Transcript March 2008

Medium Voltage Induction Motor Protection and
Diagnostics
Yi Du
Pinjia Zhang
Prof. Thomas G. Habetler
School of Electrical and Computer Engineering
Georgia Institute of Technology
Atlanta, GA
1
Medium Voltage Facilities
Measurement and instrumentation
Data acquisition and storage devices
Fiber optic interface
Potential
transformers
A
Current
transformers
B
Thermocouples or other
temperature measurement devices
D
4160 V
E
Coupling
F
480 V
Power
Supply
C
G
4160 V
480 V
2
Medium Voltage Supply
3
Medium Voltage Laboratory
4
5
6
7
8
9
10
11
12
13
14
Outline
• Introduction
• Heat transfer inside Motors
• Thermal Model-Based Approaches
• Parameter Model-Based Approaches
• Other Approaches
15
Medium voltage induction
motors



Mostly used in the petroleum,
chemical, mining and other
industries,
Rated from 2300 V to 13200 V,
They are rotor limited during
starting, and stator limited
under overload.
16
Overload Protection


Malfunctions of these motors are very
costly due to loss of productivity,
The winding insulation failure is a
typical malfunction, which is often
caused by overload.
17
Conventional Overload Relays



Conventional overload
relays utilize simple
thermal models and
embedded temperature
sensors.
Simple thermal models
can not estimate the
rotor temperature.
Disintegration of the
connection, noise
interference, and large
time constant of the
sensors often result in
false alarm or trips.
Motor
Parameters
Current
Measurement
Cooling Time
Constant
Overload
Curve
Current
Unbalance Bias
Voltage
Measurement
Ambient
Temperature
Thermal
Capacity
Check
Stator
Temperature
Alarm/Trip
Alarm/Trip
18
Requirements


Track the thermodynamic behavior of the motor's
stator and rotor under steady and transient state
conditions.
It should also take into account the important
differences in the thermal behavior due to the
motor size and the type of construction and
ventilation.
19
Possible Approaches


Higher order thermal model-based approaches
Model the thermal behavior of the motor. The thermal
parameters are calculated from the motor dimensions
and offline experiments. This approach is robust, but
measurements need be made for each motor.
Parameter-based approaches
Estimate the temperature from the variation of the
resistance of the stator and rotor. This method can
respond to changes in the cooling conditions, and is
accurate, but it is generally too sensitive.
20
Outline
• Introduction
• Heat Transfer inside Motors
• Thermal Model-Based Approaches
• Parameter Model-Based Approaches
• Other Approaches
21
Motor Losses

The temperature rise inside a motor is caused by
the losses accumulated in the motor.
22
Loss Segregations
Wlosses  Ws  Wr  Wcore  W fw  WLL
Compared with low power motors,
high power motors have larger
percentage of core loss and stray
loss, and smaller percentage of
copper loss.
Therefore, the thermal model only
considering copper loss is not
suitable for large motors.
Loss segregation for 15Hp motor
Wcopper
Wcore
Wf&w
Wstray
2 Pole
53
9
29
9
4 Pole
55
15
18
12
6 Pole
62
13
12
13
Loss segregation for 200~2000Hp motors
Wcopper
Wcore
Wf&w
Wstray
2 Pole
29
15
36
20
4 Pole
35
18
24
23
6 Pole
37
23
18
22
23
Heat Transfer
The heat transfer inside a
motor can be classified into
Conduction
- transfer of heat due to the temperature difference.
Shaft – rotor iron – rotor winding,
Stator winding – stator iron – Frame, …
Convection - transfer of heat due to the fluid motion.
Frame - external air, stator/rotor– airgap, rotor – endcap air, ...
Radiation - transfer of heat by electromagnetic radiation.
Radiation is ignored since the motor temperature is relatively low.
24
Thermal resistance and thermal capacitance
Thermal
behavior of the motor can be analyzed by
Finite element methods (Time consuming)
Lumped-parameter thermal network, composed of thermal
resistors, thermal capacitors and heat sources.
Some thermal resistances and thermal capacitances can be calculated
directly from the motor dimensions.
Other thermal resistances are complex and can only be measured
online.
Stator core to frame conduction resistance
Endwinding cooling resistance
Frame to ambient convection resistance
25
Thermal Network
Given
the difficulties to calculate certain thermal
parameters, detailed thermal models can not
guarantee good accuracy.
Simplification of the thermal network is preferred for
online monitoring.
On the other hand, the thermal network should be
complex enough to estimate the hot spot temperature.
26
Outline
• Introduction
• Heat Transfer inside Motors
• Thermal Model-Based Approaches
• Parameter Model-Based Approaches
• Other Approaches
27
Thermal Model-based Approaches
Use
thermal network to model the thermal behavior of the
motor.
The Motor is divided into homogenous components
wherein each part has a uniform temperature and heat
transfer coefficients.
The heat flow paths are determined and thermal resistors
are added between the nodes.
Losses and thermal capacitors are allocated to each node.
28
First - order Thermal Model
Stator θs
Used
in conventional relays
for its simplicity,
Do not consider the rotor
winding temperature,
The stator winding
temperature is given by,
s (t )  Ploss  Rth (1  e

t
RthCth
)  s 0  e
Ploss
Cth
Rth
θA

t
Rth Cth
Ambient
A
29
First - order Thermal Model
Thermal
resistance Rth and thermal capacitance Cth can
be directly calculated from the trip class t6x and the
service factor SF.
Assume the loss Ploss equals the stator copper loss
2
6
Rth  Cth  t6 x /[ln( 2
)]
2
6  SF
Cth
is calculated using the trip class and the winding
insulation class.
30
First - order Thermal Model
Temperature
rise is a complex combination of
distributed thermal capacitances and resistances,
single time constant is not enough.
Therefore, large margin is needed for safety
and the motor is over protected.
The rotor temperature can not be monitored.
31
Second - order Thermal Model
θs
Stator
Stator
and rotor are
modeled separately, P
Model B eliminate
the node while
maintaining the same
function.
P
Parameters are
calculated from
offline experiments
s
s
Cs
θr
Internal Cooling Air
Rotor
Rtr
Rts
Cr
Rta
θA
Pr
Ambient
Model A
θs
Stator
θr
Rotor
R3
Cs
R1
R2
θA
Cr
Pr
Ambient
Model B
32
Second - order Thermal Model
Model
C simplifies
the rotor side, and
less parameters
are needed.
Second-order
thermal model is a
good tradeoff
between accuracy
and complexity.
Model C
33
Higher - order Thermal Model
Model
the hot spot,
such as end windings,
seperately.
The thermal model
becomes complex and
it is difficult to identify
the parameters.
34
Outline
• Introduction
• Heat Transfer inside Motors
• Thermal Model-Based Approaches
• Parameter Model-Based Approaches
• Other Approaches
35
Parameter-based Approaches
Estimate the temperature from the variation
of the stator winding resistance and the rotor
bar resistance.
Ra (t b  k1 )
Rb 
t a  k1
k1 is 234.5 for 100% IACS conductivity copper
It is an online method and can respond to
changes in the cooling conditions.
36
Rotor Resistance
•Rotor resistance can be calculated in the synchronous reference
frame with the d-axis aligned with the stator current. Under the
steady state, the rotor resistance, which is independent of the stator
resistance, is given by
Rˆ r  ( ss ) 2 Lr [
s L2m
Vqs
Is
 s Ls
 Lr ]
•Rotor resistance can also be calculated in the stationary reference
frame and rotor reference frame.
•By these methods, the rotor resistance is independent of the stator
resistance and is less sensitive to the parameter variations.
37
Stator Resistance
•Stator resistance is generally calculated based on rotor
resistance.
•In the synchronous reference frame with the d-axis aligned with
the stator current, the stator resistance is given by,
e
2
ˆe
V
s

L

Rˆ s  eds  e me dr
ids
Rˆr ids
•Rotor speed can be calculated from the stator current harmonics.
38
Outline
• Introduction
• Heat transfer inside Motors
• Thermal Model-Based Approaches
• Parameter Model-Based Approaches
• Other Approaches
39
Neural Network - based Approaches
•Neural networks have been proposed to estimate the
stator resistance and rotor resistance.
•The advantages are that they do not require the motor
parameters and can be easily implemented.
•The drawbacks are they are
still sensitive to the parameter
changes since the network is
trained using the data based
on certain parameters.
40
Hybrid Approaches
•Combine thermal model – based approaches with
parameter – based approaches,
•Rotor temperature is estimated by parameter –
based approaches since it is less sensitive to the
parameter variations,
•Stator temperature is monitored by thermal model
– based approaches.
41
Signal Injection-based Approaches
Vas ,dc 2 Vab,dc
2 Vsw, dc
ˆ
Rs 


ias ,dc
3  ias ,dc
3  ias ,dc
•The stator resistance is estimated from the dc
components of the voltage and current.
•Relatively accurate since it is not affected by the
inductance of the motor.
•It is intrusive and introduces torque oscillation.
42
Overview of Fault Diagnostics for MV Motors
Distribution of MV Induction
Motor Failures
Induction Motor Fault
Categories
43
OUTLINE




Overview of Fault Diagnostics for MV Motors
Bearing Failure and its Diagnostic
Stator Winding Inter-turn Fault and its Diagnostic
Rotor Fault and its Diagnostic
–
–

Broken Rotor Bar & End-Ring Faults and their Diagnostic
Rotor Eccentricity and its Diagnostic
Conclusions
44
Overview of Fault Diagnostics for MV Motors
Distribution of MV Induction
Motor Failures
Induction Motor Fault
Categories
45
Analysis of Fault Diagnostics for MV Motors



Main differences between MV motors and
small low-voltage motors:
High Insulation Requirement for Stator
Winding: stator winding inter-turn fault
Large Output Torque: rotor and bearingrelated mechanical faults
High Thermal Stress: stator insulation failure
and rotor-related faults
46
Bearing Failure Monitoring
Bearing failure is the most common fault for
MV motors.
Reasons for Bearing Failure
 Electrical Stress:
Stator, rotor or input voltage unbalance causes
unbalanced magnetic flux, which induces shaft
current, and potential voltage between bearing
and ground.

Inner raceway
Ball
Cage
Mechanical Stress:
–

Outer raceway
Friction and rotor eccentricity can cause
mechanical failure of bearings.
Thermal Stress:
–
Overheat causes the failure of lubricant, which
lead to friction.
47
Bearing Failure Monitoring

Classification of Bearing
Failure:
Single Point Defects:
–
–
–
–

Outer raceway
Inner raceway
Ball
Cage
Generalized Roughness
Existing Methods:
• Standard vibration sensor
method
• Chemical analysis method
• Temperature monitoring
• Acoustic emission method
• Sound pressure method
• Current signature spectra
method
48
Current signature spectra methods
1.Single point defects:
 Wavelet method
 Neural network clustering method
 Adaptive time-frequency method
 Park vector trajectory method
 Other methods
2.Generalized roughness:
 Mean spectrum deviation method
Fundamentally: monitor the E-M torque harmonics corresponding to
the mechanical vibration frequencies
49
Bearing Failure Monitoring
is the power supply frequency;
is the vibration frequency;
is the corresponding stator current signature frequency.
Challenges for MV motors
For single point defects:
 Poor Signal/Noise Ratio
Due the large output torque, the torque vibration caused by bearing
failure is more difficult to observe. So the low signal/noise ratio is a
potential problem for current-based bearing diagnosis of large MV
motors.
For generalized roughness:
 Separate measurement noise and bearing failure-related vibration
noise
50
Stator Winding Inter-turn Faults

Reasons for Stator Inter-turn Fault
Electrical Stress:
–

Thermal Stress
–

Motor life is reduced by 50% for every 10°C above limit
Mechanical Stress
–

High voltage causes winding insulation failure
Friction between stator and rotor caused by rotor eccentricity
Other Stress
51
Stator Winding Inter-turn Faults
Existing Methods:
• Negative Sequence Current
• Negative Sequence Impedance
• E-M Torque Harmonics
• Current Spectrum
• Current Park Vector Trajectory Stator Inter-turn Fault
• Artificial Intelligent Methods
Fundamentally: Monitor the unbalance of stator winding
52
Stator Winding Inter-turn Fault and its
Diagnostic
Challenges:
Coupling between Negative-Sequence
and Positive-Sequence models due to
stator winding asymmetry
Stator Inter-turn Fault
Indicator (function of
slip frequency)
• How to consider voltage unbalance in power supply
• How to consider original stator winding unbalance
• How to set threshold for negative-sequence impedance
53
Rotor-related Failures






Rotor-related faults can be classified into:
Broken Rotor Bar
Broken Rotor End-Ring
Rotor Eccentricity (shaft misalignment)
Reasons for rotor-related faults:
Mechanical stress: including rotor eccentricity, and stator-rotor
friction
Thermal stress: overheat in rotor can cause rotor deterioration
Electrical stress: frequency starting and overload operations
can cause thermal stress due to large current; unbalanced flux
can induce unbalanced magnetic pull.
54
Broken Rotor Bar and End-Ring Faults

Broken rotor bar fault can cause
unbalanced magnetic flux, and thus
torque oscillation and stator current
harmonics.
• Due to large output torque, and large rotor current,
broken rotor bar fault is more common on large MV
motors than small motors
• The effects of broken rotor end-ring are the same as
broken rotor bar, in the sense that the rotor flux is
asymmetric, and induces harmonics in the stator
current.
55
Broken Rotor Bar and End-Ring Faults
Existing methods:
 Signature current analysis
 EM torque harmonics monitoring
 Slot harmonic methods
 Starting current analysis
 Pattern recognition-based methods
 Artificial intelligence-based methods
Broken Rotor Bar-related current harmonics
 Other methods

Fundamentally: monitor the signature harmonics and slot
harmonics in stator current
56
Rotor Eccentricity
Rotor eccentricity is a possible reason
for many kinds of motor faults, such as
stator insulation failure, broken rotor bar
and end-ring, and even shaft crack.
Rotor Shaft Crack
• Rotor eccentricity is mainly caused by shaft
misalignment, when the geometric center of the rotor
does not coincide with the center of the stator.
• The current harmonics related to rotor eccentricity are
57
Rotor-related Faults


For MV motors, due to the high thermal stress on rotor, and the large
output torque, especially the starting acceleration torque, rotor-related
faults are quite common.
The fundamental methods for rotor-related faults are current signature
analysis, as the signature frequencies related to broken rotor bar or
eccentricity are well-known.
Challenges for MV motors
 Separating signature harmonics from load oscillation
–

The signature harmonics in stator current are caused by the unbalanced
rotor flux, but the same harmonics can also be caused by the load
oscillation.
Diagnostics for drive-connected motors
–
the low-frequency harmonics can be cancelled or reduced by the controller.
58
Conclusions

Motor faults diagnostics:
Stator Inter-turn Fault
monitor the unbalance of stator winding

Bearing Fault
monitor the current harmonics caused by bearing-related
torque vibration

Rotor Fault
–
–
Broken Rotor Bar/End-ring
Rotor Eccentricity
monitor the current signature harmonics caused by unbalanced
rotor flux
59
Conclusion of Motor Faults and their
Diagnostics for MV Motors





Challenges for fault diagnostics of MV motors
Compensate for the effect of power supply
and original motor unbalance
Cancel the effect of load oscillation on
diagnostics
Reliable diagnosis even with low SNR
Fault diagnostics for drive-connected
systems
Remote condition monitoring
60
QUESTIONS?
61