Title of Invention	A METHOD OF DETECTING AC MOTOR MISALIGNMENT FAULTS THROUGH MODIFIED K-MEANS
Abstract	This invention relates to a method of detecting AC motor (1) misalignment faults comprising the steps of: data acquisition (3) by sensing single-phase current (2) of a 3-phase motor, high speed digitizing the current signal (3), anti-aliasing the digitalized current signal near the nyquist frequency; feature extracting (4) by estimating real-time DC level for current signal for each digitized frame, and forming a feature set; feature reductioning by removing less affecting small frequency peaks (5) from the feature set and selecting time persistent large frequency peaks (8) (spectrum fingerprint) corresponding to key components in deriving motor condition at any instant time; fault sensing in learning state by storing spectrum fingerprints (6); detecting anomalies in frequency components as fault causing component (7) and reporting fault information (9) in graphical and textual form on computer screen and taking correcting measures for healthy operation of AC motor for prolonged hours and at all possible load variation via applying expert knowledge in finding source of fault.

Title of Invention

A METHOD OF DETECTING AC MOTOR MISALIGNMENT FAULTS THROUGH MODIFIED K-MEANS

Abstract

This invention relates to a method of detecting AC motor (1) misalignment faults comprising the steps of: data acquisition (3) by sensing single-phase current (2) of a 3-phase motor, high speed digitizing the current signal (3), anti-aliasing the digitalized current signal near the nyquist frequency; feature extracting (4) by estimating real-time DC level for current signal for each digitized frame, and forming a feature set; feature reductioning by removing less affecting small frequency peaks (5) from the feature set and selecting time persistent large frequency peaks (8) (spectrum fingerprint) corresponding to key components in deriving motor condition at any instant time; fault sensing in learning state by storing spectrum fingerprints (6); detecting anomalies in frequency components as fault causing component (7) and reporting fault information (9) in graphical and textual form on computer screen and taking correcting measures for healthy operation of AC motor for prolonged hours and at all possible load variation via applying expert knowledge in finding source of fault.

Full Text	-2- FIELD OF THE INVENTION The present invention relates to a method of detecting mechanical faults in AC motor beforehand through spectral analysis of motor stator current. BACKGROUND OF THE INVENTION Induction motors are used extensively in every industry. The electric motors and motor system (comprising of motor, bearings, gear arrangement, torque transmitting shaft) experience a wide range of electrical and mechanical faults. Particularly in metal rolling process electric motors experience a wide range of mechanical problems. Misalignment, be the most common problem, causes a decrease in motor efficiency, and misaligned machinery is more prone to failure due to increased loads on bearings, seals, and couplings. Misalignment is one of the common causes, which creates most of the mechanical faults and leads to motor vibration. 'Motor Vibration' is a common problem for metal roling application and can be extremely frustrating and hazardous and often result in imminent motor failure. Apart from this a poorly aligned machine and motor vibration can cost a steel industry 20% to 30% in process down time and poor quality of metal rolling. Motor fault prognosis is becoming indispensable in the field of metal rolling and other process areas due to increasing demand of motor system in production line and life forecasting of motors and predictive maintenance planning is essential for optimized production and quality maintenance. A prior US Patent No. US 5995910 entitled "Method and system for synthesizing vibration data" discloses a teaching of a system for training a neural network to synthesize vibration data relating to the operation of a machine. The system includes a first sensor operatively coupleable to the machine, the first sensor adapted to obtain at least one vibration signal relating to the operation of the machine. The system further includes a second sensor operatively coupleable to a power lead of the machine, the second sensor adapted to obtain at least one current signal relating to the operation of a machine. Additionally, the system includes a neural network operatively coupleable to the second sensor, the neural network being trainable to generate at least one synthesized vibration signal from the current signal, wherein the synthesized vibration signal is substantially equivalent to the vibration signal obtained from the first sensor. The cited document have not mentioned regarding the detection of fault due to misalignment of motor during operation. A prior US Patent US 2005007096 entitled "System and method for proactive motor wellness diagnosis", discloses a teaching of a power meter or overload relay including a housing and a plurality of sensors configured to monitor operation of a motor. A processor disposed within the housing and configured to receive operational feedback from the plurality of sensor and proactively determine an operational wellness of the motor from the operational feedback. This prior patent is very silent in regard to fault detection due to misalignment of motor in operating condition. A prior US Patent No. US 5049815 entitled "Spectral analysis of Induction motor current to detect rotor faults with reduced false alarms", discloses a teaching of detection of Motor fault in an induction motor by analysis of a frequency spectrum of the current drawn by the motor under test, wherein success of the fault analysis requires an accurate determination of the fundamental frequency of the motor current and the motor slip frequency, method and apparatus are disclosed for determining the fundamental and slip frequency values solely from analysis of the motor current spectra. The cited document have not focused anything about the detection of fault due to misalignment of motor during operational condition. The present invention is aimed to remove the above difficulties of existing practice of maintaining AC motor health and life by early detection of motor misalignment. The present invention is mainly applicable in rolling mill area and other place where high rating motors are used. DESCRIPTION OF THE INVENTION An object of the present invention is to provide a reliable warning system regarding the AC induction motors failure especially due to mechanical misalignments. According to another object the present invention provides an unsupervised induction motor fault detection method, which maximizes the detection accuracy by eliminating the dependence on expert-knowledge for deriving expert-system rules for fault finding using motor current signature analysis means. A further object of the invention is to provide an online fault diagnostic algorithm for identification of mechanical faults of induction motors by variable speed drive. A still another object of the invention is to provided wherein a new set of feature coefficients are extracted from spectral characteristics of stator current by modified "k-means" clustering algorithm. According to the invention there is provided a method of detecting AC motor misalignment faults through modified k-means clustering comprising the steps of: data acquisition by sensing single-phase current of a 3-phase motor through Hall Effect sensor on load condition, high speed digitizing the current signal, anti-aliasing the digitalized current signal near the nyquist frequency; feature extracting by estimating real-time DC level for current signal for each digitized frame using Fourier transform and substracting from the same to eliminate the shadowing of low frequency current components resulting Fourier spectrum reflecting the operating condition of the motor and forming a feature set; feature reductioning by removing less affecting small frequency peaks from the feature set using an adaptive cutoff and selecting time persistent large frequency peaks (spectrum fingerprint) corresponding to key components in deriving motor condition at any instant time; fault sensing in learning state by storing spectrum fingerprints; detecting anomalies in frequency components as fault causing component by comparison of real-time fingerprints with the cluster information from the data base; and reporting fault information in graphical and textual form on computer screen and taking correcting measures for healthy operation of AC motor for prolonged hours and at all possible load variation via applying expert knowledge in finding source of fault. The invention will be better understood by the following description with reference to the accompanying drawing in which Figure 1 represents the performance of the invention through four interconnected and integrated sections of the method. Figure 1a and 1b show the test dates of stator DC current for healthy and faulty motor sampled at 1 KHz respectively. Figures 2a and 2b show the frequency peaks datas for normalized Fourier Transform of DC corrected frame for Healthy and faulty induction motor respectively. 1. Data Acquisition: Hall effect sensor is used to sense single-phase current of a 3-phase, induction motor. High-speed digitization of the current signal (at 1KHz) is done using a 12-bit Advantech PCI-1713 card. Digitized signal is then anti-aliased using 3rd order butterworth filter near to the nyquist frequency. 2. Feature Extraction: Real-time DC level for current signal is estimated for each digitized frame using fourier transform (Eq.2) and subtracted from the same to eliminate the shadowing of low-frequency current components, which may appear during the low-speed operation of the motor. On this DC corrected frame, proposed method applies 50% overlapping with hanning window (Eq. 3), and re-computes Fourier transform Xf) Frequency peaks in this computed normalized Fourier spectrum reflects the operating condition of the motor and each such peak forms an element of the feature set where m=1, 2,.... number of frequency peaks detected in the normalized spectrum. 3. Feature Reduction: Less affecting 'small' frequency peaks are removed from the feature set using an adaptive cutoff frequency. After adaptive thresholding only those 'large' frequency peaks are selected for further processing, which remains persistent for past 24.58 seconds. Such reduced 'large and persistenf frequency peaks (that is frequency vs. magnitude) correspond to key components in deriving the motor condition at any instant of time. These frequency peaks are labeled as 'spectrum fingerprint' of motor current. 4. Fault Sensing: This stage is operated in two states - learning state and detection state. During learning state, the system stores spectrum fingerprints for healthy operation of induction motor for prolonged hours and at all possitle load variations. Cluster information is saved to the database once sufficient learning is over. Proposed method, detects anomalies in frequency components by comparison of real-time fingerprints with the cluster information from the database during detection state. As claimed, anomalies in the motor current spectrum are detected without applying any fault-specific expert-knowledge. The fault information are reported in graphical and textual form on computer screen. The invention as narrated hereinabove is illustrated with exemplary test results as shown in Figures 1a and 1b. Figures 2a and 2b show the test datas of stator current for healthy and faulty motor (broken rotor bar) respectively as sampled at 1KHz and frames of 4096 samples are created and represent sampled data and the D.C. corrected frame. Feature set or frequency peaks for normalized Fourier transform of D.C. corrected frame is shown as circles in Figure 2a and Figure 2b. "Large' persistent peaks above adaptive threshold are extracted as 'spectrum fingerprint', as shown with stars inside circle in those figures. The invention as narrated herein should not be read and construed in a restrictive manner as various modifications, are possible within the scope and limit of the invention as defined and encampussed within the appended claims. WE CLAIM 1. A method of detecting AC motor (1) misalignment faults through modified k- means clustering comprising the steps of: - data acquisition by sensing single-phase current (2) of a 3-phase motor through Hall Effect sensor on load condition, high speed digitizing the currert signal (3), anti-aliasing the digitalized current signal near the nyquist frequency; - feature extracting (4) by estimating real-time DC level for current signal (2) for each digitized frame using Fourier transform and substracting from the same to eliminate the shadowing of low frequency current components resulting Fourier spectrum reflecting the operating condition of the motor and forming a feature set; - feature reductioning (5) by removing less affecting small frequency peaks from the feature set using an adaptive cutoff and selecting time persistent large frequency peaks (spectrum fingerprint) corresponding to key components in deriving motor condition at any instant time; - fault sensing in learning state by storing (6) spectrum fingerprints; detecting anomalies in frequency components as fault causing component by comparison of real-time fingerprints (7) with the cluster information (8) from the date base; and - reporting (9) fault information in graphical and textual form on computer screen and taking correcting measures for healthy operation of AC motor for prolonged hours and at all possible load variation via applying expert knowledge in finding source of fault. 2. A method of detecting AC motor misalignment as claimed in claim 1, wherein for high speed digitization (3) of the current signal is made at 1 KHz and digitization is carried through using a 12 bit Advantech PCI - 1713 card. 3. A method of detecting AC motor misalignment as claimed in claim 1, wherein anti-aliasing (4) is carried through using 3rd order butterworth filter. 4. A method detecting AC motor misalignment as claimed in the preceeding claims wherein real DC level for current signal (2) is estimated for each digitized frame (3) using Fourier transform equation: Estimated DC level XfDc = Xf0 and subtracted from the same to eliminate the shadowing of low frequency current components as represented by the equation DC corrected frame Xf = (Xf - XfDc), Where f signifies the current frame Xf0 is the non-sinusoidal component Fourier transform { XfK, k = 0, 2, 3 (N - 1) Sampled signal Xf (n) n = 1, 2,... no of samples. 5. A method of detecting AC motor misalignment fault as claimed in claim 1, wherein the step of feature extraction (4) is carried through DC correction, windowing with 50% overlapping, averaging and computing Fourier transform (or FFT) X-f frequency peaks to form an element of feature set for each such peak represented by ξfm, where m = 1, 2 number of frequency peaks detected in normalized spectrum. 6. A method of detecting AC motor misalignment as claimed in claim 1, wherein adaptive cutoff (5) adaptive thresholding is carried through by considering components above a fixed db difference to the fundamental ( care of db difference of 'small' with respect to 'large' components. 7. A method of detecting AC motor misalignment as claimed in claim 6, wherein after adaptive thresholding only those Marge' frequency peaks are selected (6) for further processing, which are time persistent for last 24.58 seconds. 8. A method of detecting AC motor misalignment as claimed in the preceding claims wherein the step of fault sensing is carried through in two states- learning state and detecting state and clustering (5) is carried through modified K-means considering separate radii for frequency and magnitude axis. 9. A method of detecting AC motor misalignment as claimed in claim 1, wherein anomalies in the motor current spectrum are detected without applying any fault-specific expert knowledge with is used to classify the source of the faut once detected (7). 10.A method of detecting AC motor misalignment faults through modified k- means as herein described and illustrated. ABSTRACT A METHOD OF DETECTING AC MOTOR MISALIGNMENT FAULTS THROUGH MODIFIED K-MEANS This invention relates to a method of detecting AC motor (1) misalignment faults comprising the steps of: data acquisition (3) by sensing single-phase current (2) of a 3-phase motor, high speed digitizing the current signal (3), anti-aliasing the digitalized current signal near the nyquist frequency; feature extracting (4) by estimating real-time DC level for current signal for each digitized frame, and forming a feature set; feature reductioning by removing less affecting small frequency peaks (5) from the feature set and selecting time persistent large frequency peaks (8) (spectrum fingerprint) corresponding to key components in deriving motor condition at any instant time; fault sensing in learning state by storing spectrum fingerprints (6); detecting anomalies in frequency components as fault causing component (7) and reporting fault information (9) in graphical and textual form on computer screen and taking correcting measures for healthy operation of AC motor for prolonged hours and at all possible load variation via applying expert knowledge in finding source of fault.

Full Text

-2-
FIELD OF THE INVENTION
The present invention relates to a method of detecting mechanical faults in AC
motor beforehand through spectral analysis of motor stator current.
BACKGROUND OF THE INVENTION
Induction motors are used extensively in every industry. The electric motors and
motor system (comprising of motor, bearings, gear arrangement, torque
transmitting shaft) experience a wide range of electrical and mechanical faults.
Particularly in metal rolling process electric motors experience a wide range of
mechanical problems. Misalignment, be the most common problem, causes a
decrease in motor efficiency, and misaligned machinery is more prone to failure
due to increased loads on bearings, seals, and couplings. Misalignment is one of
the common causes, which creates most of the mechanical faults and leads to
motor vibration. 'Motor Vibration' is a common problem for metal roling
application and can be extremely frustrating and hazardous and often result in
imminent motor failure.

Apart from this a poorly aligned machine and motor vibration can cost a steel
industry 20% to 30% in process down time and poor quality of metal rolling. Motor
fault prognosis is becoming indispensable in the field of metal rolling and other
process areas due to increasing demand of motor system in production line and life
forecasting of motors and predictive maintenance planning is essential for optimized
production and quality maintenance.
A prior US Patent No. US 5995910 entitled "Method and system for synthesizing
vibration data" discloses a teaching of a system for training a neural network to
synthesize vibration data relating to the operation of a machine. The system
includes a first sensor operatively coupleable to the machine, the first sensor
adapted to obtain at least one vibration signal relating to the operation of the
machine. The system further includes a second sensor operatively coupleable to a
power lead of the machine, the second sensor adapted to obtain at least one current
signal relating to the operation of a machine. Additionally, the system includes a
neural network operatively coupleable to the second sensor, the neural network
being trainable to generate at least one synthesized vibration signal from the current
signal, wherein the synthesized vibration signal is substantially equivalent to the
vibration signal obtained from the first sensor.

The cited document have not mentioned regarding the detection of fault due to
misalignment of motor during operation.

A prior US Patent US 2005007096 entitled "System and method for proactive motor
wellness diagnosis", discloses a teaching of a power meter or overload relay
including a housing and a plurality of sensors configured to monitor operation of
a motor. A processor disposed within the housing and configured to receive

operational feedback from the plurality of sensor and proactively determine an
operational wellness of the motor from the operational feedback.
This prior patent is very silent in regard to fault detection due to misalignment of
motor in operating condition.
A prior US Patent No. US 5049815 entitled "Spectral analysis of Induction motor
current to detect rotor faults with reduced false alarms", discloses a teaching of
detection of Motor fault in an induction motor by analysis of a frequency spectrum
of the current drawn by the motor under test, wherein success of the fault analysis
requires an accurate determination of the fundamental frequency of the motor
current and the motor slip frequency, method and apparatus are disclosed for
determining the fundamental and slip frequency values solely from analysis of the
motor current spectra.
The cited document have not focused anything about the detection of fault due to
misalignment of motor during operational condition.

The present invention is aimed to remove the above difficulties of existing practice
of maintaining AC motor health and life by early detection of motor misalignment.

The present invention is mainly applicable in rolling mill area and other place where
high rating motors are used.
DESCRIPTION OF THE INVENTION
An object of the present invention is to provide a reliable warning system regarding
the AC induction motors failure especially due to mechanical misalignments.

According to another object the present invention provides an unsupervised
induction motor fault detection method, which maximizes the detection accuracy
by eliminating the dependence on expert-knowledge for deriving expert-system
rules for fault finding using motor current signature analysis means.
A further object of the invention is to provide an online fault diagnostic algorithm
for identification of mechanical faults of induction motors by variable speed drive.
A still another object of the invention is to provided wherein a new set of feature
coefficients are extracted from spectral characteristics of stator current by
modified "k-means" clustering algorithm.
According to the invention there is provided a method of detecting AC motor
misalignment faults through modified k-means clustering comprising the steps
of: data acquisition by sensing single-phase current of a 3-phase motor through
Hall Effect sensor on load condition, high speed digitizing the current signal,
anti-aliasing the digitalized current signal near the nyquist frequency; feature
extracting by estimating real-time DC level for current signal for each digitized
frame using Fourier transform and substracting from the same to eliminate the

shadowing of low frequency current components resulting Fourier spectrum
reflecting the operating condition of the motor and forming a feature set; feature
reductioning by removing less affecting small frequency peaks from the feature
set using an adaptive cutoff and selecting time persistent large frequency peaks
(spectrum fingerprint) corresponding to key components in deriving motor
condition at any instant time; fault sensing in learning state by storing spectrum
fingerprints; detecting anomalies in frequency components as fault causing
component by comparison of real-time fingerprints with the cluster information
from the data base; and reporting fault information in graphical and textual form
on computer screen and taking correcting measures for healthy operation of AC
motor for prolonged hours and at all possible load variation via applying expert
knowledge in finding source of fault.
The invention will be better understood by the following description with
reference to the accompanying drawing in which
Figure 1 represents the performance of the invention through four
interconnected and integrated sections of the method.

Figure 1a
and 1b show the test dates of stator DC current for healthy and
faulty motor sampled at 1 KHz respectively.
Figures 2a
and 2b show the frequency peaks datas for normalized Fourier
Transform of DC corrected frame for Healthy and
faulty induction motor respectively.
1. Data Acquisition: Hall effect sensor is used to sense single-phase
current of a 3-phase, induction motor. High-speed digitization of the current
signal (at 1KHz) is done using a 12-bit Advantech PCI-1713 card. Digitized signal
is then anti-aliased using 3rd order butterworth filter near to the nyquist
frequency.
2. Feature Extraction: Real-time DC level for current signal is estimated for
each digitized frame using fourier transform (Eq.2) and subtracted from the
same to eliminate the shadowing of low-frequency current components, which
may appear during the low-speed operation of the motor.

On this DC corrected frame, proposed method applies 50% overlapping with
hanning window (Eq. 3), and re-computes Fourier transform Xf) Frequency peaks
in this computed normalized Fourier spectrum reflects the operating condition of
the motor and each such peak forms an element of the feature set where
m=1, 2,.... number of frequency peaks detected in the normalized spectrum.

3. Feature Reduction: Less affecting 'small' frequency peaks are removed
from the feature set using an adaptive cutoff frequency. After adaptive
thresholding only those 'large' frequency peaks are selected for further

processing, which remains persistent for past 24.58 seconds. Such reduced
'large and persistenf frequency peaks (that is frequency vs. magnitude)
correspond to key components in deriving the motor condition at any instant of
time. These frequency peaks are labeled as 'spectrum fingerprint' of motor
current.
4. Fault Sensing: This stage is operated in two states - learning state and
detection state. During learning state, the system stores spectrum fingerprints
for healthy operation of induction motor for prolonged hours and at all possitle
load variations. Cluster information is saved to the database once sufficient
learning is over.
Proposed method, detects anomalies in frequency components by comparison of
real-time fingerprints with the cluster information from the database during
detection state. As claimed, anomalies in the motor current spectrum are
detected without applying any fault-specific expert-knowledge.
The fault information are reported in graphical and textual form on computer
screen.

The invention as narrated hereinabove is illustrated with exemplary test results
as shown in Figures 1a and 1b. Figures 2a and 2b show the test datas of stator
current for healthy and faulty motor (broken rotor bar) respectively as sampled
at 1KHz and frames of 4096 samples are created and represent sampled data
and the D.C. corrected frame.
Feature set or frequency peaks for normalized Fourier transform of D.C.
corrected frame is shown as circles in Figure 2a and Figure 2b. "Large'
persistent peaks above adaptive threshold are extracted as 'spectrum
fingerprint', as shown with stars inside circle in those figures.
The invention as narrated herein should not be read and construed in a
restrictive manner as various modifications, are possible within the scope and
limit of the invention as defined and encampussed within the appended claims.

WE CLAIM
1. A method of detecting AC motor (1) misalignment faults through modified k-
means clustering comprising the steps of:
- data acquisition by sensing single-phase current (2) of a 3-phase motor
through Hall Effect sensor on load condition, high speed digitizing the currert
signal (3), anti-aliasing the digitalized current signal near the nyquist
frequency;
- feature extracting (4) by estimating real-time DC level for current signal (2)
for each digitized frame using Fourier transform and substracting from the
same to eliminate the shadowing of low frequency current components
resulting Fourier spectrum reflecting the operating condition of the motor and
forming a feature set;
- feature reductioning (5) by removing less affecting small frequency peaks
from the feature set using an adaptive cutoff and selecting time persistent
large frequency peaks (spectrum fingerprint) corresponding to key
components in deriving motor condition at any instant time;

- fault sensing in learning state by storing (6) spectrum fingerprints; detecting
anomalies in frequency components as fault causing component by
comparison of real-time fingerprints (7) with the cluster information (8) from
the date base; and
- reporting (9) fault information in graphical and textual form on computer
screen and taking correcting measures for healthy operation of AC motor for
prolonged hours and at all possible load variation via applying expert
knowledge in finding source of fault.

2. A method of detecting AC motor misalignment as claimed in claim 1, wherein
for high speed digitization (3) of the current signal is made at 1 KHz and
digitization is carried through using a 12 bit Advantech PCI - 1713 card.
3. A method of detecting AC motor misalignment as claimed in claim 1, wherein
anti-aliasing (4) is carried through using 3rd order butterworth filter.

4. A method detecting AC motor misalignment as claimed in the preceeding
claims wherein real DC level for current signal (2) is estimated for each
digitized frame (3) using Fourier transform equation: Estimated DC level XfDc
= Xf0 and subtracted from the same to eliminate the shadowing of low
frequency current components as represented by the equation
DC corrected frame Xf = (Xf - XfDc),
Where f signifies the current frame
Xf0 is the non-sinusoidal component
Fourier transform { XfK, k = 0, 2, 3 (N - 1)
Sampled signal Xf (n) n = 1, 2,... no of samples.
5. A method of detecting AC motor misalignment fault as claimed in claim 1,
wherein the step of feature extraction (4) is carried through DC correction,
windowing with 50% overlapping, averaging and computing Fourier
transform (or FFT) X-f frequency peaks to form an element of feature set for
each such peak represented by ξfm, where m = 1, 2 number of frequency
peaks detected in normalized spectrum.

6. A method of detecting AC motor misalignment as claimed in claim 1, wherein
adaptive cutoff (5) adaptive thresholding is carried through by considering
components above a fixed db difference to the fundamental ( care of db difference of 'small' with respect to 'large' components.
7. A method of detecting AC motor misalignment as claimed in claim 6, wherein
after adaptive thresholding only those Marge' frequency peaks are selected (6)
for further processing, which are time persistent for last 24.58 seconds.
8. A method of detecting AC motor misalignment as claimed in the preceding
claims wherein the step of fault sensing is carried through in two states-
learning state and detecting state and clustering (5) is carried through
modified K-means considering separate radii for frequency and magnitude
axis.

9. A method of detecting AC motor misalignment as claimed in claim 1, wherein
anomalies in the motor current spectrum are detected without applying any
fault-specific expert knowledge with is used to classify the source of the faut
once detected (7).
10.A method of detecting AC motor misalignment faults through modified k-
means as herein described and illustrated.

ABSTRACT

A METHOD OF DETECTING AC MOTOR MISALIGNMENT FAULTS THROUGH
MODIFIED K-MEANS
This invention relates to a method of detecting AC motor (1) misalignment faults
comprising the steps of: data acquisition (3) by sensing single-phase current (2) of a
3-phase motor, high speed digitizing the current signal (3), anti-aliasing the
digitalized current signal near the nyquist frequency; feature extracting (4) by
estimating real-time DC level for current signal for each digitized frame, and forming
a feature set; feature reductioning by removing less affecting small frequency peaks
(5) from the feature set and selecting time persistent large frequency peaks (8)
(spectrum fingerprint) corresponding to key components in deriving motor condition
at any instant time; fault sensing in learning state by storing spectrum fingerprints
(6); detecting anomalies in frequency components as fault causing component (7)
and reporting fault information (9) in graphical and textual form on computer screen
and taking correcting measures for healthy operation of AC motor for prolonged
hours and at all possible load variation via applying expert knowledge in finding
source of fault.

Documents:

00072-kol-2007-correspondence-1.1.pdf

00072-kol-2007-form-18.pdf

0072-kol-2007 abstract.pdf

0072-kol-2007 assignment.pdf

0072-kol-2007 claims.pdf

0072-kol-2007 correspondence others.pdf

0072-kol-2007 description(complete).pdf

0072-kol-2007 drawings.pdf

0072-kol-2007 form-1.pdf

0072-kol-2007 form-2.pdf

0072-kol-2007 form-3.pdf

7-(27-12-2011)-2-KOL-2007-CORRESPONDENCE.pdf

7-(27-12-2011)-2-KOL-2007-FORM-13.pdf

72-(27-12-2011)-2-KOL-2007-CORRESPONDENCE.pdf

72-(27-12-2011)-2-KOL-2007-FORM-13.pdf

72-KOL-2007-(02-05-2012)-CORRESPONDENCE.pdf

72-KOL-2007-(27-12-2011)-ABSTRACT.pdf

72-KOL-2007-(27-12-2011)-AMANDED CLAIMS.pdf

72-KOL-2007-(27-12-2011)-DESCRIPTION (COMPLETE).pdf

72-KOL-2007-(27-12-2011)-DRAWINGS.pdf

72-KOL-2007-(27-12-2011)-EXAMINATION REPORT REPLY RECIEVED.PDF

72-KOL-2007-(27-12-2011)-FORM-1.pdf

72-KOL-2007-(27-12-2011)-FORM-2.pdf

72-KOL-2007-(27-12-2011)-GPA.pdf

72-KOL-2007-(27-12-2011)-OTHERS.pdf

72-KOL-2007-CORRESPONDENCE 1.1.pdf

72-KOL-2007-CORRESPONDENCE 1.2.pdf

72-KOL-2007-EXAMINATION REPORT.pdf

72-KOL-2007-FORM 13.pdf

72-KOL-2007-FORM 18.pdf

72-KOL-2007-FORM 3.pdf

72-KOL-2007-GPA.pdf

72-KOL-2007-GRANTED-ABSTRACT.pdf

72-KOL-2007-GRANTED-CLAIMS.pdf

72-KOL-2007-GRANTED-DESCRIPTION (COMPLETE).pdf

72-KOL-2007-GRANTED-DRAWINGS.pdf

72-KOL-2007-GRANTED-FORM 1.pdf

72-KOL-2007-GRANTED-FORM 2.pdf

72-KOL-2007-GRANTED-SPECIFICATION.pdf

72-KOL-2007-OTHERS.pdf

72-KOL-2007-REPLY TO EXAMINATION REPORT.pdf

abstract-00072-kol-2007.jpg

« Previous Patent

Next Patent »

Patent Number

254832

Indian Patent Application Number

72/KOL/2007

PG Journal Number

52/2012

Publication Date

28-Dec-2012

Grant Date

24-Dec-2012

Date of Filing

19-Jan-2007

Name of Patentee

TATA STEEL LIMITED

Applicant Address

RESEARCH AND DEVELOPMENT AND SCIENTIFIC SERVICES DIVISION JAMSHEDPUR-831 001 INDIA

Inventors:

#	Inventor's Name	Inventor's Address
1	CHAUDHURY, S.B.	RESEARCH AND DEVELOPMENT AND SCIENTIFIC SERVICES DIVISION JAMSHEDPUR-831001
2	KUMAR, SANJEEV	RESEARCH AND DEVELOPMENT AND SCIENTIFIC SERVICES DIVISION, JAMSHEDPUR-831001
3	GUPTA, SACHIN	RESEARCH AND DEVELOPMENT AND SCIENTIFIC SERVICES DIVISION, JAMSHEDPUR-831001

PCT International Classification Number

H02K29/00

PCT International Application Number

N/A

PCT International Filing date

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1			NA