itu d e [m V ]

P u lse M a g n itu d e [m V ]
50 50 50 50

0 0 0 0

-5 0 -5 0 -5 0 -5 0

-1 0 0 -1 0 0 -1 0 0 -1 0 0

-1 5 0 -1 5 0 -1 5 0 -1 5 0

0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360
P h a s e A n g le [ d e g ] P h a s e A n g le [ d e g ]
 How to deterime Failure Mechanism
Polarity Temperature Load FAILURE
Phase Angle
Predominance Dependence Dependence MECHANISMS

45º/225º  ↑↓  Manufacturing
Defect
15º/195º
75º/255º  ?  Phase to Phase

225º + ↑↓ ↑↑ Bar/Coil
Movement

45º/225º  ↑↓  Internal
Delamination

45º – ↑↓  Thermal Cycling

15º/195º
75º/255º  ?  Contamination
 Outline
About us
PD theory
Failure mechanisms
Online PD detection
Test instruments
Result analysis
Conclusion
 Why perform PD testing
Stator windings are subject to a variety of failure mechanisms.
➢ Stator winding is second highest cause of failure (40%).

Partial Discharges occur in these voids and are symptoms of

these failure mechanisms. .

Identified levels of PD will indicate a high-risk of stator winding

failure.

PD data can aid in identifying the failure mechanism.

Specific phase angles and temperature, load, pressure effects

correlate to specific failure mechanisms
 Benefits of On-Line Testing
No interruption of your power production - no production losses
Measure at load conditions
Avoid in-service failures or unnecessary rewinds
Improve the overall reliability of generators and motors
Maximize the operating hours for the stator winding
Detect a wide range of failure mechanisms
Prioritizing Maintenance

Break & Fix Preventive Maintenance Predictive Maintenance

 PD Testing with Iris
▪ Iris is the largest supplier of on-line PD test equipments, with more
than 15,000 equipped on rotating machines

▪ NO EXPERT REQUIRED!!! Plant personnel can perform and

accurately interpret results after only 2 days of training.

▪ Iris has the largest accessible database. Over 550,000 test results
from thousands of machines

▪ Meeting the requirements of IEC 60034-27-2

▪ Proven noise separation methods have allowed for reliable and low
cost testing

▪ There is no conflict of interest with IRIS testing as we are a neutral 3rd

party

▪ IRIS(through Ontario Hydro) originally developed PD technology and

continues to innovate high quality products for predictive maintenance
testing on motors and generators.
Thank you for your attention
-
Questions?