Condition Monitoring of Three Phase
Induction Motors
Presented By:
Amandeep Sharma
Ph.D. Scholar, EE Department, NITTTR
1
CONTENTS
• INTRODUCTION
• CONDITION MONITORING
• WHAT IS VIRTUAL INSTRUMENTATION
• DAQ ASSISTANT
• DATA ACQUISITION AND PROGRAMMING
INTRODUCTION
The IM has become an essential commodity for most of the
industrial processes such as manufacturing, assembling,
and packaging.
To reduce their downtime and unexpected failures, motors
require continuous monitoring and their operating
conditions needs to be checked.
INTRODUCTION
INDUCTION MOTOR FAULTS
6
MECHANICAL FAULTS
ELECTRICAL FAULTS
FAULT DIAGNOSTIC METHODS
Many diagnostic methods have been developed so far, some of
which are summarized as follows:
1. Electromagnetic Field Monitoring (Axial Flux-Monitoring)
2. Temperature Measurement (Thermography)
3. Vibration Monitoring
4. Acoustic Noise measurement
5. Partial Discharge Analysis
6. Chemical Analysis
7. Motor Current Signature Analysis (MCSA)
8. Model-based Artificial Intelligence and Neural-Network-
based Techniques
10
FAULT DETECTION
There are three important components to fault
detection, in the order of difficulty, as follows:
i. Fault alarming, i.e. to indicate whether there is
damage;
ii. Fault localization, i.e., to identify the location of
damage; and
iii. Fault quantification, i.e. to quantify the extent
of damage.
FAULT DIAGNOSIS
12
METHODOLOGY
13
15
Bearing and broken rotor bar
fault diagnosis
16
MOTOR CURRENT SIGNATURE ANALYSIS (MCSA)
Commonly used technique, simple and low cost
Based on the spectral component analysis of the stator
current harmonics using FFT
These harmonics are caused by new rotating flux
components on account of a fault.
17 17
ADVANTAGES OF MCSA
Comparative Analysis of Broken
Rotor Fault Diagnosis
SIDEBAND HARMONICS
Effect of a broken bar fault is the appearance of side-
band components in the frequency spectrum of the
stator current.
where, k = 1, 2, 3, …; s is the slip in per unit and; fs is
the fundamental frequency of the stator current
(power supply).
Classical MCSA based Fault
Diagnosis
• FFT current spectrum for motor
under different load conditions (a)
healthy motor; (b) motor with 1
BRB; (c) motor with 3 BRB.
SIDEBAND HARMONICS
Difference of magnitude less than 50 dB is a clear
indication of presence of broken rotor bars.
If the difference is in the range between 40 and 50
dB there is probably one broken bar present,
difference less than 40 dB indicates several broken
rotor bars or presence of a broken end ring.
Apart from spectral components, broken bars are
classified using speed oscillations and torque
ripples.
PARKS VECTOR APPROACH (PVA)
Park’s transformation, reduces the motor three
phase current into two components which are in
orthogonal reference frame.
The two orthogonal current Park’s vector
components (id, iq) of the three supply current (iu,
iv, iw) are given by:
25
PARK’S VECTOR ANALYSIS
The concept behind the current monitoring
using Park’s method is the absence of homo-
polar or zero sequence current components in a
Y-connected induction motor due to absence of
neutral wire.
For a healthy motor, under ideal conditions the
Park’s vector gives a circular pattern centered at
the origin of the coordinates.
Park’s Current
Vector Analysis
• Parks current vector for motor
under (a) no load and full load
healthy condition; (b) no load and
full load conditions with 1 BRB; (c)
no load and full load with 3 BRB
fault
Front Panel and Block Diagram of
Virtual Instrument Developed for PVA
31
EXPERIMENTAL SETUP
32
NN BASED FAULT DETECTION
CLASSIFIER BASED FAULT DETECTION
METHODS
FEATURE EXTRACTION
TIME DOMAIN FEATURES FREQUENCY DOMAIN TIME-FREQUENCY FEATURES
FEATURES
MEAN, RMS, VARIANCE, MEAN FREQUENCY, RMS DWT/ CWT COEFFICIENTS
KURTOSIS, SKEWNESS, SHAPE FREQUECCY, FREQUENCY
FACTOR, IMPULSE FACTOR, CENTRE, ROOT VARIANCE
CREST FACTOR, ENTROPY, ETC. FREQUENCY
NEURAL NETWORK BASED FAULT
DIAGNOSIS
NEURAL NETWORK SIMULINK
MODEL TRAINED FOR BEARING
FAULT DETECTION
CLASSIFIERS
• Support vector machines (SVM)
• Linear discriminant analysis (LDA)
• k-Nearest neighbors (k-NN)
• Decision Trees
• Artificial neural network (ANN)
DATA FUSION ALGORITHMS
•Data-level Fusion
•Feature-level Fusion
•Decision-level Fusion
Majority Voting Decision Fusion
Bayesian Belief Fusion
Multi-agent Fusion
41
