CN104091078B - Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories - Google Patents
Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories Download PDFInfo
- Publication number
- CN104091078B CN104091078B CN201410333952.5A CN201410333952A CN104091078B CN 104091078 B CN104091078 B CN 104091078B CN 201410333952 A CN201410333952 A CN 201410333952A CN 104091078 B CN104091078 B CN 104091078B
- Authority
- CN
- China
- Prior art keywords
- msub
- product
- mtd
- mrow
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000004519 manufacturing process Methods 0.000 claims abstract description 24
- 230000004927 fusion Effects 0.000 claims abstract description 23
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000011524 similarity measure Methods 0.000 claims abstract description 6
- 230000000875 corresponding effect Effects 0.000 claims description 18
- 238000005259 measurement Methods 0.000 claims description 14
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000005516 engineering process Methods 0.000 claims description 6
- 239000011247 coating layer Substances 0.000 claims description 5
- 230000002596 correlated effect Effects 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 3
- 239000000155 melt Substances 0.000 claims description 2
- 238000012360 testing method Methods 0.000 claims description 2
- 238000011084 recovery Methods 0.000 abstract description 4
- 230000015572 biosynthetic process Effects 0.000 description 2
- 239000010410 layer Substances 0.000 description 2
- 238000003786 synthesis reaction Methods 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004381 surface treatment Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- General Factory Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories, the technical problem for causing product individual to follow the trail of open circuit in process of production for solving existing method.Technical scheme is associating between analysis product physics and multi-source information in manufacturing process and data structure, it is scattered with imperfect multi-source information model to establish product in Discrete Manufacturing Process, the characteristic of failure indication part and the historical data base similarity measure of product are calculated by the Euclidean distance based on class center and improved coefficient of variation weighting method, fusion identification is finally carried out to its characteristic layer using D S evidence theories, the effective mark for recovering to remedy failure product, and then the technical problem for solving the failure of DM codes and causing product individual to follow the trail of open circuit in process of production.It is demonstrated experimentally that the inventive method prevents that the recognition correct rate of tracking failure more than 95% remedy reliably, has good practicality by efficiency high, recovery.
Description
Technical field
The present invention relates to a kind of product identification failure means to save the situation, more particularly to a kind of production based on D-S evidence theory
Product Multi-information acquisition indicating failure means to save the situation.
Background technology
Document " a kind of image synthesis preprocess method Xi'an University of Technologys journal for the identification of workpiece 3D bar codes,
2002,18(4):332-336. " discloses a kind of image synthesis preprocess method for recovering identification for workpiece 3D bar codes, laser
The bar code of workpiece surface is etched in, the industrial camera intake 3D bar code images equipped with secondary light source, through histogram equalization, figure
It is final by national standard decoding after the processing such as projective transformation, the extraction of adaptive threshold law spacing wave, the correction of the bar sky information true and false
Workpiece numbering is obtained, the identification recovery of product and tracking under the concrete condition of some processes is partly met and requires.Described in document
Method mainly to know method for distinguishing again after the image preprocessing by bar code, nonetheless, product lifecycle production process
In the technique such as abrasion, pollution, heat treatment and surface treatment DM codes and cleartext information is distinguished and is failed, and
The picture breakdown of product identification than it is more serious when can not can not extract feature from image zooming-out reliable characteristic or completely, so,
Existing technical measures can not prevent product identification completely and not failed in its all production link, cause product individual in life
Open circuit is followed the trail of during production.
The content of the invention
In order to overcome the shortcomings of that existing method causes product individual to follow the trail of open circuit in process of production, the present invention provides a kind of
Product Multi-information acquisition indicating failure means to save the situation based on D-S evidence theory.This method analysis product physics and manufacturing process
Association and data structure between middle multi-source information, it is scattered with imperfect multi-source information mould to establish product in Discrete Manufacturing Process
Type, pass through the characteristic of the Euclidean distance based on class center and improved coefficient of variation weighting method to failure indication part and production
The historical data base similarity measure of product calculates, and finally carries out fusion identification to its characteristic layer using D-S evidence theory, effectively
The mark of failure product is remedied in recovery, and then prevents DM codes from failing and causing product individual to follow the trail of open circuit in process of production.
The technical solution adopted for the present invention to solve the technical problems is:A kind of product based on D-S evidence theory is believed more
Breath fusion indicating failure means to save the situation, is characterized in using following steps:
Step 1: the information source that analysis is related to product identification, including product physical message and manufacturing process information, and will
Multi-information acquisition is incorporated into during product represents that failure remedies.If a multi-source information model, wherein object set are
P={ x1,x2,x3,.....xi} (1)
It is the main body of the multi-source information, element xi, i=1,2,3 ... and ..n } represent different information sources;Property set is
C={ c1,c2,c3,.....cj} (2)
It is the description of product external information attribute;The codomain of attribute is discrete values, is designated as Um(m≤j), wherein m are represented
The attribute of product different aforementioned sources, j represent all properties of product different aforementioned sources, and object set P and property set C set of relations are
F={ fi:m≤j} (3)
Wherein, fm:P→Um(m≤j), fiRepresent product multi-source information object set, fmRepresent the category of actual different aforementioned sources
Property collection, expresses contacting between object set U and property set C, is the basis of information source.E is set again to be determined by property set C
Status attribute collection, i.e.,
E={ e1,e2,e3,.....ej} (4)
Formula (4) represents the state that each attribute of product object is included, its codomain Um' (m≤j) is qualitative value regulation
State is divided into indicating failure and not failed, and property set C is referred to as conditional attribute collection.Product multi-source information model is represented with following formula:
I={ P, C, F, E } (5)
For multiple information sources outside arbitrary product, above-mentioned model is expressed as a two-dimensional data table.
Step 2: when mark fails, scattered and incomplete product physics and manufacture information are gathered, to meet to produce
The requirement of product fusion identification, similitude between judgement sample frequently with neighbour's criterion, using in failure indication and Sample Storehouse not
The degree of generic center matching is high, repeatedly the correlated characteristic of measurement failure product and progress and historical record data
Similarity measure calculates.
1. the product multiple features model based on record.
If (X(1),X(2),....,X(n)) be product multi-source information n dimensions it is overall, therefrom obtain sample data
(x11,x12,...,x1n)T, (x21,x22,...,x2n)T... (xN1,xN2,...,xNn)T
I-th of observation data is designated as in process of producing product
Xi=(xi1,xi2,...,xin)TI=1,2 ..., N (6)
Formula (6) represents multiple features of single measurement product.Introduce data observation matrix
Formula (7) is N × n matrix, and N rows are N number of sample X1,X2,....,XN, form the n dimensions from product multiple features
Overall (X1,X2,....,Xn) sample.Observing matrix X n row are n variable X respectively(1),X(2),....,X(n)Tested in n times
Middle taken value.It is designated as
Xj=(x1j,x2j,...,xNj)TJ=1,2 ..., N (8)
2. the euclidean distance method based on class center is estimated.
Provided with M classification:w1,w2,...wM, have N per classiIndividual sample, is expressed as
For any sample X=(x to be identified1,x2,...,xn), calculate distance
WhereinFor the class center of the i-th class.If compare X to all kinds of distances to meet:
Then X ∈ wi (9)
3. improved coefficient of variation weighting method.
The weight of each feature is calculated according to statistics, the historical data of certain feature is relatively stable in production process change, batch
The distance between secondary middle part is bigger closer to the weight shared by its state, if Xij(i=1,2..., n;J=1,2..., m) be
The historical data of ith feature;Wherein Xj(j=1,2..., m) is the historical data of j-th of feature:
Then average is
Variance is:
The corresponding coefficient of variation is:
For product feature Xj(j=1,2..., m) is according to the historical data of evaluation object, this each attribute change amplitude
The smaller corresponding weight of degree is:Weight is multiple features in process of producing product
Weight shared by ith attribute value,
Step 3: the weight of part, length, intensity, case hardness and coating layer thickness feature are designated as:Wherein, di' to survey feature, repeatedly measure the feature and historical data of failure indication
One group of distance is obtained, the characteristic value of product is designated as:{dij| i=1 ..., 5, j=1 ..., n }, wherein, n is characterized history measurement
Number.Its corresponding five brief inference function, is designated as:m1,m2,...,m5.It is m that the task of basic brief inference, which is exactly,1,
m2,...,m5The probable value that corresponding power set element assigns.It is designated as using its distance average as standard scaleFeature is all
Relation dull and independent, that the power of size and supporting evidence is inversely proportional.If data are close to going through corresponding to actual measurement feature
History data then think that supporting evidence is strong, otherwise bigger with gap, then support weaker.Distributed for the burnt first A of i features confidence level
(Beli(A)) it is:
The closest similarity of each feature and historical data is higher, that is, is interpreted as likelihood degree maximum, on the contrary and Europe
Formula distance is bigger, then likelihood support is weaker, whereinIt is each feature as defined in standard technology apart from reasonable change value.It is right
(Pl is distributed in the burnt first A of i features likelihood degreei(A)), basic brief inference is constructed for each part feature:
From evidence section relation, refusal evidence section certainty value is:mi(B)=1-Pli(A);Uncertain evidence section
Certainty value is:mi(Θ)=1-mi(A)-mi(B)。
Fusion Features are carried out using fusion formula:If Bel1And Bel2For same identification framework Θ two belief functions, m1
And m2Respectively its corresponding Basic Probability As-signment, burnt member are respectively A1,...,AkAnd B1,...,Br, then combinatorial formula is m=m1
⊕m2Determined by following formula:
Wherein, K1The size to conflict between evidence is represented,If K1=1 shows m1And m2It is complete
Full contradiction, it is impossible to be combined to Basic Probability As-signment.For more Evidence Combination Methods, it is combined by team.
The beneficial effects of the invention are as follows:Associating between this method analysis product physics and multi-source information in manufacturing process and
Data structure, it is scattered with imperfect multi-source information model to establish product in Discrete Manufacturing Process, by based on the European of class center
Distance and improved coefficient of variation weighting method are surveyed to the characteristic of failure indication part and the historical data base similarity of product
Degree calculates, and finally carries out fusion identification to its characteristic layer using D-S evidence theory, the effective mark for recovering to remedy failure product
Know, and then prevent DM codes from failing and causing product individual to follow the trail of open circuit in process of production.It is demonstrated experimentally that the inventive method prevents
The recognition correct rate of failure is followed the trail of more than 95%, efficiency high, recovery is remedied reliably, has good practicality.
The present invention is elaborated with reference to the accompanying drawings and detailed description.
Brief description of the drawings
The flow chart of product Multi-information acquisition indicating failure means to save the situation of Fig. 1 present invention based on D-S evidence theory.
Embodiment
Reference picture 1.Product Multi-information acquisition indicating failure means to save the situation of the invention based on D-S evidence theory comprises the following steps that:
Step 1: it is scattered with imperfect multi-source information model to establish product.
The analysis information source related to product identification, including product physical message and manufacturing process information, and by multi information
Fusion is incorporated into during product represents that failure remedies, for the ease of the mathematical description to product multi-source information, if a multi-source information
Model, wherein object set are
P={ x1,x2,x3,.....xi} (1)
It is the main body of the multi-source information, element xi, i=1,2,3 ... and ..n } represent different information sources;Property set is
C={ c1,c2,c3,.....cj} (2)
It is the description of product external information attribute;The codomain of attribute is discrete values, and is designated as Um(m≤j), object set
P and property set C set of relations is
F={ fi:m≤j} (3)
Wherein, fm:P→Um(m≤j), contacting between object set U and property set C is expressed, be the basis of information source.Again
If E is the status attribute collection that is determined by property set C, i.e.,
E={ e1,e2,e3,.....ej} (4)
Formula (4) represents the state that each attribute of product object is included, its codomain Um' (m≤j) is qualitative value regulation
State is divided into indicating failure and not failed, and property set C is referred to as conditional attribute collection.Product multi-source information model is represented with following formula:
I={ P, C, F, E } (5)
For multiple information sources outside arbitrary product, above-mentioned model is expressed as a two-dimensional data table.Note does not occur
Dash number is A, B, C before failure, and part is designated as 1,2,3 after indicating failure.The scattered Incomplete information of the product of failure indication
Historical record such as enters, including the characteristic parameter of weight, part length, intensity, case hardness, coating layer thickness is as shown in table 1:
The product of table 1 does not fail history multi-source information
Step 2: product multiple features similarity measure calculates.
When mark fails, by scattered and incomplete product physics and information gathering is manufactured, to meet that product melts
The requirement of identification is closed, the similitude between judgement sample is frequently with neighbour's criterion, using inhomogeneity in failure indication and Sample Storehouse
The degree of other center matching is high, repeatedly measures the correlated characteristic of failure product and carries out similar to historical record data
Property Likelihood Computation.
1. the product multiple features model based on record.
If (X(1),X(2),....,X(n)) be product multi-source information n dimensions it is overall, therefrom obtain sample data
(x11,x12,...,x1n)T, (x21,x22,...,x2n)T... (xN1,xN2,...,xNn)T
I-th of observation data is designated as in process of producing product
Xi=(xi1,xi2,...,xin)TI=1,2 ..., N (6)
Represent multiple features of single measurement product.Introduce data observation matrix
It is N × n matrix, N rows are N number of sample X1,X2,....,XN, form the overall (X of n dimensions from product multiple features1,
X2,....,Xn) sample.Observing matrix X n row are n variable X respectively(1),X(2),....,X(n)Taken in n times test
Value.It is designated as
Xj=(x1j,x2j,...,xNj)TJ=1,2 ..., N (8)
2. the euclidean distance method based on class center is estimated.
Provided with M classification:w1,w2,...wM, have N per classiIndividual sample, is expressed as
For any sample X=(x to be identified1,x2,...,xn), calculate distance
WhereinFor the class center of the i-th class.If compare X to all kinds of distances to meet:
Then X ∈ wi (9)
3. improved coefficient of variation weighting method.
The weight of each feature is calculated according to statistics, the historical data of certain feature is relatively stable in production process change, batch
The distance between secondary middle part is bigger closer to the weight shared by its state, if Xij(i=1,2..., n;J=1,2..., m) be
The historical data of ith feature;Wherein Xj(j=1,2..., m) is the historical data of j-th of feature:
Then average is
Variance is:
The corresponding coefficient of variation is:
For product feature Xj(j=1,2..., m) is according to the historical data of evaluation object, this each attribute change amplitude
The smaller corresponding weight of degree is:Weight is multiple features in process of producing product
Weight shared by ith attribute value,
According to history making Information Statistics analysis and the weight coefficient of improved coefficient of variation weighting methodCalculate, determine weight in multi-source information feature, length, intensity, case hardness, coating layer thickness institute
It is respectively 0.09,0.11,0.21,0.36,0.23 to account for weight.
If di' to survey feature, repeatedly measure failure indication and obtain one group of distance, the feature of product with historical data
Value is designated as:{dij| i=1 ..., 5, j=1 ..., n }, the present embodiment is chosen 3 different failure indication parts and repeatedly measured simultaneously
With the Euclidean distance of the coefficient of variation weighting method of historical part A, B, C difference computed improved, corresponding 5 of failure part 1
10 data variable weight Euclidean distances of pattern measurement, average value and the minimum distance value for asking for its character pair are as follows:
Step 3: based on multiple features fusion DM codes identification theoretical D-S.
The feature of the part of the present embodiment includes five weight, part length, intensity, case hardness, coating layer thickness features,
Five features are designated as:Wherein di' to survey feature, repeatedly measure the feature of failure indication
One group of distance is obtained with historical data, the characteristic value of product is designated as:{dij| i=1 ..., 5, j=1 ..., n }, wherein n is spy
Levy history measurement number.Its corresponding five brief inference function, is designated as:m1,m2,...,m5.The task of basic brief inference is just
It is for m1,m2,...,m5Corresponding power set element is assigned appropriately, in accordance with the probable value of explanation.Using its distance average as mark
Object staff degree is designated asFeature is all to be dull and independent, relation that the power of size and supporting evidence is inversely proportional.If actual measurement is special
Data corresponding to sign are believed that supporting evidence is strong close to historical data, otherwise bigger with gap, then support weaker.It is designated as not losing
Dash number is A, B, C before effect, and part is designated as 1,2,3 after indicating failure.Repeatedly measurement failure indication obtains with historical data
To one group of distance, pass through the power of similarity measure calculating variable weight Euclidean distance supporting evidence.So for the burnt first A's of i features
Confidence level distributes (Beli(A)) it is:
The closest similarity of each feature and historical data is higher, that is, is interpreted as likelihood degree maximum, on the contrary and Europe
Formula distance is bigger, then likelihood support is weaker, whereinIt is each feature as defined in standard technology apart from reasonable change value, than
Such as undulating value of machining tolerance, weight.SetIt is each feature as defined in standard technology apart from reasonable change value, (point
It is not designated as), distributed for the burnt first A of i features likelihood degree
(Pli(A)), for each part example the basic brief inference of each latent structure:
From evidence section relation, refusal evidence section certainty value is:mi(B)=1-Pli(A);Uncertain evidence section
Certainty value is:mi(Θ)=1-mi(A)-mi(B)。
Fusion Features are carried out using fusion formula:If Bel1And Bel2For same identification framework Θ two belief functions, m1
And m2Respectively its corresponding Basic Probability As-signment, burnt member are respectively A1,...,AkAnd B1,...,Br, then combinatorial formula is m=m1
⊕m2Determined by following formula:
Wherein K1The size to conflict between evidence is represented,If K1=1 shows m1And m2Completely
Contradiction, it is impossible to be combined to Basic Probability As-signment.For more Evidence Combination Methods, it is combined using the above method by team.
Refusing evidence section certainty value is:mi(B)=1-Pli(A) it is, thus, part 1 and each feature of part A, B, C
Constructed the following table is the basic reliability of this example:
The part 1 of the failure indication of table 2 distributes with part A features confidence level
The identification between failure indication part 1 and part A is calculated, according to Dempster fusion rules, first merges weight
Feature and length characteristic, if the result after fusion is m(1)(A),m(1)(B), m(1)(C):
K(1)=m1(A)m2(B)+m1(B)m2(A)=0.4404
m(1)(C)=0.095
The part 1 of the failure indication of table 3 and part A Fusion Features results
By the result m after fusion(1)(A),m(1)(B),m(1)And m (C)3(A),m3(B),m3(C) Dempster fusions are carried out,
This by team merges whole features to class, and part 1 and part A obtains last fusion results and is:m(4)(A)=0.931, m(4)(B)=
0.068,m(4)(C)=0.001.Identification between this calculating failure indication part 1 of class and part B, m(4)(A)=0.338, m(4)(B)=0.662, m(4)(C)=0.Identification between this calculating failure indication part 1 of class and part C, m(4)(A)=
0.488, m(4)(B)=0.512, m(4)(C)=0.The support of target is more general than what original single features were drawn after Fusion Features
Rate is all high, and failure indication part 1 and part A matching degree highest order 0.931, failure indication part 1 and part B matching degree is most
A high position 0.338, failure indication part 1 and the matching degree highest order 0.488 of part C, after multiple features fusion, failure indication
Part 1 and part A identification highests, and uncertainty reduces, and can recover to remedy the mark of part, continues fusion correlated characteristic and carries
The reliability of high target identification.
Claims (1)
1. a kind of product Multi-information acquisition indicating failure means to save the situation based on D-S evidence theory, it is characterised in that including following
Step:
Step 1: the information source that analysis is related to product identification, including product physical message and manufacturing process information, and will believe more
Breath fusion is incorporated into product identification failure and remedied;If a product multi-source information model, wherein object set are P={ x1,x2,
x3,.....xi} (1)
It is the main body of the multi-source information, element xi, i=1,2,3 ... and ..n } represent different information sources;Property set is
C={ c1,c2,c3,.....cJ} (2)
It is the description of product external information attribute;The codomain of attribute is discrete values, is designated as Um(m≤J), wherein m represent product
The attribute of different aforementioned sources, J represent all properties of product different aforementioned sources, and object set P and property set C set of relations are
F={ fi:m≤J} (3)
Wherein, fm:P→Um(m≤J), fiRepresent product multi-source information object set, fmThe property set of actual different aforementioned sources is represented,
Contacting between object set P and property set C is expressed, is the basis of information source;The state for setting E again to be determined by property set C
Property set, i.e.,
E={ e1,e2,e3,.....eJ} (4)
Formula (4) represents the state that each attribute of product object is included, its codomain Um' (m≤J) is qualitative value specified states
It is divided into indicating failure and does not fail, property set C is referred to as conditional attribute collection;Product multi-source information model is represented with following formula:
I={ P, C, F, E } (5)
For multiple information sources outside arbitrary product, above-mentioned model is expressed as a two-dimensional data table;
Step 2: when mark fails, scattered and incomplete product physics and manufacture information are gathered, to meet that product melts
Close the requirement of identification, the similitude between judgement sample use neighbour's criterion, will failure indication with it is different classes of in Sample Storehouse
Center is matched, repeatedly the correlated characteristic of measurement failure product and progress and the similarity measure meter of historical record data
Calculate;
1. the product multiple features model based on record;
If (X(1),X(2),....,X(n)) be product multi-source information n dimensions it is overall, therefrom obtain sample data
(x11,x12,...,x1n)T, (x21,x22,...,x2n)T... (xN1,xN2,...,xNn)T
I-th of observation data is designated as in process of producing product
Xi=(xi1,xi2,...,xin)TI=1,2 ..., N (6)
Formula (6) represents multiple features of single measurement product;Introduce data observation matrix
<mrow>
<mi>X</mi>
<mo>=</mo>
<msub>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>11</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mn>12</mn>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mrow>
<mn>1</mn>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>21</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mn>22</mn>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mrow>
<mn>2</mn>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mrow>
<mi>N</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mrow>
<mi>N</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>x</mi>
<mrow>
<mi>N</mi>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mrow>
<mi>N</mi>
<mo>&times;</mo>
<mi>n</mi>
</mrow>
</msub>
<mo>=</mo>
<mo>&lsqb;</mo>
<msub>
<mi>X</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msub>
<mo>,</mo>
<msub>
<mi>X</mi>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</msub>
<mo>...</mo>
<msub>
<mi>X</mi>
<mrow>
<mo>(</mo>
<mi>n</mi>
<mo>)</mo>
</mrow>
</msub>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
Formula (7) is N × n matrix, and N rows are N number of sample X1,X2,....,XN, it is overall to form the n dimensions from product multiple features;
Observing matrix X n row are n variable X respectively(1),X(2),....,X(n)The value taken in n times test;It is designated as Xj=(x1j,
x2j,...,xNj)TJ=1,2 ..., N (8)
2. the euclidean distance method based on class center is estimated;
Provided with M classification:w1,w2,...wM, have N per classiIndividual sample, is expressed as
For any sample X=(x to be identified1,x2,...,xn), calculate distanceIts
InFor the class center of the i-th class;If compare X to all kinds of distances to meet:
J=1,2 ..., M, i ≠ j then X ∈ wi (9)
3. improved coefficient of variation weighting method;
The weight of each feature is calculated according to statistics, the historical data of certain feature is relatively stable in production process change, in batch
The distance between part is bigger closer to the weight shared by its state, if xij, i=1,2..., n;J=1,2..., j-th of m
I-th historical data of feature;Wherein xj, j=1,2..., m is j-th of product feature:
Then average is
Variance is:
The corresponding coefficient of variation is:
For product feature xj, j=1,2..., m, according to the historical data of evaluation object, each smaller institute of attribute change amplitude degree
The weight accounted for is bigger, and corresponding weight is:QjFor j-th of multiple features in process of producing product
Weight shared by feature,
Step 3: the weight w of part, length l, intensity h, case hardness s and coating layer thickness t features are designated as:Wherein, diTo survey feature, the feature for repeatedly measuring failure indication obtains with historical data
To one group of distance, the characteristic value of product is designated as:{dij| i=1 ..., 5, j=1 ..., n }, wherein, n is characterized history measurement number
Mesh;Its corresponding five brief inference function, is designated as:θ1,θ2,...,θ5;It is θ that the task of basic brief inference, which is exactly,1,
θ2,...,θ5The probable value that corresponding power set element assigns;It is designated as using its distance average as standard scaleMeasurement is lost
Criterion knowledge is designated as d with historical data apart from minimum valuemin, feature all to be dull and independent, size and supporting evidence it is strong
The weak relation being inversely proportional;Think that supporting evidence is strong if data corresponding to actual measurement feature are close to historical data, on the contrary and gap
It is bigger, then support weaker;Bel is distributed for the burnt first A of i features confidence leveli(A) it is:
<mrow>
<msub>
<mi>Bel</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msub>
<mi>&theta;</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>min</mi>
</msub>
<msub>
<mover>
<mi>d</mi>
<mo>&OverBar;</mo>
</mover>
<mi>i</mi>
</msub>
</mfrac>
<mo>,</mo>
<mfrac>
<msub>
<mi>d</mi>
<mi>min</mi>
</msub>
<msub>
<mover>
<mi>d</mi>
<mo>&OverBar;</mo>
</mover>
<mi>i</mi>
</msub>
</mfrac>
<mo>&le;</mo>
<mn>1</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>13</mn>
<mo>)</mo>
</mrow>
</mrow>
The closest similarity of each feature and historical data is higher, that is, it is bigger to be interpreted as likelihood degree, otherwise Euclidean distance
Bigger, then likelihood support is weaker, whereinIt is each feature as defined in standard technology apart from reasonable change value;It is special for i
The burnt first A of sign likelihood degree distribution Pli(A), basic brief inference is constructed for each part feature:
From evidence section relation, refusal evidence section certainty value is:θi(B)=1-Pli(A);Uncertain evidence section reliability
It is worth and is:θi(Θ)=1- θi(A)-θi(B);
Fusion Features are carried out using fusion formula:If Bel1And Bel2For same identification framework Θ two belief functions, θ1And θ2
Respectively its corresponding Basic Probability As-signment, burnt member are respectively A1,...,AkAnd B1,...,Br, then combinatorial formula is θ=θ1⊕θ2
Determined by following formula:
Wherein,K1<The size to conflict between 1 expression evidence, if K1=1 shows θ1And θ2Complete lance
Shield, it is impossible to be combined to Basic Probability As-signment;For more Evidence Combination Methods, it is combined by team.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410333952.5A CN104091078B (en) | 2014-07-14 | 2014-07-14 | Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410333952.5A CN104091078B (en) | 2014-07-14 | 2014-07-14 | Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104091078A CN104091078A (en) | 2014-10-08 |
CN104091078B true CN104091078B (en) | 2017-12-12 |
Family
ID=51638794
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410333952.5A Expired - Fee Related CN104091078B (en) | 2014-07-14 | 2014-07-14 | Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104091078B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104537432B (en) * | 2014-12-16 | 2018-01-05 | 华南理工大学 | The decision-making technique of power system Multiobjective Optimal Operation based on evidential reasoning |
CN105515184B (en) * | 2015-12-04 | 2018-07-17 | 国网河南省电力公司电力科学研究院 | Multisensor many reference amounts distribution synergic monitoring system based on wireless sensor network |
CN105913144B (en) * | 2016-04-07 | 2018-07-03 | 北京航空航天大学 | A kind of method for predicting service life of product based on goal orientation Optimum Matching similitude |
CN106056358B (en) * | 2016-05-27 | 2019-05-07 | 南昌航空大学 | A kind of process failure mode automatic identifying method |
CN111639065B (en) * | 2020-04-17 | 2022-10-11 | 太原理工大学 | Polycrystalline silicon ingot casting quality prediction method and system based on batching data |
CN112288308A (en) * | 2020-11-11 | 2021-01-29 | 华能澜沧江水电股份有限公司 | DS evidence theory landslide evaluation method based on multiple coefficient of variation weighting |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202694336U (en) * | 2012-06-01 | 2013-01-23 | 深圳市创荣发电子有限公司 | Device with data transmission circuit |
-
2014
- 2014-07-14 CN CN201410333952.5A patent/CN104091078B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202694336U (en) * | 2012-06-01 | 2013-01-23 | 深圳市创荣发电子有限公司 | Device with data transmission circuit |
Non-Patent Citations (3)
Title |
---|
Data Acquisition and Quality Assurance Integration Platform Based on DPM;Jian Wang et al.;《applied Mechanics and Materials Vols》;20130910;第2815-2821页 * |
基于D-S证据理论的信息融合算法;江涛;《计算机科学》;20131130;第40卷(第11A期);第120-124页 * |
基于D-S证据理论的组合数据融合算法;陆三兰等;《微电子学与计算机》;20110131;第28卷(第1期);第95-98页 * |
Also Published As
Publication number | Publication date |
---|---|
CN104091078A (en) | 2014-10-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104091078B (en) | Product Multi-information acquisition indicating failure means to save the situation based on D S evidence theories | |
CN105095238B (en) | For detecting the decision tree generation method of fraudulent trading | |
CN105975941B (en) | A kind of multi-direction vehicle detection identifying system based on deep learning | |
CN106575367B (en) | Method and system for the face critical point detection based on multitask | |
CN104931960B (en) | Trend message and radar target situation information full flight path segment data correlating method | |
CN110070010A (en) | A kind of face character correlating method identified again based on pedestrian | |
US20150134578A1 (en) | Discriminator, discrimination program, and discrimination method | |
CN102393912B (en) | Comprehensive target identification method based on uncertain reasoning | |
CN106326868A (en) | Face identification method based on cosine similarity measure learning | |
CN107977656A (en) | A kind of pedestrian recognition methods and system again | |
CN106127733B (en) | The method and apparatus of human body target identification | |
CN102722887A (en) | Image registration method and device | |
CN110348437A (en) | It is a kind of based on Weakly supervised study with block the object detection method of perception | |
CN108268632A (en) | Abnormal information data identifies machine learning method | |
CN107392929A (en) | A kind of intelligent target detection and dimension measurement method based on human vision model | |
CN108710900A (en) | A kind of multi-platform sensor measurement data fusion method based on D-S reasonings | |
CN105976397A (en) | Target tracking method based on half nonnegative optimization integration learning | |
CN107247954A (en) | A kind of image outlier detection method based on deep neural network | |
Kansizoglou et al. | Haseparator: Hyperplane-assisted softmax | |
CN107944340B (en) | Pedestrian re-identification method combining direct measurement and indirect measurement | |
CN107644203A (en) | A kind of feature point detecting method of form adaptive classification | |
CN104331711B (en) | SAR image recognition methods based on multiple dimensioned fuzzy mearue and semi-supervised learning | |
Agarwal et al. | One-Shot learning based classification for segregation of plastic waste | |
CN112241844A (en) | Method and device for determining and updating environmental risk source background list of drinking water source area | |
CN104462826B (en) | The detection of multisensor evidences conflict and measure based on Singular Value Decomposition Using |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171212 |