U.S. patent application number 15/352703 was filed with the patent office on 2017-03-02 for method for verifying bad pattern in time series sensing data and apparatus thereof.
This patent application is currently assigned to SAMSUNG SDS CO., LTD.. The applicant listed for this patent is SAMSUNG SDS CO., LTD.. Invention is credited to Dae Jung AHN, WooYoung JUNG, Dae Hong SEO, Kae Young SHIN.
Application Number | 20170060664 15/352703 |
Document ID | / |
Family ID | 52020342 |
Filed Date | 2017-03-02 |
United States Patent
Application |
20170060664 |
Kind Code |
A1 |
SHIN; Kae Young ; et
al. |
March 2, 2017 |
METHOD FOR VERIFYING BAD PATTERN IN TIME SERIES SENSING DATA AND
APPARATUS THEREOF
Abstract
A method for verifying bad pattern in time series sensing data
by calculating a bad pattern error rate, which can be applied to
the time series sensing data measured and produced from a
predetermined sensor provided in predetermined equipment, and an
apparatus thereof are provided. The method includes receiving
information on the bad pattern applied to time series sensing data
measured by a suspicious sensor, accessing the time series sensing
data of each product, generated by the suspicious sensor during a
verification period, calculating similarity measures between the
bad pattern based on the bad pattern information and the time
series sensing data for each product, and calculating an error rate
of the bad pattern based on the similarity measures.
Inventors: |
SHIN; Kae Young; (Yongin-si,
KR) ; AHN; Dae Jung; (Yongin-si, KR) ; SEO;
Dae Hong; (Seoul, KR) ; JUNG; WooYoung;
(Yongin-si, KR) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
SAMSUNG SDS CO., LTD. |
Gangnam-Gu |
|
KR |
|
|
Assignee: |
SAMSUNG SDS CO., LTD.
Gangnam-Gu
KR
|
Family ID: |
52020342 |
Appl. No.: |
15/352703 |
Filed: |
November 16, 2016 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
14306967 |
Jun 17, 2014 |
9547544 |
|
|
15352703 |
|
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06F 11/0772 20130101;
G06F 11/076 20130101; G06F 11/079 20130101; G01R 31/3171 20130101;
G06F 11/0706 20130101; G06F 11/0703 20130101; G06F 11/0736
20130101; H03M 13/015 20130101 |
International
Class: |
G06F 11/07 20060101
G06F011/07 |
Foreign Application Data
Date |
Code |
Application Number |
Jun 18, 2013 |
KR |
10-2013-0069678 |
Claims
1. A method for monitoring sensing data of equipment, the method
comprising: receiving information on a bad pattern applied to
sensing data measured from a sensor; receiving sensing data
information from the sensor; calculating similarity measures of an
Euclidean ratio, a correlation and a trend similarity between the
bad pattern based on bad pattern information and the sensing data,
each of the Euclidean ratio, the correlation and the trend
similarity having a value in a range between 0, which means
non-similarity, and 1, which means sameness; and generating a
warning signal when a calculated value of at least one from among
the Euclidean ratio, the correlation and the trend similarity is 1,
wherein the trend similarity is a measured value pattern one of
increasing and decreasing according to a passage of time.
2. Production equipment comprising: one or more sensors generating
sensing data; a sensing data stream receiving unit configured to
receive a sensing data stream measured from the one or more
sensors; a bad pattern information receiving unit configured to
receive information on a bad pattern applied to sensing data
measured by each of the one or more sensors; a similarity measure
calculator configured to calculate similarity measures of an
Euclidean ratio, a correlation and a trend similarity between the
bad pattern based on the bad pattern information and the sensing
data, each of the Euclidean ratio, the correlation and the trend
similarity having a value in a range between 0, which means
non-similarity, and 1, which means sameness; and a warning signal
output unit configured to generate a warning signal when a
calculated value of at least one from among the Euclidean ratio,
the correlation and the trend similarity is 1.
3. The production equipment of claim 2, further comprising a
stoppage controller configured to receive the warning signal from
the warning signal output unit and generate a control signal for
controlling a production operation to be stopped.
4. The production equipment of claim 2, wherein the sensing data
based on the bad pattern information has a pattern including
measured values in a range between an upper limit and a lower limit
of a predefined measured value.
5. A reference pattern detecting apparatus comprising: a sensing
data stream receiving unit configured to receive a sensing data
stream measured from one or more sensors; a reference pattern
information receiving unit configured to receive information on a
reference pattern applied to the sensing data generated from each
of the one or more sensors; and a similarity measure calculator
configured to calculate and output similarity measures of an
Euclidean ratio, a correlation and a trend similarity between the
reference pattern based on reference pattern information and the
sensing data, each of the Euclidean ratio, the correlation and the
trend similarity having a value in a range between 0, which means
non-similarity, and 1, which means sameness.
6. The reference pattern detecting apparatus of claim 5, further
comprising a warning signal output unit configured to generate and
output a warning signal when the value of at least one from among
the Euclidean ratio, the correlation and the trend similarity,
calculated by the similarity measure calculator, is 1.
7. The reference pattern detecting apparatus of claim 5, wherein
the one or more sensors include a first sensor measuring energy
consumption of a building.
8. The reference pattern detecting apparatus of claim 6, wherein
the one or more sensors include a first sensor measuring building
energy consumption and a second sensor measuring an internal
environment parameter of a building, the reference pattern
information receiving unit receives first reference pattern
information applied to the first sensor and second reference
pattern information applied to the second sensor, the similarity
measure calculator calculates and outputs similarity measures of an
Euclidean ratio, a correlation and a trend similarity between a
first reference pattern based on the first reference pattern
information and the sensing data of the first sensor, and
calculates and outputs similarity measures of the Euclidean ratio,
the correlation and the trend similarity between a second reference
pattern based on the second reference pattern information and the
sensing data of the second sensor, and the warning signal output
unit generates and outputs a warning signal when the calculated
value of at least one from among the Euclidean ratio, the
correlation and the trend similarity of the first sensor is 1 and
the calculated value of at least one from among the Euclidean
ratio, the correlation and the trend similarity of the second
sensor is 1.
Description
CROSS-REFERENCE TO RELATED APPLICATION
[0001] [1] This is a divisional of U.S. patent application Ser. No.
14/306,967, filed Jun. 17, 2014, which claims priority from Korean
Patent Application No. 10-2013-0069678 filed on Jun. 18, 2013 in
the Korean Intellectual Property Office, and all the benefits
accruing therefrom under 35 U.S.C. 119, the contents of which in
its entirety are herein incorporated by reference.
BACKGROUND
[0002] 1. Field of the Invention
[0003] The present invention relates to a method for verifying bad
pattern in time series sensing data and an apparatus thereof. More
particularly, the present invention relates to a method for
verifying bad pattern in time series sensing data by calculating a
bad pattern error rate, which can be applied to time series sensing
data measured and produced from a predetermined sensor provided in
predetermined equipment, and an apparatus thereof.
[0004] 2. Description of the Related Art
[0005] In producing a product, it is very important to maintain a
product quality while achieve a production yield. Therefore, in
order to increase the production yield, it is necessary to find and
diagnose an abnormal condition of the progress of a process or
equipment at an early stage. To this end, fault detection &
classification (FDC) technology has been introduced to monitor the
condition of the process, to detect probable faults and to classify
types of the faults.
SUMMARY
[0006] The present invention provides an apparatus and method for
verifying bad pattern in time series sensing data.
[0007] The present invention also provides an apparatus and method
for verifying reliability of bad pattern by calculating an error
rate of the bad pattern by comparing a predetermined bad pattern
with time series sensing data at the time of producing each
product.
[0008] The present invention also provides an apparatus and method
for verifying bad pattern, to which a method for calculating
similarity measures between time series sensing data generated in
the course of producing a product and the bad pattern specialized
to the time series sensing data.
[0009] The present invention also provides an apparatus and method
for verifying bad pattern by calculating a bad pattern error rate
based on the failure ratio of products having time series sensing
data determined to be similar to the bad pattern.
[0010] The present invention also provides a method and equipment
for determining a fault by calculating a similarity measure between
a predetermined bad pattern and time series sensing data measured
by a sensor provided within the equipment.
[0011] The present invention also provides production equipment
having one or more sensors, which determines a similarity measure
between the time series sensing data and a predetermined bad
pattern on a real time basis and generates a warning signal
according to the determination result.
[0012] The present invention also provides production equipment
having one or more sensors, which determines similarity between the
time series sensing data and a predetermined bad pattern on a real
time basis and generates a warning signal according to the
determination result, and a method for monitoring sensor-measured
data of the production equipment.
[0013] The present invention also provides a bad pattern detecting
apparatus, which receives time series sensing data from production
equipment having one or more sensors, determines similarity between
the time series sensing data and a predetermined bad pattern on a
real time basis and generates a warning signal according to the
determination result, and a method therefore.
[0014] These and other objects of the present invention will be
described in or be apparent from the following description of the
exemplary embodiments.
[0015] According to an aspect of the present invention, there is
provided a method for verifying bad pattern in time series sensing
data, the method comprising receiving bad pattern information
applied to sensing data measured by a sensor, accessing the sensing
data of each product, generated by the sensor, calculating
similarity measures between the bad pattern based on the bad
pattern information and the accessed sensing data, and calculating
an error rate of the bad pattern based on the similarity
measures.
[0016] According to another aspect of the present invention, there
is provided an apparatus for verifying a bad pattern in sensing
data, the apparatus comprising a verification parameter receiving
unit configured to receive verification parameters including
information on the bad pattern applied to sensing data of a sensor
and information on a verification method, a sensing data extracting
unit configured to access the sensing data of each product,
generated by the sensor based on verification method information, a
similarity measure calculating unit configured to calculate
similarity measures between the bad pattern based on the bad
pattern information and the sensing data, for each product, and a
bad pattern verification unit configured to calculate an error rate
of the bad pattern based on the similarity measures.
[0017] According to still another aspect of the present invention,
there is provided a method for monitoring sensing data of
equipment, the method comprising receiving information on a bad
pattern applied to sensing data measured from a sensor, receiving
sensing data information from the sensor, calculating similarity
measures of an Euclidean ratio, a correlation and a trend
similarity between the bad pattern based on bad pattern information
and the sensing data, each of the Euclidean ratio, the correlation
and the trend similarity having a value in a range between 0, which
means non-similarity, and 1, which means sameness, and generating a
warning signal when a calculated value of at least one from among
the Euclidean ratio, the correlation and the trend similarity is 1,
wherein the trend similarity is a measured value pattern one of
increasing and decreasing according to a passage of time.
[0018] According to a further aspect of the present invention,
there is provided production equipment comprising one or more
sensors generating sensing data, a sensing data stream receiving
unit configured to receive a sensing data stream measured from the
one or more sensors, a bad pattern information receiving unit
configured to receive information on a bad pattern applied to
sensing data measured by each of the one or more sensors, a
similarity measure calculator configured to calculate similarity
measures of an Euclidean ratio, a correlation and a trend
similarity between the bad pattern based on the bad pattern
information and the sensing data, each of the Euclidean ratio, the
correlation and the trend similarity having a value in a range
between 0, which means non-similarity, and 1, which means sameness,
and a warning signal output unit configured to generate a warning
signal when a calculated value of at least one from among the
Euclidean ratio, the correlation and the trend similarity is 1.
[0019] According to a further aspect of the present invention,
there is provided a reference pattern detecting apparatus
comprising a sensing data stream receiving unit configured to
receive a sensing data stream measured from one or more sensors, a
reference pattern information receiving unit configured to receive
information on a reference pattern applied to the sensing data
generated from each of the one or more sensors, and a similarity
measure calculator configured to calculate and output similarity
measures of an Euclidean ratio, a correlation and a trend
similarity between the reference pattern based on reference pattern
information and the sensing data, each of the Euclidean ratio, the
correlation and the trend similarity having a value in a range
between 0, which means non-similarity, and 1, which means
sameness.
[0020] As described above, according to the present invention, the
reliability of a predetermined bad pattern can be verified using an
actual production result, thereby obtaining bad pattern with high
reliability.
[0021] In addition, time series sensing data generated from each of
one or more sensors is analyzed to determine whether the time
series sensing data is similar to a predetermined bad pattern, and
if yes, a warning signal is generated, thereby determining a value
of the time series sensing data is within a predetermined range and
monitoring whether the sensor-measured data of the production
equipment represents the predetermined bad pattern.
[0022] Further, according to the present invention, appropriate
measured values can be taken in a case where generation of time
series sensing data similar to the bad pattern is detected during
production equipment monitoring process.
BRIEF DESCRIPTION OF THE DRAWINGS
[0023] The above and other features and advantages of the present
invention will become more apparent by describing in detail
exemplary embodiments thereof with reference to the attached
drawings in which:
[0024] FIG. 1 is a flowchart of a method for verifying bad pattern
in time series sensing data according to an embodiment of the
present invention;
[0025] FIGS. 2A to 2I illustrate examples of bad pattern in time
series sensing data;
[0026] FIG. 3 is a conceptual diagram illustrating time series
sensing data of a product used in verifying bad pattern in the bad
pattern verifying method shown in FIG. 1;
[0027] FIG. 4 is a conceptual diagram illustrating a process for
calculating a bad pattern error rate in the bad pattern verifying
method shown in FIG. 1;
[0028] FIG. 5 is a flowchart illustrating a process of
pre-processing time series sensing data of a product used in
verifying bad pattern in the bad pattern verifying method shown in
FIG. 1;
[0029] FIG. 6 is a conceptual diagram illustrating a divisional
compression process in the pre-processing process shown in FIG.
5;
[0030] FIG. 7 is a conceptual diagram illustrating a normalization
process in the pre-processing method shown in FIG. 5;
[0031] FIGS. 8A and 8B are conceptual diagrams illustrating a
symbolization process in the pre-processing process shown in FIG.
5;
[0032] FIG. 9 illustrates a verification result by the bad pattern
verifying method shown in FIG. 1;
[0033] FIG. 10 is a schematic diagram of a bad pattern verification
system according to another embodiment of the present
invention;
[0034] FIG. 11 is a block diagram of a bad pattern verification
system according to still another embodiment of the present
invention;
[0035] FIG. 12 is another schematic diagram of the bad pattern
verification system shown in FIG. 11;
[0036] FIGS. 13A and 13B illustrate results of monitoring sensing
data in production equipment according to still another embodiment
of the present invention and in conventional production
equipment;
[0037] FIG. 14 is a flowchart of a monitoring process of sensing
data generated from a sensor of production equipment according to
still another embodiment of the present invention;
[0038] FIG. 15 is a block diagram of production equipment according
to still another embodiment of the present invention;
[0039] FIG. 16 is a schematic diagram of a reference pattern
detection system according to still another embodiment of the
present invention;
[0040] FIG. 17 is a schematic diagram of a reference pattern
detection system according to still another embodiment of the
present invention;
[0041] FIG. 18 is a first block diagram of a reference pattern
detection system according to still another embodiment of the
present invention; and
[0042] FIG. 19 is a second block diagram schematic diagram of a
reference pattern detection system according to still another
embodiment of the present invention.
DETAILED DESCRIPTION OF THE EMBODIMENTS
[0043] Advantages and features of the present invention and methods
of accomplishing the same may be understood more readily by
reference to the following detailed description of exemplary
embodiments and the accompanying drawings. The present invention
may, however, be embodied in many different forms and should not be
construed as being limited to the embodiments set forth herein.
Rather, these embodiments are provided so that this disclosure will
be thorough and complete and will fully convey the concept of the
invention to those skilled in the art, and the present invention
will only be defined by the appended claims. Like reference
numerals refer to like elements throughout the specification.
[0044] The terminology used herein is for the purpose of describing
particular embodiments only and is not intended to be limiting of
the invention. As used herein, the singular forms "a", "an" and
"the" are intended to include the plural forms as well, unless the
context clearly indicates otherwise. It will be further understood
that the terms "comprises" and/or "comprising," when used in this
specification, specify the presence of stated features, integers,
steps, operations, elements, and/or components, but do not preclude
the presence or addition of one or more other features, integers,
steps, operations, elements, components, and/or groups thereof.
[0045] Throughout the specification of the present invention, the
term "production equipment" may be used to refer to particular
equipment of a particular process. For example, assuming that
certain products are produced through a PHOTO process, a DRY
process and a DEPOSITION process, one among various equipments of
each process or one of multiple production lines of each process
having multiple equipments sequentially arranged may be allocated
to each of the products to allow the product to pass through the
allocated equipment or production line. The production equipment
refers to one among multiple equipments constituting each
process.
[0046] In addition, throughout the specification of the present
invention, the term "time series sensing data" may mean data having
measured values generated by a sensor incorporated in the
production equipment serially recorded according to the passage of
time.
[0047] FIG. 1 is a flowchart of a method for verifying bad pattern
in time series sensing data according to an embodiment of the
present invention.
[0048] First, parameters for verification are received from a user
terminal (S100). The verification parameters include information
required for verification. For example, the verification parameters
may include at least one of information on data of bad pattern and
suspicious equipment and a suspicious sensor to which the bad
pattern are applied, information on a verification period for
extracting time series sensing data used in verifying the bad
pattern, and information used in pre-processing the time series
sensing data used in verifying the bad pattern. Here, the
"suspicious" equipment and the "suspicious" sensor may be so
designated because they are suspected to cause a fault.
[0049] The bad pattern data may be the time series sensing data of
the bad pattern, address information for accessing the time series
sensing data of the bad pattern, or pre-processed data of the time
series sensing data of the bad pattern.
[0050] The term "bad pattern" used in the specification of the
present invention will now be briefly described with reference to
FIGS. 2A to 2I. A temporary or usual fault generated in the
production equipment or a fault generated under a particular
condition may be reflected on values measured by a sensor provided
in the production equipment, and time series sensing data of the
measured values may have a predetermined pattern indicating the
fault.
[0051] FIG. 2A illustrates a normal time series sensing data
pattern of a particular sensor provided in particular equipment and
FIGS. 2B to 2I illustrate "bad pattern" in cases where different
kinds of faults are generated. FIG. 2B illustrates fault of a
process due to rapid progress. FIG. 2C illustrates fault of a
process due to slow progress. FIG. 2D illustrates late start and
early shift of a process. FIG. 2E illustrates late start and late
progress/shift of a process. FIG. 2F illustrates the overall
progress is delayed. FIG. 2G illustrates that sensor-measured
values are drifted upward/downward as a whole while a process time
is normal. FIG. 2H illustrates that sensor-measured values are
deviated from a normal range in a particular point in a process for
a short period of time. FIG. 2I illustrates that sensor-measured
values are deviated from a normal range in a particular point in a
process for a long period of time.
[0052] In the specification of the present invention, the term "bad
pattern" is used to mean patterns in time series sensing data
measured by a sensor provided within the production equipment to
indicate temporary or usual faults generated in the production
equipment or faults generated under a particular condition. For
example, the "bad pattern" may include combinations of one or two
or more of the bad pattern shown in FIGS. 2B to 2I.
[0053] Referring back to FIG. 1, after the verification parameters
are received, time series sensing data of suspicious equipment for
a suspicious sensor is extracted (S110). When only particular
suspicious equipment is designated in the verification parameters
without designating a particular suspicious sensor, time series
sensing data for all sensors provided in the particular suspicious
equipment are extracted. The time series sensing data to be
extracted will now be described briefly with reference to FIG. 3.
The time series sensing data may be extracted from, for example,
fault detection & classification (FDC) data.
[0054] FIG. 3 is a conceptual diagram illustrating time series
sensing data of a product used in verifying bad pattern in the bad
pattern verifying method shown in FIG. 1.
[0055] As shown in FIG. 3, a product may be produced through
multiple processes from a Fab-in process to a Fab-out process.
Specifically, when equipment in charge of a particular process is
designated as suspicious equipment, the time series sensing data
generated by the suspicious equipment are collected for a
predetermined period to be used in verifying bad pattern. FIG. 3
illustrates time series sensing data generated by a sensor provided
in suspicious equipment in a process `0405` are collected for a
period from one month earlier to one month later than the present
time. The collected time series sensing data may include time
series sensing data pieces measured whenever processes, in which
each product passes through the suspicious equipment, are carried
out. A point in time for extracting the time series sensing data
may be based on information on a verification period for extracting
the time series sensing data used in verifying the bad pattern
included in the verification parameters.
[0056] Referring back to FIG. 1, after the extracting of the time
series sensing data is completed (S110), in order to calculate
similarity measures between the extracted time series sensing data
and the bad pattern, the extracted time series sensing data may be
pre-processed (S120). The bad pattern and the extracted time series
sensing data should be pre-processed in the same manner. Therefore,
in a case where pre-processed bad pattern are not indicated by the
bad pattern information included in the verification parameters,
the bad pattern are pre-processed in the same manner with the
extracted time series sensing data. The pre-processing according to
the present invention will later be described in more detail with
reference to FIG. 5.
[0057] Next, after the pre-processing (S120), similarity measures
between the bad pattern and the pre-processed extracted time series
sensing data are calculated (S130). For example, when FDC data of
the particular suspicious equipment are collected for the last one
month and 1000 products have pass through the suspicious equipment
for the last one month, there exists time series sensing data for
each of the 1000 products, and similarity measures between the bad
pattern and each of 1000 time series sensing data pieces are
calculated. The respective 1000 time series sensing data pieces may
be arranged on the basis of the similarity measures.
[0058] After the calculating of the similarity measures is
completed (S130), fault determining information of a product
corresponding to each of the time series sensing data pieces is
accessed, and the calculated similarity measures are compared with
good/bad determination results (S140). That is to say, a number of
cases of being determined as a good product, and the time series
sensing data of the product being determined to be similar to the
bad pattern. According to the result of the comparing (S140), a bad
pattern error rate is calculated (S150).
[0059] The error rate may be calculated as a value obtained by
dividing a number of products determined as good products even if
they are determined to be similar to the bad pattern by a number of
products having time series sensing data determined to be similar
to the bad pattern.
[0060] FIG. 4 is a conceptual diagram illustrating a method for
calculating a bad pattern error rate in the bad pattern verifying
method shown in FIG. 1.
[0061] As shown in FIG. 4, time series sensing data pieces for 9
products are extracted, and among that, there exist 5 time series
sensing data pieces similar to the bad pattern, as represented by
graphical representations (1), (5), (6), (8) and (9). However one
time series sensing data piece (5) is determined as a good product.
Therefore, the calculated error rate may be 1/5 (=20%).
[0062] According to the present embodiment of the present
invention, the error rate is calculated as a ratio of the number of
the bad pattern suspected as patterns causing faults to the number
of products determined to be similar to the bad pattern and
determined as good products. This embodiment numerically calculates
the reliability of the bad pattern using time series sensing data
generated during actual production as verification data.
[0063] In order to accurate calculate the error rate, it is
important to accurately calculate similarity measures between the
bad pattern and the time series sensing data of each product. In
addition, in order to accurately calculate similarity measures
between the bad pattern and the time series sensing data of each
product, an appropriate pre-processing process should be performed.
Hereinafter, the pre-processing process according to the present
embodiment will be described with reference to FIGS. 5 to 8B.
[0064] FIG. 5 is a flowchart illustrating a process of
pre-processing time series sensing data of a product used in
verifying bad pattern in the bad pattern verifying method shown in
FIG. 1.
[0065] As shown in FIG. 5, the time series sensing data may be
pre-processed through processes of divisional compression (S122),
normalization (S124) and symbolization (S126).
[0066] First, although not shown in FIG. 5, prior to the divisional
compression (S122), compensation of missing values may be
performed. Since there are missing parts in measured values of the
time series sensing data, the missing values can be compensated
from other measured values using interpolation.
[0067] In the divisional compression of the time series sensing
data (S122), the measured value data sensed on a short time basis,
for example, in units of seconds, are divided into a predetermined
number (W) of sections and only one representative value is stored
in each section. For example, when 10 measured value data pieces
each section is compressed into one representative value data
section, the time series sensing data of the measured values may be
compressed to have a tenth of a data size.
[0068] The representative value of each section may be a mean value
of the measured values in the section.
[0069] The predetermined number (W) of sections may be a value
included in the verification parameters.
[0070] Assuming that the divided sections are denoted by C.sub.1,
C.sub.2, . . . , C.sub.w and values of sensor-measured time series
sensing data are denoted by D.sub.1, . . . , D.sub.n, the divided
sections may be defined in Equation (1):
C i = ( n w ( i - 1 ) + 1 , n w i ) ( 1 ) ##EQU00001##
[0071] Table 1 demonstrates a case where time series sensing data
including 10 measured value data are divided into 6 sections, and
Table 2 demonstrates a case where time series sensing data
including 10 measured value data are divided into 3 sections.
TABLE-US-00001 TABLE 1 i Starting point End point Number 1 1 2 2 2
3 4 2 3 5 5 1 4 6 7 2 5 8 9 2 6 10 10 1 10
TABLE-US-00002 TABLE 2 i Starting point End point Number 1 1 4 4 2
5 7 3 3 8 10 3 10
[0072] In addition, the mean value of the divided sections C.sub.i
may be calculated, as mathematically expressed in Equation (2):
C _ i = 1 n w i - j + 1 j = n w ( i - 1 ) + 1 n w i D j ( 2 )
##EQU00002##
where .left brkt-top. .right brkt-bot. means a ceiling.
[0073] FIG. 6 is a conceptual diagram illustrating a divisional
compression process (S122) in the pre-processing process shown in
FIG. 5.
[0074] As shown in FIG. 6, when the predetermined number (W) of
sections included in the verification parameters is 5, a time axis
of the time series sensing data is divided into 5 sections. In
addition, mean values (11, 12, 13, 14, 15) of the time series
sensing data 10 for each section is calculated. After the
divisional compression (S122), only the mean values 11, 12, 13, 14
and 15, rather than all of the 10 time series sensing data pieces,
are stored, thereby achieving a data compressing effect. In
addition, as only the mean values are stored for each section, some
of noise values are averaged, thereby achieving a noise removing
effect.
[0075] FIG. 7 is a conceptual diagram illustrating a normalization
process (S123) in the pre-processing process shown in FIG. 5.
[0076] The time series sensing data resulting after performing the
divisional compression process shown in FIG. 6 may be normalized
using a mean and a variance of the mean values of the respective
sections for effective comparison between Good and Bad groups
(S123). The upper graphical representation of FIG. 7 illustrates
divisionally compressed time series sensing data prior to
normalization and the lower graphical representation of FIG. 7
illustrates time series sensing data after normalization.
[0077] FIGS. 8A and 8B are conceptual diagrams illustrating a
symbolization process (S124) in the pre-processing process shown in
FIG. 5.
[0078] The time series sensing data resulting after the
normalization shown in FIG. 7 may be SAX (Symbolic Aggregate
approXimation) converted into symbols on the basis of critical
values for the time series sensing data. Here, a number of symbols
(.alpha.) may be included in the verification parameters. When the
measured values of the time series sensing data are symbolized into
a symbols, the critical value y.sub.i, as the basis of symbol
allocation can be calculated as mathematically expressed in
Equation (3):
y i = .PHI. - 1 ( i .alpha. ) ( 3 ) ##EQU00003##
where i=1, 2, . . . , .alpha.-1 and .PHI..sup.-1() denotes an
inverse function of standard normal distribution.
[0079] FIG. 8A illustrates an example of symbolizing measured
values of time series sensing data into 4 symbols A, B, C and D
(.alpha.=4). As the result of the symbolizing, the time series
sensing data shown in FIG. 8A may be converted into "ACDBBDCA".
[0080] As shown in FIG. 8B, the symbolizing may include converting
letters, e.g., alphabetical letters, into digit numbers. FIG. 8B
illustrates an example of symbolizing measured values of time
series sensing data into 5 symbols 1, 2, 3, 4 and 5 (.alpha.=5).
That is to say, when a normalized measured value is y1 or less, a
section corresponding to the normalized measured value is
symbolized into "1". When a normalized measured value is in a range
of y1 to y2, a section corresponding to the normalized measured
value is symbolized into "2". When a normalized measured value is
in a range of y2 to y3, a section corresponding to the normalized
measured value is symbolized into "3". When a normalized measured
value is in a range of y3 to y4, a section corresponding to the
normalized measured value is symbolized into "4". When a normalized
measured value is y4 or greater, a section corresponding to the
normalized measured value is symbolized into "5".
[0081] FIG. 8B illustrates that the time series sensing data
generated at the time of producing a wafer 0310 is symbolized into
"34245", the time series sensing data generated at the time of
producing a wafer 0314 is symbolized into "24245", and the time
series sensing data generated at the time of producing a wafer 0320
is symbolized into "43245".
[0082] The pre-processing process of the time series sensing data
generated at the time of producing each product has been hitherto
described. As shown in FIG. 5, the pre-processing process is
performed through the divisional compression (S122), the
normalization (S124) and the symbolization (S126) in sequence.
However, according to an embodiment of the present invention, only
the normalization (S124) and the symbolization (S126) may be
sequentially performed without performing the divisional
compression (S122).
[0083] Hereinafter, a verification result by the bad pattern
verifying method shown in FIG. 1 will be described with reference
to FIG. 9.
[0084] FIG. 9 illustrates that a bad pattern is symbolized into "65
67 68 65 65". As shown in FIG. 9, when a number of products (Glass
IDs y1 to y5) satisfying conditions of suspicious equipment,
suspicious sensor and verification period defined as verification
parameters is 5 in total, the time series sensing data generated at
the time of producing each product is pre-processed and a symbol
array (SAXed sensor data) is then calculated.
[0085] According to an embodiment of the present invention, 3 kinds
of similarity measures (a first similarity measure, a second
similarity measure, and a third similarity measure) between the
symbol array of the bad pattern and the symbol array of the time
series sensing data can be calculated. The respective similarity
measures are the same with one another in that all of the
similarity measures have values in a range between 0 and 1, but are
different from one another in that similarity measures between two
symbol arrays are evaluated in terms of different parameters. In
order to calculate the similarity measures, it is assumed that a
predetermined digit number is matched to each symbol or each symbol
is comprised of digit numbers.
[0086] The first similarity measure is a similarity measure of an
Euclidean ratio between the symbol array of the bad pattern and the
symbol array of the time series sensing data of the measured
values. The first similarity measure can be calculated, as
expressed in Equation (4):
X=(x.sub.1,x.sub.2, . . . ,x.sub.n)
Y.sup.1=(y.sub.1.sup.1,y.sub.2.sup.1, . . . y.sub.n.sup.1)
Y.sup.2=(y.sub.1.sup.2,y.sub.2.sup.2, . . . y.sub.n.sup.2)
Y.sup.m=(y.sub.1.sup.m,y.sub.2.sup.m, . . . y.sub.n.sup.m)
[0087] where X is a symbol array of the bad pattern, Y.sup.k is a
symbol array of the time series sensing data of the measured values
for a product k,
[0088] Euclidean ratio between X and
Y k == 1 - d k max k ( d k ) , where d k = i = 1 n ( x i - y i k )
2 . ##EQU00004##
[0089] For example, when X=(1, 3, 5, 5, 1), Y.sup.1=(4, 3, 6, 6,
0), and Y.sup.2=(4, 2, 4, 4, 7), d.sup.1=3.4641 and d.sup.2=6.9282.
Thus, the Euclidean ratio of X to Y.sup.1 will be 0.5 and the
Euclidean ratio of X to Y.sup.2 will be 0.
[0090] The second similarity measure may be a similarity measure of
a correlation (r) between the symbol array of the bad pattern and
the symbol array of the time series sensing data. The correlation
between two variables is widely known in the art and a detailed
description thereof will not be given.
[0091] The third similarity measure may be a trend similarity
between the symbol array of the bad pattern and the symbol array of
the time series sensing data. The trend similarity is a similarity
measure between symbol increasing/decreasing index arrays
indicating whether a symbol value of the immediately previous
section on a time axis increases and decreases.
[0092] The trend index similarity between the symbol array
X=(x.sub.1, x.sub.2, . . . , x.sub.n) of the bad pattern and the
symbol array Y=(y.sub.1, y.sub.2, . . . , y.sub.n) of the time
series sensing data can be calculated, as expressed in Equation
(5):
[0093] 1. Calculate:
.DELTA.x.sub.i=x.sub.i+1-x.sub.i
.DELTA.y.sub.i=y.sub.i+1-y.sub.i
[0094] 2. Calculate .DELTA.xy.sub.i by obtaining signs of
(.DELTA.x.sub.i, .DELTA.y.sub.i).fwdarw.
[0095] If sign (.DELTA.x.sub.i)=sign (.DELTA.y.sub.i),
.DELTA.xy.sub.i=1, Otherwise .DELTA.xy.sub.i=0.
3. Trend index similarity = .SIGMA..DELTA. xy n - 1 .
##EQU00005##
[0096] For example, when the symbol array of the bad pattern, X=(1,
3, 5, 5, 1), and the symbol array of the time series sensing data,
Y=(4, 3, 6, 6, 0), .DELTA.X.sub.i=(2, 2, 0, -4), .DELTA.Y.sub.i=(2,
2, 0, -4), .DELTA.XY.sub.i=(0, 1, 1, 1), so that the trend index
similarity becomes 0.75.
[0097] According to the present embodiment of the present
invention, when one of the first to third similarity measures of
the measured time series sensing data for a particular product is
1, the measured time series sensing data is determined to be
similar to the bad pattern and it is checked whether the particular
product has been determined as a good/bad product. If the measured
time series sensing data for the particular product has been
determined to be similar to the bad pattern and the particular
product has been determined as a good product, the more cases, the
higher bad pattern error rate.
[0098] According to the present embodiment of the present
invention, the bad pattern error rate may be defined as a
proportion of products determined as good products among products
having at least one of the first and third similarity measures. In
such a case as shown in FIG. 9, for example, since the symbol array
of measured time series sensing data for products y2 and y1 is
determined as to be similar to the symbol array of the bad pattern
and there is no product determined as a good product, the bad
pattern error rate becomes 0.
[0099] Hereinafter, the configuration and operation of a bad
pattern verification system according to another embodiment of the
present invention will be described with reference to FIG. 10.
[0100] As shown in FIG. 10, the bad pattern verification system 200
according to the present embodiment may include a suspicious bad
pattern selection device 230 and a bad pattern verification device
240.
[0101] After measured values are generated by sensors provided in
equipment 210 when the products pass through the equipment 210, the
suspicious bad pattern selection device 230 selects suspicious
equipment, a suspicious sensor and bad pattern expected to cause
particular faults by analyzing time series sensing data stored in a
sensor-measured value storage device 220 in linkage with product
inspection data stored in a product inspection data storage device
270.
[0102] A user terminal 260 generates verification parameters for
producing verification processes for the suspicious equipment, the
suspicious sensor and the bad pattern, which are selected by the
suspicious bad pattern selection device 230 according to
manipulation by a manager, and provides the generated verification
parameters to the bad pattern verification device 240.
[0103] The bad pattern verification device 240 receives information
on bad pattern applied to time series sensing data of the
suspicious sensor, accesses the time series sensing data of each
product, generated by the suspicious sensor and stored in the
sensor-measured value storage device 220 during a verification
period, calculates similarity measures between the bad pattern
based on the bad pattern information and the time series sensing
data, for each product, and calculates a bad pattern error rate
based on the similarity measures. The bad pattern verification
device 240 may provide the calculated bad pattern error rate to the
user terminal 260.
[0104] Hereinafter, the configuration and operation of a bad
pattern verification system according to still another embodiment
of the present invention will be described with reference to FIG.
11. As shown in FIG. 11, the bad pattern verification device 240
according to the present embodiment may include a verification
parameter receiving unit 241, a time series sensing data extracting
unit 242, a pre-processing unit 243, a similarity measure
calculating unit 244, a determining information receiving unit 245
and a bad pattern verification unit 246.
[0105] The verification parameter receiving unit 241 receives the
verification parameters including bad pattern information applied
to the suspicious sensor measured time series sensing data and
verification method information. The verification parameters may be
received from the user terminal 260. The verification method
information may include verification period information, a number
(W) of sections used in divisional compression, and a number (a) of
symbols used in symbolization.
[0106] The time series sensing data extracting unit 242 accesses
the time series sensing data of each product, generated by the
suspicious sensor during the verification period based on the
verification method information.
[0107] The pre-processing unit 243 may perform a pre-processing
process on the time series sensing data of each product, extracted
by the time series sensing data extracting unit 242. The
pre-processing unit 243 may perform the pre-processing process
having been described above with reference to FIGS. 5 to 8B. For
example, the pre-processing unit 243 may sequentially perform the
divisional compression (S122), the normalization (S124) using a
mean and a variance, and the SAX based symbolization (S126).
[0108] The similarity measure calculating unit 244 calculates a
first similarity measure, a second similarity measure and a third
similarity measure between the symbol array of the bad pattern and
the symbol array of the time series sensing data, for each product.
Here, each of the first to third similarity measures has a value in
a range between 0, which means non-similarity, and 1, which means
the sameness.
[0109] The first similarity measure is an Euclidean ratio between
the bad pattern based on the bad pattern information and the symbol
array of the time series sensing data, the second similarity
measure is a correlation between the symbol array of the bad
pattern and the symbol array of the time series sensing data, and
the third similarity measure is a trend similarity between the
symbol array of the bad pattern and the symbol array of the time
series sensing data. The trend similarity is a similarity measure
between symbol increasing/decreasing index arrays indicating
whether a symbol value of the immediately previous section on a
time axis increases or decreases.
[0110] The bad pattern verification unit 246 calculates the bad
pattern error rate based on the similarity measures calculated by
the similarity measure calculating unit 244. The bad pattern
verification unit 246 receives fault determining information from
the determining information receiving unit 245 querying the fault
determining information of each product, calculates a number of
products determined as good products among products having a value
1 as a value of at least one of the first and second similarity
measures, and calculates the error rate using the number of the
products determined as good products.
[0111] FIG. 12 is another schematic diagram of the bad pattern
verification system different from the bad pattern verification
system shown in FIG. 11. A bad pattern verification device 240 may
have the same configuration as shown in FIG. 12. The bad pattern
verification device 240 may include a processor 43 for executing
commands, a storage 41 in which bad pattern verification program
data is stored, a memory 44, a network interface 42
transmitting/receiving data with respect to an external device, and
a data bus 40 connected to the storage 41, the network interface
42, the processor 43 and the memory 44 to serve as a data movement
path.
[0112] Bad pattern verification program data may be stored in the
storage 41. The bad pattern verification program may include a
module receiving information on bad pattern applied to time series
sensing data measured by a suspicious sensor, a module accessing
the time series sensing data of each product, generated by the
suspicious sensor during a verification period, a module
calculating the similarity measures between the bad pattern based
on the bad pattern information and the time series sensing data,
for each product, and a module calculating the bad pattern error
rate based on the similarity measures.
[0113] FIGS. 13A and 13B illustrate results of monitoring sensing
data in production equipment according to still another embodiment
of the present invention and in conventional production equipment.
In a monitoring process by the conventional production equipment,
an upper limit and a lower limit are just specified and managed for
a value measured by a sensor provided in production equipment.
However, even if a bad pattern is generated, generation of the bad
pattern may be undetectable when the bad pattern is positioned
between the upper limit and the lower limit, and generation of a
fault due to the generation of the bad pattern may not be avoided.
A problem arising in the conventional equipment is illustrated in
FIG. 13A.
[0114] Meanwhile, in the monitoring process in the production
equipment according to still another embodiment of the present
invention, a bad pattern of the sensing data that is input in
advance can be detected, and a warning signal may be generated when
the bad pattern is detected. An effect exhibited in the present
embodiment is shown in FIG. 13B.
[0115] FIG. 14 is a flowchart of a monitoring process of sensing
data generated from a sensor of production equipment according to
still another embodiment of the present invention. The monitoring
process of the sensing data illustrated in FIG. 14 may be performed
by the production equipment.
[0116] First, bad pattern information applied to time series
sensing data measured by a sensor targeted for monitoring is
received (S200). Here, in addition to the bad pattern information,
information on a pre-processing process of the time series sensing
data, for example, a number (W) of sections used in divisional
compression, and a number (a) of symbols used in symbolization, may
further be received.
[0117] Next, the time series sensing data is received from the
monitoring target sensor produced according to the operation of the
production equipment (S210).
[0118] Next, a pre-processing process is performed on the time
series sensing data (S220). In the pre-processing process, the same
pre-processing process as described above with reference to FIGS. 5
to 8B may be used.
[0119] Next, a similarity measure of an Euclidean ratio, a
similarity measure of a correlation, and a trend similarity between
the bad pattern based on the bad pattern information and the symbol
array of the time series sensing data are calculated (S230). Here,
each of the Euclidean ratio, the correlation and the trend
similarity has a value in a range between 0, which means
non-similarity, and 1, which means the sameness.
[0120] Next, when the bad pattern based on the bad pattern
information and the time series sensing data pattern are determined
to be similar to each other, a warning signal is output (S240).
According to an embodiment of the present invention, when the bad
pattern based on the bad pattern information and the time series
sensing data pattern are determined to be similar to each other,
further to the outputting of the warning signal, the production
process may be automatically interrupted.
[0121] According to an embodiment of the present invention, when
the calculated value of at least one of the Euclidean ratio, the
correlation and the trend similarity is 1, the bad pattern based on
the bad pattern information and the time series data pattern may be
determined to be similar to each other.
[0122] The trend similarity is a similarity measure of a pattern
increasing/decreasing according to the passage of time. As to a
method of calculating the trend similarity measure, the Equation
(5) can be referred to.
[0123] FIG. 15 is a block diagram of production equipment according
to still another embodiment of the present invention.
[0124] As shown in FIG. 15, the production equipment 210 according
to an embodiment of the present invention includes one or more
sensors 211 and modules calculating similarity measures with
respect to bad pattern by analyzing FDC time series sensing data
supplied from the sensors 211 and outputting a warning signal based
on the calculated similarity measures.
[0125] First, the time series sensing data stream receiving unit
212 receives a time series sensing data stream measured from the
one or more sensors 211. The time series sensing data stream
receiving unit 212 transfers the time series sensing data stream to
the pre-processing unit 213. The pre-processing unit 213 performs a
pre-processing process on the time series sensing data stream. The
pre-processing unit 213 may perform the same pre-processing process
described above with reference to FIGS. 5 to 8B. For example, the
pre-processing unit 213 may sequentially perform processes of
divisional compression (S122), normalization (S124) and SAX based
symbolization (S126).
[0126] The bad pattern information receiving unit 214 receives
detection parameters including information on bad pattern and
information on a detection method. The detection parameters may be
received from the user terminal 260. The detection method
information may include verification period information, a number
(W) of sections used in divisional compression, and a number
(.alpha.) of symbols used in symbolization. The time series sensing
data based on the bad pattern information may have a pattern
including measured values in a range between an upper limit and a
lower limit of a predefined measured value. That is to say, the
production equipment according to the present embodiment may detect
bad pattern comprised of measured values not exceeding the range
between the upper limit and the lower limit of the predefined
measured value.
[0127] The similarity measure calculating unit 244 calculates a
first similarity measure, a second similarity measure and a third
similarity measure between the symbol array of the bad pattern and
the symbol array of the measured time series sensing data, for each
product. Here, each of the first to third similarity measures has a
value in a range between 0, which means non-similarity, and 1,
which means the sameness.
[0128] When at least one of the Euclidean ratio, the correlation
and the trend similarity, calculated by the similarity measure
calculating unit 215, has a value of 1, the warning signal output
unit 216 generates a warning signal.
[0129] The production equipment according to the present embodiment
may further include a stoppage controlling unit 217 receiving the
warning signal from the warning signal output unit 216 and
generating a control signal for controlling a production operation
to be stopped.
[0130] FIG. 16 is a schematic diagram of a reference pattern
detection system according to still another embodiment of the
present invention.
[0131] The reference pattern detection system according to the
present embodiment may receive time series sensing data from the
one or more sensors and may detect generation of a reference
pattern. Unlike the "bad pattern", the "reference pattern" is used
to indicate a temporary or usual fault generated in the production
equipment or a fault generated under a particular condition, but
not limited thereto. Rather, the "reference pattern" may be broadly
used to indicate time series sensing data patterns to be detected
from the time series sensing data received from the sensors.
Therefore, the "reference pattern" includes "bad pattern".
[0132] As shown in FIG. 16, the reference pattern detection system
according to the present embodiment may include a reference pattern
detecting apparatus 280 receiving time series sensing data from one
or more production equipment 210, detecting generation of bad
pattern and, when the generation of bad pattern is detected, and
outputting a warning signal.
[0133] The reference pattern detecting apparatus 280 may receive
the bad pattern information and information on the equipment 210
and sensors, to which bad pattern are to be applied, from the
reference pattern providing apparatus 290 connected through the
network 250.
[0134] The warning signal generated by the bad pattern detecting
apparatus 280 may be transmitted to a mobile terminal 261 of a
production manager, for example, on a real time basis.
[0135] FIG. 17 is a schematic diagram of a reference pattern
detection system according to still another embodiment of the
present invention.
[0136] As shown in FIG. 17, the reference pattern detection system
according to the present embodiment may include a reference pattern
detecting apparatus 280 receiving time series sensing data from at
least one of an environment sensor and an energy consumption
sensor, which are installed in a building 211, detecting generation
of a reference pattern, and outputting a warning signal when the
generation of the reference pattern is detected. The reference
pattern detection system according to the present embodiment may be
connected to a building energy management system (BEMS) or may be
used as a component of the BEMS.
[0137] The reference pattern detecting apparatus 280 may receive
reference pattern information for the environment sensor and
reference pattern information for the energy consumption sensor,
suggesting that a predetermined abnormal situation has occurred,
from the reference pattern providing apparatus 290 connected
through the network 250.
[0138] According to an embodiment of the present invention, the
predetermined "abnormal situation" may occur when a reference
pattern for the environment sensor and a reference pattern for the
energy consumption sensor are both detected. The reference pattern
detecting apparatus 280 may detect generation of the reference
pattern for the environment sensor by monitoring sensing data
supplied from the environment sensor and may detect generation of
the reference pattern for the energy consumption sensor by
monitoring sensing data supplied from the energy consumption
sensor, thereby determining occurrence of the "abnormal
situation".
[0139] FIG. 18 is a first block diagram of a reference pattern
detection system according to still another embodiment of the
present invention.
[0140] As shown in FIG. 18, the reference pattern detecting
apparatus 280 according to the embodiment of the present invention
may include a time series sensing data stream receiving unit 212, a
pre-processing unit 213, a similarity measure calculating unit 215,
and a reference pattern information receiving unit 214.
[0141] First, the time series sensing data stream receiving unit
212 receives a time series sensing data stream measured from one or
more sensors connected through the network. The time series sensing
data stream receiving unit 212 transfers the measured time series
sensing data stream to the pre-processing unit 213. The
pre-processing unit 213 performs a pre-processing process on the
time series sensing data stream. The pre-processing unit 213 may
perform the same pre-processing process described above with
reference to FIGS. 5 to 8B. For example, the pre-processing unit
213 may sequentially perform processes of divisional compression
(S122), normalization using a variance and a mean value (S124) and
SAX based symbolization (S126).
[0142] The reference pattern information receiving unit 214
receives detection parameters including information on bad pattern
and information on a detection method. The detection parameters may
be received from the user terminal 260. The detection method
information may include verification period information, a number
(W) of sections used in divisional compression, and a number (a) of
symbols used in symbolization.
[0143] The reference pattern information may include the "bad
pattern". The bad pattern may be patterns including measured values
in a range between an upper limit and a lower limit of a predefined
measured value. That is to say, the reference pattern detecting
apparatus 280 according to the present embodiment may detect bad
pattern comprised of measured values not exceeding the range
between the upper limit and the lower limit of the predefined
measured value.
[0144] The similarity measure calculating unit 244 calculates and
outputs a first similarity measure, a second similarity measure and
a third similarity measure between the symbol array of the
reference pattern and the symbol array of the measured time series
sensing data, for each product. Here, each of the first to third
similarity measures has a value in a range between 0, which means
non-similarity, and 1, which means the sameness.
[0145] The reference pattern detecting apparatus 280 according to
the present embodiment may be connected to a building energy
management system (BEMS) or may be used as a component of the BEMS.
Here, the one or more sensors may include a first sensor that
measures energy consumption of a building. Here, the reference
pattern may be a pattern applied to the first sensor to indicate
occurrence of a particular situation.
[0146] FIG. 19 is a second block diagram schematic diagram of a
reference pattern detection system according to still another
embodiment of the present invention.
[0147] As shown in FIG. 19, the reference pattern detecting
apparatus according to still another embodiment of the present
invention may further include a warning signal output unit 216.
[0148] When at least one of the Euclidean ratio, the correlation
and the trend similarity, calculated by the similarity measure
calculating unit 215, has a value of 1, the warning signal output
unit 216 generates a warning signal.
[0149] The warning signal may be supplied to production equipment
from which the bad pattern are sensed through a network.
[0150] According to an embodiment of the present invention, in a
predetermined "abnormal situation" occurring when a reference
pattern for an environment sensor and a reference pattern for an
energy consumption sensor are both detected, the reference pattern
detecting apparatus may generate and output the warning signal.
Here, the one or more sensors include a first sensor that measures
energy consumption of a building and a second sensor that measures
an internal environment parameter of the building. The reference
pattern information receiving unit 214 receives first reference
pattern information applied to the first sensor and second
reference pattern information applied to the second sensor. The
similarity measure calculating unit 215 calculates and outputs a
similarity measure of an Euclidean ratio, a similarity measure of a
correlation and a trend similarity measure between the first
reference pattern based on the first reference pattern information
and sensing data of the first sensor, and calculates and outputs a
similarity measure of an Euclidean ratio, a similarity measure of a
correlation and a trend similarity measure between the second
reference pattern based on the second reference pattern information
and sensing data of the second sensor.
[0151] When the calculated value of at least one of the Euclidean
ratio, the correlation and the trend similarity of the first sensor
is 1 and when the calculated value of at least one of the Euclidean
ratio, the correlation and the trend similarity of the second
sensor is 1, the warning signal output unit 216 generates and
outputs the warning signal.
[0152] While the present invention has been particularly shown and
described with reference to exemplary embodiments thereof, it will
be understood by those of ordinary skill in the art that various
changes in form and details may be made therein without departing
from the spirit and scope of the present invention as defined by
the following claims. It is therefore desired that the present
embodiments be considered in all respects as illustrative and not
restrictive, reference being made to the appended claims rather
than the foregoing description to indicate the scope of the
invention.
[0153] This invention, explained by referring FIGS. 1-15, may be
implemented by using a computer readable code on a non-transitory
machine-readable medium. For example, a computer program product
comprising a non-transitory machine-readable medium storing
instructions that, when executed by at least one programmable
processor, cause the at least one programmable processor to perform
operations comprising may be provided for the implementation of
this invention.
[0154] The foregoing is illustrative of the present invention and
is not to be construed as limiting thereof. Although a few
embodiments of the present invention have been described, those
skilled in the art will readily appreciate that many modifications
are possible in the embodiments without materially departing from
the novel teachings and advantages of the present invention.
Accordingly, all such modifications are intended to be included
within the scope of the present invention as defined in the claims.
Therefore, it is to be understood that the foregoing is
illustrative of the present invention and is not to be construed as
limited to the specific embodiments disclosed, and that
modifications to the disclosed embodiments, as well as other
embodiments, are intended to be included within the scope of the
appended claims. The present invention is defined by the following
claims, with equivalents of the claims to be included therein.
* * * * *