CN102706573A - Fault classification diagnosis method of equipment - Google Patents

Fault classification diagnosis method of equipment Download PDF

Info

Publication number
CN102706573A
CN102706573A CN2012100691763A CN201210069176A CN102706573A CN 102706573 A CN102706573 A CN 102706573A CN 2012100691763 A CN2012100691763 A CN 2012100691763A CN 201210069176 A CN201210069176 A CN 201210069176A CN 102706573 A CN102706573 A CN 102706573A
Authority
CN
China
Prior art keywords
vector machine
square method
supporting vector
method supporting
compound
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.)
Pending
Application number
CN2012100691763A
Other languages
Chinese (zh)
Inventor
陈勇旗
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ningbo University
Original Assignee
Ningbo University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Ningbo University filed Critical Ningbo University
Priority to CN2012100691763A priority Critical patent/CN102706573A/en
Publication of CN102706573A publication Critical patent/CN102706573A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)

Abstract

The invention discloses a fault classification diagnosis method of equipment, which is characterized by comprising the following steps of: constructing a classifier model of a feature weighting least square support vector machine; constructing a classifier model of a compound-type feather weighting least square support vector machine; constructing a model parameter of the classifier model of the compound-type feather weighting least square support vector machine; inputting a fault training sample of the equipment, and training the classifier model of the compound-type feather weighting least square support vector machine; and inputting a fault sample to be measured into the trained classifier model of the compound-type feather weighting least square support vector machine for fault classification diagnosis. The fault classification diagnostic method of equipment has the advantage that the fault classification accuracy and fault diagnosis precision of the equipment are improved.

Description

A kind of failure modes diagnostic method of equipment
Technical field
The present invention relates to a kind of method for diagnosing faults of equipment, more particularly, to a kind of failure modes diagnostic method of equipment.
Background technology
The later stage nineties, Vapnik proposes the machine learning method of SVMs on the basis of Statistical Learning Theory.SVMs can solve the problems, such as higher-dimension and the local extremum problem for being difficult to overcome in neural network algorithm, achieve certain achievement in field of diagnosis about equipment fault at present again while small sample problem is solved.The SVMs that Vapnik is proposed thinks that the importance of the data sample of each in training sample is identical in the training process, but in actual applications, the significance level of each data sample is different, particularly in the case where training sample has noise, discounting for different importance of each data sample to training process, it will be easy to over-fitting occur in the training process.To solve the above problems, Chun-Fu Lin and Sheng-De Wang propose Fuzzy Support Vector Machine, the concept that fuzzy support vector machine introduces fuzzy membership is distinguished to the importance of each data sample, it being assigned according to the importance of each data sample in the training process and being subordinate to angle value accordingly, the size for being subordinate to angle value determines the corresponding importance of the data sample.Fuzzy support vector machine inhibits the influence of noise and trouble point to training process to a certain extent, improves the precision of failure modes and fitting, progressively replaces SVMs in field of diagnosis about equipment fault at present.Fuzzy support vector machine is although it is believed that the importance of each data sample in the training process is different, but it is all equal but to give tacit consent to contribution of each different characteristic of each data sample to the performance of SVMs in itself.But when fuzzy support vector machine is applied to the fault diagnosis field of equipment, the characteristic attribute species of each data sample in training sample is more, some characteristic indexs of data sample occupy main status in failure modes, and some characteristic indexs are probably redundancy index, if the significance level of the various features attribute of each data sample is considered as identical, so in failure diagnostic process, the classification degree of accuracy of failure will be reduced, so that the precision to fault diagnosis has undesirable effect.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of classification degree of accuracy height, the failure modes diagnostic method of the high equipment of diagnostic accuracy.
The present invention solve the technical scheme that is used of above-mentioned technical problem for：A kind of failure modes diagnostic method of equipment, bag
Include following steps：
（1）The sorter model of construction feature Weighted Least Squares Support Vector Machines；
（2）Build the sorter model of compound characteristic weighing least square method supporting vector machine；
（3）Build the model parameter of the sorter model of compound characteristic weighing least square method supporting vector machine；
（4）The failure training sample of equipment is inputted, the classification to compound characteristic weighing least square method supporting vector machine
Device model is trained；
（5）By the compound characteristic weighing least square method supporting vector machine after the fault sample to be measured input training of equipment
Failure modes diagnosis is carried out in sorter model.
Described step（1）In the building process of sorter model of characteristic weighing least square method supporting vector machine be：
（1）Build the sorter model of least square method supporting vector machine：Set first
Figure DEST_PATH_IMAGE001
Individual training sample
Figure 412033DEST_PATH_IMAGE002
,
x iForiIndividual input data,,
Figure 314130DEST_PATH_IMAGE004
Representx iThe input space at place, d representation space dimensions,
Figure DEST_PATH_IMAGE005
Represent input datax iCorresponding classification,,Represent training sample from the input space
Figure 211865DEST_PATH_IMAGE004
To the Nonlinear Mapping of high-dimensional feature space, pass throughBy training sample from the input space
Figure 825566DEST_PATH_IMAGE008
High-dimensional feature space is mapped to, the functional expression stated in the classification problem of high-dimensional feature space least square method supporting vector machine is:
             
Figure DEST_PATH_IMAGE009
             s.t.
Figure 539444DEST_PATH_IMAGE010
Wherein min represents to minimize,wThe weight vector of high-dimensional feature space is represented,bBiasing coefficient is represented,c Represent penalty coefficient,
Figure DEST_PATH_IMAGE011
Represent slack variable,
Figure 647077DEST_PATH_IMAGE012
The wrong degree divided is represented, subscript T represents transformation of ownership computing, and s.t. represents constraints；Then Lagrangian is introduced, the expression functional expression for obtaining the minimum value of Lagrangian is：
Wherein
Figure 158131DEST_PATH_IMAGE014
It is Lagrange multiplier,
Figure DEST_PATH_IMAGE015
It is correspondingFor supporting vector, make L pairsw、b、
Figure 857283DEST_PATH_IMAGE011
、
Figure 186633DEST_PATH_IMAGE014
Partial derivative be equal to zero, while introduce kernel function
Figure DEST_PATH_IMAGE017
（i=1,2,3 ... n；j=1,2,3 ... n）, calculate
Figure 182271DEST_PATH_IMAGE018
With
Figure DEST_PATH_IMAGE019
Concrete outcome, according to
Figure 913466DEST_PATH_IMAGE018
WithResult, the categorised decision function for obtaining least square method supporting vector machine is, that is, construct the sorter model of least square method supporting vector machine；
（2）The sorter model of construction feature Weighted Least Squares Support Vector Machines on the basis of the sorter model of least square method supporting vector machine：First set sample characteristics weighting parameter be,p=1,2,3 ..., m,
Figure 788384DEST_PATH_IMAGE022
,Sequence number is characterized, feature weight matrix is constructed on this basis
Figure 439333DEST_PATH_IMAGE024
, least square method supporting vector machine is improved using feature weight matrix, calculating of the smaller feature of weights on nonlinear mapping function and kernel function is influenceed smaller, the classification problem of the least square method supporting vector machine after improvement can be expressed as functional expression：
Figure DEST_PATH_IMAGE025
        
s.t. 
Figure 665915DEST_PATH_IMAGE026
  
Wherein
Figure DEST_PATH_IMAGE027
,
Figure 969857DEST_PATH_IMAGE028
,
Figure DEST_PATH_IMAGE029
,,
Figure DEST_PATH_IMAGE031
,
Figure 13086DEST_PATH_IMAGE032
, then optimized by method of Lagrange multipliers, while introducing kernel function, can be calculated following matrix equality：
Figure DEST_PATH_IMAGE033
s.t. ,
Figure 201807DEST_PATH_IMAGE034
,
Figure DEST_PATH_IMAGE035
,
Figure 611448DEST_PATH_IMAGE036
Wherein, gaussian kernel function is：
Figure DEST_PATH_IMAGE037
, wherein σ represents the radial width parameter of gaussian kernel function, and the categorised decision function for obtaining characteristic weighing least square method supporting vector machine is： , that is, construct the sorter model of characteristic weighing least square method supporting vector machine.
Described step（2）In the sorter model of compound characteristic weighing least square method supporting vector machine be in feature
On the basis of the sorter model of Weighted Least Squares Support Vector Machines, fitness function is used as using the classification accuracy of characteristic weighing least square method supporting vector machine, selection is iterated to the weight vector of characteristic weighing least square method supporting vector machine by compound algorithm, obtain best initial weights vector, best initial weights vector is brought into the sorter model of characteristic weighing least square method supporting vector machine again, the sorter model of compound characteristic weighing least square method supporting vector machine is obtained.
4. the failure modes diagnostic method of a kind of equipment according to claim 3, it is characterised in that the process that described compound algorithm is iterated selection to the weight vector of characteristic weighing least square method supporting vector machine comprises the following steps：
（1）Generate initial composite type
Figure DEST_PATH_IMAGE039
Individual weights summit
Figure 740126DEST_PATH_IMAGE040
,
Figure DEST_PATH_IMAGE041
, and it is converted into meeting one by one to the feasible individual set of constraints, using these weights summits as initial weight set；
（2）The nicety of grading of the characteristic weighing least square method supporting vector machine on compound each weights summit is calculated, best initial weights point of the nicety of grading as fitness function is determined
Figure 18661DEST_PATH_IMAGE042
, worst weights point
Figure DEST_PATH_IMAGE043
With suboptimum weights point
Figure 963483DEST_PATH_IMAGE044
；
（3）Calculate the center on compound weights summit：
Figure DEST_PATH_IMAGE045
；
（4）Calculate compound pip
Figure 378284DEST_PATH_IMAGE046
,
Figure DEST_PATH_IMAGE047
For reflectance factor,
Figure 117570DEST_PATH_IMAGE048
；
If the fitness value of pip
Figure DEST_PATH_IMAGE049
Better than current optimum point
Figure 617822DEST_PATH_IMAGE042
, then step is gone to（5）；
If the fitness value of pip
Figure 369264DEST_PATH_IMAGE049
Better than current time advantage, then, put
Figure DEST_PATH_IMAGE051
, go to step（7）；
If the fitness value of pip
Figure 739569DEST_PATH_IMAGE049
It is only most not good enough better than current
Figure 26194DEST_PATH_IMAGE043
, compression pointIf, the fitness value of compression point
Figure DEST_PATH_IMAGE053
Better than currently most not good enough
Figure 762254DEST_PATH_IMAGE043
, then
Figure 909202DEST_PATH_IMAGE054
, put
Figure 999518DEST_PATH_IMAGE051
, then step is gone to（7）If, the fitness value of compression point
Figure 181100DEST_PATH_IMAGE053
It is worse than current most not good enough
Figure 698669DEST_PATH_IMAGE043
, then step is gone to（6）；
If the fitness value of pip
Figure 598492DEST_PATH_IMAGE049
It is worse than current most not good enough
Figure DEST_PATH_IMAGE055
, then compression point
Figure 303885DEST_PATH_IMAGE056
If, the fitness value of compression point
Figure 339974DEST_PATH_IMAGE053
Better than currently most not good enough
Figure 28444DEST_PATH_IMAGE043
, then
Figure 149984DEST_PATH_IMAGE054
, put
Figure 582103DEST_PATH_IMAGE051
, then step is gone to（7）If, the fitness value of compression point
Figure 738277DEST_PATH_IMAGE053
It is worse than current most not good enough
Figure 597649DEST_PATH_IMAGE043
, then step is gone to（6）；
（5）Calculate inflexion point
Figure DEST_PATH_IMAGE057
,
Figure 268802DEST_PATH_IMAGE058
For flare factor,
Figure DEST_PATH_IMAGE059
,
Figure 239032DEST_PATH_IMAGE060
, put
Figure 312030DEST_PATH_IMAGE051
, go to step（7）；
（6）Halve operation, keep optimum pointConstant, remaining each summit is all towards optimum point
Figure 441365DEST_PATH_IMAGE042
It is close, be close to formula
Figure DEST_PATH_IMAGE061
, put
Figure 215286DEST_PATH_IMAGE051
, go to step（7）；
（7）If
Figure 408369DEST_PATH_IMAGE062
, wherein 0 ＜
Figure DEST_PATH_IMAGE063
＜ 0.1, show this it is compound have been adjusted to optimum state, iteration terminates, obtain best initial weights
Figure 671861DEST_PATH_IMAGE064
, otherwise go to step（2）.
The model parameter of the sorter model of described compound characteristic weighing least square method supporting vector machine includes weight vector
Optimizing space, initial weight Vector Groups, penalty coefficient, the radial width of gaussian kernel function, reflectance factor, the compressed coefficient, flare factor and iterations.
Compared with prior art, the advantage of the invention is that first on the basis of least square method supporting vector machine construction feature Weighted Least Squares Support Vector Machines, then selection is iterated to the weight vector of characteristic weighing least square method supporting vector machine by compound algorithm, build the sorter model of compound characteristic weighing least square method supporting vector machine, corresponding weight vector can be set to the size of the contribution margin of the performance of SVMs according to each different characteristic of input data, the weak related or incoherent feature of input data is avoided to adversely affect classification results, improve the failure modes degree of accuracy and the fault diagnosis precision of equipment.
Brief description of the drawings
Fig. 1 is workflow diagram of the invention.
Embodiment
The present invention is described in further detail below in conjunction with accompanying drawing embodiment.
Embodiment one：The failure modes diagnostic method of a kind of equipment, it is characterised in that comprise the following steps：
（1）The sorter model of construction feature Weighted Least Squares Support Vector Machines；
（2）Build the sorter model of compound characteristic weighing least square method supporting vector machine；
（3）Build the model parameter of the sorter model of compound characteristic weighing least square method supporting vector machine；
（4）The failure training sample of equipment is inputted, the classification to compound characteristic weighing least square method supporting vector machine
Device model is trained；
（5）By the compound characteristic weighing least square method supporting vector machine after the fault sample to be measured input training of equipment
Failure modes diagnosis is carried out in sorter model.
In above-mentioned specific embodiment, the building process of the sorter model of characteristic weighing least square method supporting vector machine includes： （1）Build the sorter model of least square method supporting vector machine： 
Setting
Figure DEST_PATH_IMAGE065
Individual training sample
Figure 52026DEST_PATH_IMAGE002
, input data
Figure 567321DEST_PATH_IMAGE003
, 
Figure 614912DEST_PATH_IMAGE004
Representx iThe input space at place, d represents the dimension of the input space,
Figure 986987DEST_PATH_IMAGE005
Represent input datax iCorresponding classification,
Figure 857379DEST_PATH_IMAGE006
.Generally, it is directly extremely difficult in input space searching optimal separating hyper plane, it is therefore desirable to introduce nonlinear mapping function
Figure 910785DEST_PATH_IMAGE007
,
Figure 812882DEST_PATH_IMAGE007
Represent training sample from the input spaceTo the Nonlinear Mapping of high-dimensional feature space, pass through
Figure 913879DEST_PATH_IMAGE007
By training sample from the input space
Figure 567715DEST_PATH_IMAGE008
High-dimensional feature space is mapped to, optimal separating hyper plane is constructed in this high-dimensional feature space, its function expression is：  
Figure 527580DEST_PATH_IMAGE066
                                  (1)      
Wherein,wThe weight vector of high-dimensional feature space is represented,bBiasing coefficient is represented, subscript T represents transformation of ownership computing, input dataFollowing two classes situation is divided into by optimal separating hyper plane： 
Figure 241459DEST_PATH_IMAGE068
              (2)
When can not the classification of input data is completely separable with an optimal separating hyper plane（Part input data is by mistake point）, slack variable can be introduced, optimal separating hyper plane is met formula：
Figure 541039DEST_PATH_IMAGE010
                               (3)
Formula（3）For the categorised decision function of least square method supporting vector machine two, when
Figure DEST_PATH_IMAGE069
When, input data
Figure 240190DEST_PATH_IMAGE067
Correctly classified, and worked as
Figure DEST_PATH_IMAGE071
When, input data
Figure 572470DEST_PATH_IMAGE067
By mistake point.Therefore,
Figure 833687DEST_PATH_IMAGE012
Input data is weighed
Figure 502566DEST_PATH_IMAGE067
By the degree of mistake point, meanwhile, optimal separating hyper plane must make input data
Figure 558247DEST_PATH_IMAGE067
Classification space Margin=
Figure 312576DEST_PATH_IMAGE072
Maximum, in view of the foregoing, the Solve problems of optimal separating hyper plane can be changed into the following object function of solution：
        
s.t.
Figure 900869DEST_PATH_IMAGE010
                          (4) 
Formula（4）The classification problem of least square method supporting vector machine is described, wherein min represents the minimum value found a function,c Penalty coefficient is represented, s.t. represents constraints；Above optimization problem can be solved by method of Lagrange multipliers, and specific formula is as follows：
Figure 127451DEST_PATH_IMAGE013
  (5)                                   
Wherein
Figure DEST_PATH_IMAGE073
It is Lagrange multiplier,
Figure 431394DEST_PATH_IMAGE015
It is corresponding
Figure 34413DEST_PATH_IMAGE016
For supporting vector.Above formula is optimized according to optimality condition, makes L pairsw、b、
Figure 474622DEST_PATH_IMAGE011
、
Figure 75368DEST_PATH_IMAGE073
Partial derivative be equal to zero, can obtain the functional expression group of following optimal conditions：
Figure 592238DEST_PATH_IMAGE074
                           (6)
Figure DEST_PATH_IMAGE075
                               （7）
Figure 998948DEST_PATH_IMAGE076
                （8）
Figure DEST_PATH_IMAGE077
                                  （9）                                                                                                    
By functional expression group（6）~（9）It is further converted into matrix form：
Figure 293663DEST_PATH_IMAGE078
                                  (10)
s.t.
Figure 127627DEST_PATH_IMAGE032
,
Figure 140583DEST_PATH_IMAGE034
,
Figure DEST_PATH_IMAGE079
, 
Due to formula
Figure 500206DEST_PATH_IMAGE007
Without any relevant information, therefore above formula can not be calculated, but SVMs possesses a very important characteristic：I.e. without it is understood that, but introduce kernel function
Figure DEST_PATH_IMAGE081
（i=1,2,3 ... n；j=1,2,3 ... n）, we can calculate biasing coefficient
Figure 8252DEST_PATH_IMAGE018
And Lagrange multiplierConcrete outcome, substituted into the categorised decision functional expression of least square method supporting vector machine two（3）In, you can obtain the function expression of the sorter model of least square method supporting vector machine：
(11)
（2）The sorter model of construction feature Weighted Least Squares Support Vector Machines on the basis of the sorter model of least square method supporting vector machine：Set sample characteristics weighting parameter be
Figure 963756DEST_PATH_IMAGE021
,p=1,2,3 ..., m,,
Figure 64753DEST_PATH_IMAGE023
Sequence number is characterized, feature weight matrix is constructed on this basis
Figure 351377DEST_PATH_IMAGE024
, least square method supporting vector machine is improved using feature weight matrix, calculating of the smaller feature of weights on nonlinear mapping function and kernel function is influenceed smaller, the classification problem of the least square method supporting vector machine after improvement can be expressed as functional expression：
Figure 678454DEST_PATH_IMAGE082
        
s.t. 
Figure DEST_PATH_IMAGE083
  
Wherein
Figure 759542DEST_PATH_IMAGE084
,
Figure 234386DEST_PATH_IMAGE028
,
Figure 262385DEST_PATH_IMAGE030
,
Figure 506284DEST_PATH_IMAGE029
,
Figure 23853DEST_PATH_IMAGE031
,
Figure 658097DEST_PATH_IMAGE032
, then optimized by method of Lagrange multipliers, while introducing kernel function, can be calculated following matrix equality：
Figure 289454DEST_PATH_IMAGE033
s.t. 
Figure 591122DEST_PATH_IMAGE032
,
Figure 279593DEST_PATH_IMAGE034
,
Figure 401132DEST_PATH_IMAGE035
,
Figure 833251DEST_PATH_IMAGE036
Selection weighting kernel function, is trained using training sample and weighted, the sorter model of construction feature Weighted Least Squares Support Vector Machines；The use gaussian kernel function in linear kernel function, Polynomial kernel function or gaussian kernel function, the present embodiment can be used by weighting kernel function, obtain the categorised decision function of characteristic weighing least square method supporting vector machine（That is the sorter model of characteristic weighing least square method supporting vector machine）For: 
Figure 989426DEST_PATH_IMAGE038
The expression formula of wherein gaussian kernel function is：
Figure 848797DEST_PATH_IMAGE037
Wherein σ represents the radial width parameter of gaussian kernel function, from the expression formula of gaussian kernel function, if in the presence of a feature weight
Figure DEST_PATH_IMAGE085
,
Figure 519950DEST_PATH_IMAGE086
Smaller, the calculating influence on gaussian kernel function is smaller, and the influence to classification results is smaller, when
Figure DEST_PATH_IMAGE087
When, the attribute is disappeared from the calculating process of kernel function, and classification results are had no effect.
In above-mentioned specific embodiment, the sorter model of compound characteristic weighing least square method supporting vector machine is added in feature
On the basis of the sorter model for weighing least square method supporting vector machine, fitness function is used as using the classification accuracy of characteristic weighing least square method supporting vector machine, selection is iterated to the weight vector of characteristic weighing least square method supporting vector machine by compound algorithm, obtain best initial weights vector, best initial weights vector is brought into the sorter model of characteristic weighing least square method supporting vector machine again, the sorter model of compound characteristic weighing least square method supporting vector machine is obtained.
Compound algorithm（Compound optimal searching algorithm）It is to solve for a kind of with direct local search algorithm widely of nonlinear optimization design problem.It is so-called it is compound be in the Constraint feasible domain of n-dimensional space, it is made up of m summit compound, calculate the fitness function value on each summit, then m fitness function value is compared one by one, remove the worst summit of fitness function value, construction one meets constraints, and fitness function value is preferably put as new summit simultaneously, utilize the new summit, it is new compound so as to constantly construct, draw close the new compound optimum point constantly into feasible zone, i.e. each summit is constantly approached towards the optimal point of fitness function in an iterative process, the approximate solution of convergence criterion is met until getting.Compound algorithm is needed only to launch, expand, compress and rotate etc. and calculated in iterative process, and need not as simplex method calculating target function one, second dervative, one-dimensional optimization direction search procedure is also omit, it is therefore, without any requirement to object function.
Feature weight
Figure 490180DEST_PATH_IMAGE086
Represent the size of each characteristic attribute contribution degree, so as to determine the nicety of grading of least square method supporting vector machine, selection is iterated to the optimal value of feature weight using compound algorithm in compound characteristic weighing least square method supporting vector machine, its iterative selection process includes step：
（1）Generate initial composite type
Figure 563178DEST_PATH_IMAGE039
Individual weights summit
Figure 531134DEST_PATH_IMAGE040
,
Figure 689583DEST_PATH_IMAGE041
, and it is converted into meeting one by one to the feasible individual set of constraints, using these weights summits as initial weight set；
（2）The nicety of grading of the characteristic weighing least square method supporting vector machine on compound each weights summit is calculated, best initial weights point of the nicety of grading as fitness function is determined
Figure 401187DEST_PATH_IMAGE042
, worst weights point
Figure 331622DEST_PATH_IMAGE043
With suboptimum weights point
Figure 470479DEST_PATH_IMAGE044
；
（3）Calculate the center on compound weights summit：
Figure 116224DEST_PATH_IMAGE045
；
（4）Calculate compound pip
Figure 693836DEST_PATH_IMAGE046
,
Figure 475847DEST_PATH_IMAGE047
For reflectance factor,；
If the fitness value of pip
Figure 918647DEST_PATH_IMAGE049
Better than current optimum point
Figure 34370DEST_PATH_IMAGE042
, then step is gone to（5）；
If the fitness value of pip
Figure 874150DEST_PATH_IMAGE049
Better than current time advantage
Figure 417127DEST_PATH_IMAGE044
, then
Figure 975147DEST_PATH_IMAGE050
, put, go to step（7）；
If the fitness value of pip
Figure 323269DEST_PATH_IMAGE049
It is only most not good enough better than current
Figure 317375DEST_PATH_IMAGE043
, compression point
Figure 97112DEST_PATH_IMAGE052
If, the fitness value of compression pointBetter than currently most not good enough
Figure 431327DEST_PATH_IMAGE043
, then, put
Figure 583140DEST_PATH_IMAGE051
, then step is gone to（7）If, the fitness value of compression point
Figure 844357DEST_PATH_IMAGE053
It is worse than current most not good enough
Figure 513236DEST_PATH_IMAGE043
, then step is gone to（6）；
If the fitness value of pip
Figure 568916DEST_PATH_IMAGE049
It is worse than current most not good enough
Figure 323246DEST_PATH_IMAGE055
, then compression point
Figure 122575DEST_PATH_IMAGE056
If, the fitness value of compression point
Figure 645960DEST_PATH_IMAGE053
Better than currently most not good enough
Figure 138121DEST_PATH_IMAGE043
, then
Figure 444993DEST_PATH_IMAGE054
, put
Figure 720117DEST_PATH_IMAGE051
, then step is gone to（7）If, the fitness value of compression point
Figure 425904DEST_PATH_IMAGE053
It is worse than current most not good enough
Figure 761071DEST_PATH_IMAGE043
, then step is gone to（6）；
（5）Calculate inflexion point
Figure 552309DEST_PATH_IMAGE057
,
Figure 631124DEST_PATH_IMAGE058
For flare factor,
Figure 191418DEST_PATH_IMAGE059
,, put
Figure 976020DEST_PATH_IMAGE051
, go to step（7）；
（6）Halve operation, keep optimum point
Figure 920843DEST_PATH_IMAGE042
Constant, remaining each summit is all towards optimum point
Figure 273327DEST_PATH_IMAGE042
It is close, be close to formula
Figure 278192DEST_PATH_IMAGE061
, put
Figure 716126DEST_PATH_IMAGE051
, go to step（7）；
（7）If
Figure 199060DEST_PATH_IMAGE062
, wherein
Figure 671630DEST_PATH_IMAGE063
For a less value, its span can be 0 ＜
Figure 850326DEST_PATH_IMAGE063
＜ 0.1, show this it is compound have been adjusted to optimum state, iteration terminates, obtain best initial weights
Figure 775557DEST_PATH_IMAGE064
, otherwise go to step（2）.
In above-mentioned specific embodiment, the model parameter of the sorter model of compound characteristic weighing least square method supporting vector machine
Including weight vector optimizing space, initial weight Vector Groups, penalty coefficient, the radial width of gaussian kernel function, reflectance factor, the compressed coefficient, flare factor and iterations.
Embodiment two：When the gear distress to large-scale floating crane equipment carries out classification diagnosis, it is first according in embodiment one
The step of build the sorter model of compound characteristic weighing least square method supporting vector machine, gear distress training sample in the present embodiment is included in gear wear fault sample, rotor pitting fault sample, gear slight crack fault sample and fault-free sample, gear distress training sample correspondence formulax i , the quantity of sample can be selected as needed, and six characteristic attributes that each sample is included are：Peak factor, kurtosis, pulse index, margin index, meshing frequency ratio and rotation frequency ratio, i.e.,
Figure 62181DEST_PATH_IMAGE029
In m=6, represent that each sample has 6 characteristic attributes.In the failure modes algorithm carried herein, weight vector is set firstwOptimizing space be 0 ﹤w≤ 1, initial weight Vector Groups include 10 groups of weight vectors, and penalty coefficient is 1.2896, the quadratic power of the radial width of gaussian kernel function is 0.0881, and reflectance factor is 1.5, and the compressed coefficient is 0.5, flare factor 1.2, iterations is 50 times, and 10 groups of weight vectors are respectively、
Figure DEST_PATH_IMAGE089
、
Figure 532663DEST_PATH_IMAGE090
、
Figure 945190DEST_PATH_IMAGE089
、
Figure DEST_PATH_IMAGE091
、
Figure 35506DEST_PATH_IMAGE092
、
Figure DEST_PATH_IMAGE093
、
Figure 279405DEST_PATH_IMAGE094
、
Figure DEST_PATH_IMAGE095
With
Figure 531395DEST_PATH_IMAGE096
；Gear distress training sample is inputted and is trained in the sorter model of compound characteristic weighing least square method supporting vector machine, optimal weight vector is being exported after 50 iteration
Figure DEST_PATH_IMAGE097
With nicety of grading 97.5%, use the method for the present invention, classification diagnosis precision to the gear distress of large-scale floating crane equipment can reach 97.5%, if under the same conditions, when carrying out classification diagnosis to the gear distress of large-scale floating crane equipment using the model classifiers of least square method supporting vector machine, its classification diagnosis precision is 95%, as shown in table 1.
Table 1：The result table of comparisons of the method for the method of the present invention and prior art to gear distress classification diagnosis
  Least square method supporting vector machine Based on compound characteristic weighing least square method supporting vector machine
（
Figure 493535DEST_PATH_IMAGE098
,
Figure DEST_PATH_IMAGE099
）		 （1.2896,0.0881） (1.2896,0.0881)
Weights
Testing classification precision 95% 97.5%
As it can be seen from table 1 the compound characteristic weighing least square in least square method supporting vector machine and the present embodiment
SVMs all uses identical penalty coefficient and gaussian kernel function parameter（1.7202,0.1382）The purpose is to the training sample for ensureing two class support vector machines is consistent in the distribution of high-dimensional feature space, the weights of least square method supporting vector machine are all chosen for 1, it is 95% to obtain least square method supporting vector machine failure modes precision, and the weights of compound characteristic weighing least square method supporting vector machine are by compound algorithm picks in the present embodiment
Figure DEST_PATH_IMAGE103
, the failure modes precision that the compound characteristic weighing least square method supporting vector machine in final the present embodiment is obtained is 97.5%.Gear distress sample shown in table 2 is the test sample of the sorter model of compound characteristic weighing least square method supporting vector machine.
Table 2：Gear under test failure（Or it is normal）The input data table of sample
Figure 911768DEST_PATH_IMAGE104
In summary, the raising of failure modes precision of the invention is considerable, and iterations is more, and failure modes precision is higher, and when it is applied to the complexity equipment comprising various faults species and test data, safety in production can be played an important role.

Claims (5)

1. the failure modes diagnostic method of a kind of equipment, it is characterised in that comprise the following steps：
（1）The sorter model of construction feature Weighted Least Squares Support Vector Machines；
（2）Build the sorter model of compound characteristic weighing least square method supporting vector machine；
（3）Build the model parameter of the sorter model of compound characteristic weighing least square method supporting vector machine；
（4）The failure training sample of equipment is inputted, the classification to compound characteristic weighing least square method supporting vector machine
Device model is trained；
（5）By the compound characteristic weighing least square method supporting vector machine after the fault sample to be measured input training of equipment
Failure modes diagnosis is carried out in sorter model.
2. a kind of failure modes diagnostic method of equipment described in claim 1, it is characterised in that described step（1）
In the building process of sorter model of characteristic weighing least square method supporting vector machine be：
（1）Build the sorter model of least square method supporting vector machine：Set first
Figure 879414DEST_PATH_IMAGE001
Individual training sample,
x iForiIndividual input data,
Figure 513974DEST_PATH_IMAGE003
,
Figure 96134DEST_PATH_IMAGE004
Representx iThe input space at place, d representation space dimensions,
Figure 582611DEST_PATH_IMAGE005
Represent input datax iCorresponding classification,,
Figure 127566DEST_PATH_IMAGE007
Represent training sample from the input space
Figure 998570DEST_PATH_IMAGE004
To the Nonlinear Mapping of high-dimensional feature space, pass through
Figure 667449DEST_PATH_IMAGE007
By training sample from the input space
Figure 113342DEST_PATH_IMAGE008
High-dimensional feature space is mapped to, the functional expression stated in the classification problem of high-dimensional feature space least square method supporting vector machine is:
             
Figure 539776DEST_PATH_IMAGE009
             s.t.
Figure 542367DEST_PATH_IMAGE010
Wherein min represents to minimize,wThe weight vector of high-dimensional feature space is represented,bBiasing coefficient is represented,c Represent penalty coefficient,
Figure 252703DEST_PATH_IMAGE011
Represent slack variable,
Figure 416968DEST_PATH_IMAGE012
The wrong degree divided is represented, subscript T represents transformation of ownership computing, and s.t. represents constraints；Then Lagrangian is introduced, the expression functional expression for obtaining the minimum value of Lagrangian is：
Figure 596276DEST_PATH_IMAGE013
Wherein
Figure 323930DEST_PATH_IMAGE014
It is Lagrange multiplier,
Figure 639505DEST_PATH_IMAGE015
It is corresponding
Figure 240250DEST_PATH_IMAGE016
For supporting vector, make L pairsw、b、
Figure 893473DEST_PATH_IMAGE011
、
Figure 237867DEST_PATH_IMAGE014
Partial derivative be equal to zero, while introduce kernel function
Figure 673527DEST_PATH_IMAGE017
（i=1,2,3 ... n；j=1,2,3 ... n）, calculate
Figure 366546DEST_PATH_IMAGE018
WithConcrete outcome, according to
Figure 402952DEST_PATH_IMAGE018
With
Figure 942386DEST_PATH_IMAGE019
Result, the categorised decision function for obtaining least square method supporting vector machine is
Figure 619355DEST_PATH_IMAGE020
, that is, construct the sorter model of least square method supporting vector machine；
（2）The sorter model of construction feature Weighted Least Squares Support Vector Machines on the basis of the sorter model of least square method supporting vector machine：First set sample characteristics weighting parameter be
Figure 994973DEST_PATH_IMAGE021
,p=1,2,3 ..., m,
Figure 868120DEST_PATH_IMAGE022
,
Figure 75110DEST_PATH_IMAGE023
Sequence number is characterized, feature weight matrix is constructed on this basis
Figure 126243DEST_PATH_IMAGE024
, least square method supporting vector machine is improved using feature weight matrix, calculating of the smaller feature of weights on nonlinear mapping function and kernel function is influenceed smaller, the classification problem of the least square method supporting vector machine after improvement can be expressed as functional expression：
Figure 51474DEST_PATH_IMAGE025
        
s.t. 
Figure 477381DEST_PATH_IMAGE026
  
Wherein
Figure 742140DEST_PATH_IMAGE027
,
Figure 26491DEST_PATH_IMAGE028
,
Figure 360389DEST_PATH_IMAGE029
,
Figure 326071DEST_PATH_IMAGE030
,
Figure 507654DEST_PATH_IMAGE031
,
Figure 149856DEST_PATH_IMAGE032
, then optimized by method of Lagrange multipliers, while introducing kernel function, can be calculated following matrix equality：
Figure 49679DEST_PATH_IMAGE033
s.t. 
Figure 553473DEST_PATH_IMAGE032
,
Figure 589562DEST_PATH_IMAGE034
,
Figure 402666DEST_PATH_IMAGE035
,
Figure 524206DEST_PATH_IMAGE036
Wherein, gaussian kernel function is：
Figure 831691DEST_PATH_IMAGE037
, wherein σ represents the radial width parameter of gaussian kernel function, and the categorised decision function for obtaining characteristic weighing least square method supporting vector machine is： 
Figure 177746DEST_PATH_IMAGE038
, that is, construct the sorter model of characteristic weighing least square method supporting vector machine.
3. a kind of failure modes diagnostic method of equipment according to claim 1, it is characterised in that described step
（2）In the sorter model of compound characteristic weighing least square method supporting vector machine be on the basis of the sorter model of characteristic weighing least square method supporting vector machine, fitness function is used as using the classification accuracy of characteristic weighing least square method supporting vector machine, selection is iterated to the weight vector of characteristic weighing least square method supporting vector machine by compound algorithm, obtain best initial weights vector, best initial weights vector is brought into the sorter model of characteristic weighing least square method supporting vector machine again, the sorter model of compound characteristic weighing least square method supporting vector machine is obtained.
4. the failure modes diagnostic method of a kind of equipment according to claim 3, it is characterised in that the process that described compound algorithm is iterated selection to the weight vector of characteristic weighing least square method supporting vector machine comprises the following steps：
（1）Generate initial composite typeIndividual weights summit
Figure 521319DEST_PATH_IMAGE040
,
Figure 429233DEST_PATH_IMAGE041
, and it is converted into meeting one by one to the feasible individual set of constraints, using these weights summits as initial weight set；
（2）The nicety of grading of the characteristic weighing least square method supporting vector machine on compound each weights summit is calculated, best initial weights point of the nicety of grading as fitness function is determined
Figure 626865DEST_PATH_IMAGE042
, worst weights point
Figure 532504DEST_PATH_IMAGE043
With suboptimum weights point
Figure 628636DEST_PATH_IMAGE044
；
（3）Calculate the center on compound weights summit：
Figure 527190DEST_PATH_IMAGE045
；
（4）Calculate compound pip
Figure 595641DEST_PATH_IMAGE046
,
Figure 734498DEST_PATH_IMAGE047
For reflectance factor,
Figure 239297DEST_PATH_IMAGE048
；
If the fitness value of pip
Figure 754592DEST_PATH_IMAGE049
Better than current optimum point, then step is gone to（5）；
If the fitness value of pip
Figure 177188DEST_PATH_IMAGE049
Better than current time advantage
Figure 982333DEST_PATH_IMAGE044
, then
Figure 973423DEST_PATH_IMAGE050
, put
Figure 153DEST_PATH_IMAGE051
, go to step（7）；
If the fitness value of pip
Figure 480813DEST_PATH_IMAGE049
It is only most not good enough better than current
Figure 976517DEST_PATH_IMAGE043
, compression point
Figure 568035DEST_PATH_IMAGE052
If, the fitness value of compression point
Figure 714852DEST_PATH_IMAGE053
Better than currently most not good enough
Figure 366413DEST_PATH_IMAGE043
, then
Figure 83833DEST_PATH_IMAGE054
, put, then step is gone to（7）If, the fitness value of compression point
Figure 480365DEST_PATH_IMAGE053
It is worse than current most not good enough, then step is gone to（6）；
If the fitness value of pip
Figure 507544DEST_PATH_IMAGE049
It is worse than current most not good enough
Figure 896325DEST_PATH_IMAGE055
, then compression point
Figure 565203DEST_PATH_IMAGE056
If, the fitness value of compression point
Figure 496250DEST_PATH_IMAGE053
Better than currently most not good enough
Figure 437530DEST_PATH_IMAGE043
, then
Figure 440121DEST_PATH_IMAGE054
, put
Figure 901190DEST_PATH_IMAGE051
, then step is gone to（7）If, the fitness value of compression point
Figure 65455DEST_PATH_IMAGE053
It is worse than current most not good enough
Figure 494031DEST_PATH_IMAGE043
, then step is gone to（6）；
（5）Calculate inflexion point
Figure 972417DEST_PATH_IMAGE057
,
Figure 350309DEST_PATH_IMAGE058
For flare factor,
Figure 138005DEST_PATH_IMAGE059
,
Figure 539030DEST_PATH_IMAGE060
, put
Figure 883424DEST_PATH_IMAGE051
, go to step（7）；
（6）Halve operation, keep optimum point
Figure 293984DEST_PATH_IMAGE042
Constant, remaining each summit is all towards optimum pointIt is close, be close to formula
Figure 953952DEST_PATH_IMAGE061
, put
Figure 23408DEST_PATH_IMAGE051
, go to step（7）；
（7）If
Figure 375892DEST_PATH_IMAGE062
, wherein 0 ＜
Figure 256124DEST_PATH_IMAGE063
＜ 0.1, show this it is compound have been adjusted to optimum state, iteration terminates, obtain best initial weights
Figure 881009DEST_PATH_IMAGE064
, otherwise go to step（2）.
5. the failure modes diagnostic method of a kind of equipment according to claim 1, it is characterised in that the model parameter of the sorter model of described compound characteristic weighing least square method supporting vector machine includes weight vector optimizing space, initial weight Vector Groups, penalty coefficient, the radial width of gaussian kernel function, reflectance factor, the compressed coefficient, flare factor and iterations.
CN2012100691763A 2012-03-15 2012-03-15 Fault classification diagnosis method of equipment Pending CN102706573A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2012100691763A CN102706573A (en) 2012-03-15 2012-03-15 Fault classification diagnosis method of equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2012100691763A CN102706573A (en) 2012-03-15 2012-03-15 Fault classification diagnosis method of equipment

Publications (1)

Publication Number Publication Date
CN102706573A true CN102706573A (en) 2012-10-03

Family

ID=46899516

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2012100691763A Pending CN102706573A (en) 2012-03-15 2012-03-15 Fault classification diagnosis method of equipment

Country Status (1)

Country Link
CN (1) CN102706573A (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103454113A (en) * 2013-09-16 2013-12-18 中国人民解放军国防科学技术大学 Method for monitoring health of rotary machine suitable for working condition changing condition
CN103499924A (en) * 2013-09-22 2014-01-08 浙江大学 Weighted system and method for having control over standard noxious substance emission of pesticide effluent incinerator
CN103512765A (en) * 2013-09-13 2014-01-15 中国科学院苏州生物医学工程技术研究所 Fault detection method for variable learning rate wavelet BP neural network of blood type centrifugal machine
CN104142254A (en) * 2014-08-06 2014-11-12 浙江大学 Fault diagnosis method for flow passage of steam turbine
CN104330257A (en) * 2014-11-26 2015-02-04 天津工业大学 Fault diagnosis method of planetary gear transmission system
CN105158137A (en) * 2015-07-28 2015-12-16 昆明理工大学 Detection method of air permeability of tipping paper based on least square support vector machine
CN105842273A (en) * 2016-03-16 2016-08-10 华北电力大学（保定） Acquiring method and system for compression factor
CN106546439A (en) * 2016-10-13 2017-03-29 南京航空航天大学 A kind of combined failure diagnostic method of hydraulic AGC system
CN108764265A (en) * 2018-03-26 2018-11-06 海南电网有限责任公司电力科学研究院 A kind of method for diagnosing faults based on algorithm of support vector machine
CN109063743A (en) * 2018-07-06 2018-12-21 云南大学 The construction method of medical data disaggregated model based on semi-supervised multi-task learning
CN109635864A (en) * 2018-12-06 2019-04-16 佛山科学技术学院 A kind of fault tolerant control method and device based on data
CN109828230A (en) * 2019-04-02 2019-05-31 国网新疆电力有限公司电力科学研究院 The localization method of automatic calibration of electric energy meter assembly line epitope failure
CN110991295A (en) * 2019-11-26 2020-04-10 电子科技大学 Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network
CN112816553A (en) * 2021-01-22 2021-05-18 国电锅炉压力容器检验有限公司 Heat-resistant steel aging grade evaluation method based on support vector machine
CN113779236A (en) * 2021-08-11 2021-12-10 齐维维 Method and device for problem classification based on artificial intelligence

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103512765A (en) * 2013-09-13 2014-01-15 中国科学院苏州生物医学工程技术研究所 Fault detection method for variable learning rate wavelet BP neural network of blood type centrifugal machine
CN103454113B (en) * 2013-09-16 2016-02-03 中国人民解放军国防科学技术大学 A kind of rotating machinery health monitor method be applicable in working conditions change situation
CN103454113A (en) * 2013-09-16 2013-12-18 中国人民解放军国防科学技术大学 Method for monitoring health of rotary machine suitable for working condition changing condition
CN103499924A (en) * 2013-09-22 2014-01-08 浙江大学 Weighted system and method for having control over standard noxious substance emission of pesticide effluent incinerator
CN103499924B (en) * 2013-09-22 2015-12-23 浙江大学 The pesticide waste liquid incinerator hazardous emission controls up to par system and method for weighting
CN104142254A (en) * 2014-08-06 2014-11-12 浙江大学 Fault diagnosis method for flow passage of steam turbine
CN104142254B (en) * 2014-08-06 2017-01-18 浙江大学 Fault diagnosis method for flow passage of steam turbine
CN104330257A (en) * 2014-11-26 2015-02-04 天津工业大学 Fault diagnosis method of planetary gear transmission system
CN105158137A (en) * 2015-07-28 2015-12-16 昆明理工大学 Detection method of air permeability of tipping paper based on least square support vector machine
CN105842273B (en) * 2016-03-16 2018-08-28 华北电力大学（保定） A kind of acquisition methods and system of compressibility factor
CN105842273A (en) * 2016-03-16 2016-08-10 华北电力大学（保定） Acquiring method and system for compression factor
CN106546439A (en) * 2016-10-13 2017-03-29 南京航空航天大学 A kind of combined failure diagnostic method of hydraulic AGC system
CN106546439B (en) * 2016-10-13 2019-01-15 南京航空航天大学 A kind of combined failure diagnostic method of hydraulic AGC system
CN108764265A (en) * 2018-03-26 2018-11-06 海南电网有限责任公司电力科学研究院 A kind of method for diagnosing faults based on algorithm of support vector machine
CN109063743A (en) * 2018-07-06 2018-12-21 云南大学 The construction method of medical data disaggregated model based on semi-supervised multi-task learning
CN109635864A (en) * 2018-12-06 2019-04-16 佛山科学技术学院 A kind of fault tolerant control method and device based on data
CN109828230A (en) * 2019-04-02 2019-05-31 国网新疆电力有限公司电力科学研究院 The localization method of automatic calibration of electric energy meter assembly line epitope failure
CN110991295A (en) * 2019-11-26 2020-04-10 电子科技大学 Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network
CN110991295B (en) * 2019-11-26 2022-05-06 电子科技大学 Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network
CN112816553A (en) * 2021-01-22 2021-05-18 国电锅炉压力容器检验有限公司 Heat-resistant steel aging grade evaluation method based on support vector machine
CN113779236A (en) * 2021-08-11 2021-12-10 齐维维 Method and device for problem classification based on artificial intelligence

Similar Documents

Publication Publication Date Title
CN102706573A (en) Fault classification diagnosis method of equipment
Zhao et al. Intelligent fault diagnosis of multichannel motor–rotor system based on multimanifold deep extreme learning machine
CN109766921B (en) Vibration data fault classification method based on depth field self-adaption
CN108319962A (en) A kind of Tool Wear Monitoring method based on convolutional neural networks
CN106548230A (en) Diagnosis Method of Transformer Faults based on Modified particle swarm optimization neutral net
CN109186973A (en) A kind of mechanical failure diagnostic method of unsupervised deep learning network
CN108875772B (en) Fault classification model and method based on stacked sparse Gaussian Bernoulli limited Boltzmann machine and reinforcement learning
CN108875771A (en) A kind of failure modes model and method being limited Boltzmann machine and Recognition with Recurrent Neural Network based on sparse Gauss Bernoulli Jacob
CN104751229A (en) Bearing fault diagnosis method capable of recovering missing data of back propagation neural network estimation values
CN111199252A (en) Fault diagnosis method for intelligent operation and maintenance system of power communication network
CN107944349A (en) GIS partial discharge type-scheme recognition methods based on particle swarm support vector machine
CN110132554A (en) A kind of rotary machinery fault diagnosis method that depth Laplce encodes certainly
CN110909802A (en) Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor
Guo et al. Improved adversarial learning for fault feature generation of wind turbine gearbox
CN110363230A (en) Stacking integrated sewage handling failure diagnostic method based on weighting base classifier
CN111310722A (en) Power equipment image fault identification method based on improved neural network
Su et al. Generative adversarial networks for gearbox of wind turbine with unbalanced data sets in fault diagnosis
Gu et al. Identification of concurrent control chart patterns with singular spectrum analysis and learning vector quantization
CN106485326B (en) A kind of hardness detection method in ore reduction production process
Wang et al. An effective multi-channel fault diagnosis approach for rotating machinery based on multivariate generalized refined composite multi-scale sample entropy
CN115587290A (en) Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network
Aqlan et al. Integrating artificial neural networks and cluster analysis to assess energy efficiency of buildings
CN113159225A (en) Multivariable industrial process fault classification method
CN110674893A (en) Self-adaptive correction method for diagnosis experience in rotary machine fault diagnosis knowledge base
CN112257773B (en) Mechanical equipment fault diagnosis method based on multiple measurement vectors of wireless sensor network

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C12 Rejection of a patent application after its publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20121003