CN106886467B - Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution - Google Patents
Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution Download PDFInfo
- Publication number
- CN106886467B CN106886467B CN201710102212.4A CN201710102212A CN106886467B CN 106886467 B CN106886467 B CN 106886467B CN 201710102212 A CN201710102212 A CN 201710102212A CN 106886467 B CN106886467 B CN 106886467B
- Authority
- CN
- China
- Prior art keywords
- individual
- evolution
- multitask
- elite
- target
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/0703—Error or fault processing not based on redundancy, i.e. by taking additional measures to deal with the error or fault not making use of redundancy in operation, in hardware, or in data representation
- G06F11/0751—Error or fault detection not based on redundancy
Abstract
Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution the invention discloses a kind of, multi-objective Evolutionary Algorithm is respectively adopted to each mission mode first to obtain testing preferred elite individual collections, wherein the individual in multi-objective Evolutionary Algorithm is that testing scheme selects vector, evolution target is evolution target of the preset testability index under the mission mode, the constraint condition of each evolution target is constraint condition of the testability index under the mission mode, then it is preferred multitask integration test to be carried out using multi-objective Evolutionary Algorithm according to the elite individual collections of each mission mode, the element of its individual is serial number of each mission mode elite individual in corresponding set, multitask testing scheme selection vector is obtained according to the elite individual collections acquired, as preferred non-domination solution is tested in multitask.Multitask under multi-target condition can be got more quickly to using the present invention and test preferred non-domination solution, and solving result can be made more accurate.
Description
Technical field
The invention belongs to Technique of Fault Diagnosis in Systems fields, more specifically, are related to a kind of based on the more mesh of grouping-synthesis
Preferred method is tested in the multitask that mark is evolved.
Background technique
As semiconductor integrated circuit develops to integration and miniaturization both direction, electronic system becomes increasingly complex,
Measuring point is set also more and more inconvenient in circuit, since measuring point sharply reduces, fault diagnosis difficulty is caused to increase.Often grind
This and the inverted situation of maintenance cost is made, increases maintenance personal's number, when requiring to get higher, train to their industrial grade
Between it is elongated.In order to mitigate the maintenance difficulties of equipment in the future, system should just consider Testability Design in the initial stage of design.
Testability refers to the degree that the state of system can be detected accurately.For the quality of evaluation system measurability, need
Testability Analysis is carried out to system.Testable parameter refers to using the malfunction of simulation to tests such as the measuring point settings of system
Property design test, test degree in the hope of reasonably estimating system.
In the troubleshooting issue for large scale electronic equipment system, how testing scheme is selected, make fault detection rate
(FDR, fault diagnose rate), false alarm rate (FAR, fault alarm rate) and test every expense (time,
Economy etc.) etc. testabilities index meet constraint condition simultaneously and even tend to more preferable, be that academic or engineering field is constantly explored
The problem of.
For the above test optimal selection problem for considering multiple testability indexes simultaneously, multi-objective optimization question can be considered as.
Multi-objective optimization question is to discuss how under certain constraints to find to meet multiple targets and be attained by optimal solution.One
As in the case of, be between each sub-goal of multi-objective optimization question it is contradictory, the improvement of a sub-goal is possible to cause
The reduced performance of another or another several sub-goals, that is, to make simultaneously multiple sub-goals be optimal together value be can not
Can, and can only carry out coordinating among them and compromise processing, make each sub-goal all being optimal as much as possible.More mesh
The mathematical form of mark optimization problem can be described as follows:
Min y=f (x)=[f1(x),f2(x),…,fn(x)…,fN(x)], n=1,2 ..., N
gi..., (x)≤0, i=1,2 P
hj..., (x)=0, j=1,2 Q
X=[x1,x2,…,xd,…,xD]
xd_min≤xd≤xd_max, d=1,2 ..., D
Wherein, x indicates that D ties up decision vector, and y indicates object vector, fn(x) indicate that n-th of objective function, N indicate optimization
Target sum, gi(x)≤0 i-th of inequality constraints is indicated, P indicates the quantity of inequality constraints, hj(x)=0 indicate j-th etc.
Formula constraint, Q indicate the quantity of equality constraint, and X indicates the decision space that decision vector is formed, and Y indicates the mesh that object vector is formed
Mark space.giAnd h (x)≤0j(x)=0 the feasible zone understood, x are determinedd_maxAnd xd_minFor the bound of every dimensional vector search.
Essential distinction with single-object problem is that the solution of multi-objective optimization question is simultaneously not exclusive, but there are one
The optimal solution set that group is made of numerous Pareto (Pareto) optimal solution, each element in set are known as Pareto optimal solution
Or Pareto optimal.
Multi-objective Evolutionary Algorithm is an analoglike biological heredity mechanism and the probability optimization searching method of overall importance formed,
The 1990s, mid-term started to rapidly develop, and development can be divided into two stages: the first stage, there are mainly two types of methods i.e.
It is not based on the method for Pareto optimization and the method based on Pareto optimization;Second stage is exactly outside proposing on this basis
Portion collects this concept, and outside collection storage is all non-dominant individuals for working as former generation, so that disaggregation be made to keep preferable degree of distribution.
The multi-objective genetic algorithm that this period proposes more emphasizes the efficiency and validity of algorithm.In the two stages, compare
Typical multi-objective genetic algorithm has NSGA-II.For this algorithm, advantage be time complexity is reduced, but
It is that its disadvantage is also obvious, Li et al. people is it is demonstrated experimentally that crowding distance (crowding distance) plan in NSGA-II
Slightly even reaction can be played to the evolution of population on higher-dimension multi-objective problem.
The basic principle of multi-objective Evolutionary Algorithm is described as follows: the population that multi-objective Evolutionary Algorithm generates at random from one group goes out
Hair, by executing selection to population, intersecting and the evolutional operations such as variation, through excessive for evolution, in population, individual fitness is not
It is disconnected to improve, thus the Pareto optimal solution set of Step wise approximation multi-objective optimization question.It is different from single goal evolution algorithm, multiple target
Evolution algorithm has special fitness evaluation mechanism.It is most of in order to give full play to the collective search advantage of evolution algorithm
MOEA is all made of the fitness evaluation method based on Pareto sequence.
In electronic system malfunction testing field, with increasingly complicated, the single task test optimal selection problem of Large-Scale Equipment system
Gradually it is evolved into multitask test optimal selection problem.For a complication system, only each subtask testing scheme runs succeeded,
Final overall test task can just succeed.In order to ensure overall system runs succeeded, each submodule also will be executed into all
Function correctly executes completely for example, guided missile will ensure to explode and destroy target, its submodule is (energy supply, Target Acquisition, steady
Determine device and navigation) also have to all correct execute.So the incipient fault of modules must be detected.
Based on the above demand, the multitask of system is tested in design, and each mission mode can have the test of their own to need
It asks.The purpose that measuring point is chosen is that optimal test set is selected from original test set, meets all individual task modes several
Testing requirement.And in final overall test scheme, also to meet overall testing requirement.Comparing traditional single task test needs
It asks, multitask test design has that optimization aim is excessive, constraint condition is excessive, if it is excellent directly to continue to use single task test
Algorithm is selected, then there can be the disadvantages of runing time is too long, disaggregation search is incomplete, last solution is not global non-domination solution.In addition,
In multitask test preferably, need to carry out each mission mode and overall tasks to test preferably.So multitask is surveyed
Optimal selection problem is tried, if only simply combined global optimization target and single task optimization aim, using multiple target
Evolution algorithm is solved, and will cause that target is too many, the excessively high problem of dimension, this be to search solution it is unfavorable, fitted
The improvement of answering property.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of based on grouping-synthesis multi-target evolution
Preferred method is tested in multitask, realizes that the multitask test of efficiently and accurately under multi-target condition is preferred.
For achieving the above object, the present invention is based on the multitasks of grouping-synthesis multi-target evolution to test preferred method
The following steps are included:
S1: matrix is relied on according to the multitask test that system information obtains system, selects multitask to survey according to actual needs
The testability index of examination preferably reference, including fault detection rate or Percent Isolated determine system totality and each mission mode
Under the evolution target of each testability index and the constraint condition of fault detection rate or Percent Isolated;Define testing scheme
Choose vector X=[b1,b2,…,bQ], q=1,2 ..., Q, Q indicate testing scheme quantity, bqValue is 0 or 1, bq=0 indicates the
Q testing scheme is not chosen, bq=1 indicates that the q testing scheme is chosen, and remembers testability index F under m-th of mission moden's
Calculation formula isN=1,2 ..., N, N indicate testability index quantity, m=1,2 ..., M, and M indicates task
Mode quantity remembers the testability index F of system totalitynCalculation formula be
S2: multi-objective Evolutionary Algorithm is respectively adopted to each mission mode and obtains testing preferred elite individual collections Ym,
Wherein the individual in multi-objective Evolutionary Algorithm is that testing scheme selects vector X=[b1,b2,…,bQ], evolution target is step S1
Selected in N number of testability index FnEvolution target under the mission mode, the constraint condition of each evolution target are to survey
Constraint condition of the examination property index under the mission mode, each individual is tested according to multitask and relies on matrix according to the task
Testability index calculation formula under modeCalculate its evolution target value;
S3: according to the elite individual collections Y of the obtained each mission mode of step S2m, multi-objective Evolutionary Algorithm is set
Individual be X '=[x1′,x2′,…,x′M], wherein x 'mIndicate the elite individual serial number of m-th of mission mode, value range
For 1≤x 'm≤|Ym|, | Ym| indicate elite individual collections YmIn elite individual amount;Evolution target is selected in step S1
N number of testability index FnIn the evolution target of system totality, the constraint condition of each evolution target is that testability index is being
The constraint condition for totality of uniting;Each individual is tested according to multitask and relies on matrix according to the meter of system overall test index
Calculate formulaIts evolution target value is calculated, wherein Xall=y (x1′)|y(x2′)|…|y(x′M), y (x 'm) table
Show elite individual collections YmMiddle xth 'mA elite individual;
The multi-objective Evolutionary Algorithm is run according to arrangement above, obtains elite individual collections Z, wherein each elite individual Zd
=[zd1,zd2,…,zdM] correspond to a multitask testing scheme selection vectorI.e.zdmThe m task mould in d-th of elite individual in expression elite individual collections Z
The elite individual serial number of formula, y (zdm) indicate elite individual collections YmIn zdmA elite individual, d=1,2 ..., | Z |, | Z | table
Show elite individual amount in elite individual collections Z;To obtain | Z | a multitask testing scheme selects vector, as multitask
Test preferred non-domination solution.
The present invention is based on the multitasks of grouping-synthesis multi-target evolution to test preferred method, first to each mission mode
Multi-objective Evolutionary Algorithm is respectively adopted to obtain testing preferred elite individual collections, wherein the individual in multi-objective Evolutionary Algorithm is
Testing scheme selects vector, and evolution target is evolution target of the preset testability index under the mission mode, each evolution
The constraint condition of target is constraint condition of the testability index under the mission mode, then according to the elite of each mission mode
Individual collections are preferred using multi-objective Evolutionary Algorithm progress multitask integration test, and the element of individual is each mission mode essence
Serial number of the English individual in corresponding set, after obtaining the preferred elite individual collections of multitask integration test, according to the essence acquired
English individual collections obtain multitask testing scheme selection vector, and preferred non-domination solution is tested in as multitask.Using the present invention
Multitask under multi-target condition can be got more quickly to than direct integrated solution and tests preferred non-domination solution, and is avoided that non-
The generation of global non-domination solution keeps solving result more accurate.
Detailed description of the invention
Fig. 1 is that the present invention is based on the specific embodiments of grouping-synthesis multi-target evolution multitask test preferred method
Flow chart;
Fig. 2 is the flow chart of multi-objective Genetic evolution algorithm employed in the present embodiment.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art
Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps
When can desalinate main contents of the invention, these descriptions will be ignored herein.
Embodiment
The present invention tests preferred feature according to multitask, using grouping-synthesis multi-target evolution, to realize that multitask is surveyed
Examination is preferred.Fig. 1 is that the present invention is based on the specific embodiment streams of grouping-synthesis multi-target evolution multitask test preferred method
Cheng Tu.As shown in Figure 1, the present invention is based on the specific steps packets of grouping-synthesis multi-target evolution multitask test preferred method
It includes:
S101: system related data is obtained:
Matrix is relied on according to the multitask test that system information obtains system, selects multitask test excellent according to actual needs
The testability index of reference, including fault detection rate or Percent Isolated are selected, determines that system is each totally and under each mission mode
The evolution target of a testability index and the constraint condition of fault detection rate or Percent Isolated.Testing scheme is defined to choose
Vector X=[b1,b2,…,bQ], q=1,2 ..., Q, Q indicate testing scheme quantity, bqValue is 0 or 1, bq=0 indicates q-th
Testing scheme is not chosen, bq=1 indicates that q-th of testing scheme is chosen, and remembers testability index F under m-th of mission modenCalculating
Formula isN=1,2 ..., N, N indicate testability index quantity, m=1,2 ..., M, and M indicates mission mode
Quantity remembers the testability index F of system totalitynCalculation formula be
The multitask test of system, which relies on matrix, to be obtained by collection system information, is referred to wherein containing testability
The data being marked on each measuring point.Table 1 is that the fault detection rate of multitask test relies on matrix example.
Table 1
As shown in table 1, smIndicate that m-th of mission mode, m=1,2 ..., M, M indicate mission mode quantity; fmkIndicate the
K-th of fault mode, k=1,2 ..., K in m mission modem, KmIndicate fault mode quantity in m-th of mission mode;λmk
Indicate the probability of malfunction of k-th of fault mode in m-th of mission mode;tqIndicate q-th of testing scheme, q=1,2 ..., Q, Q
Indicate testing scheme quantity;pmIndicate the corresponding probability of malfunction of m-th of mission mode, cqIndicate the corresponding survey of q-th of testing scheme
Try time overhead, dmkqIndicate q-th of testing scheme to the fault detection rate of k-th of fault mode in m-th of mission mode.
In test optimal selection problem, testability index faulty verification and measurement ratio (FDR, fault usually of interest
Diagnose rate), Percent Isolated (FIR, fault isolation rate), false alarm rate (FAR, fault alarm
Rate), testing time expense (TC, time cost) and the economic expense (PC, price cost) of test etc., can basis
Actual needs is to be selected.To simplify the explanation, selection considers fault detection rate FDR, false alarm rate FAR and survey in the present embodiment
These three testability indexes of time overhead TC are tried, can be calculated according to the dependence matrix of multitask test test.With fault detection
For rate FDR, the fault detection rate FDR of each mission modemCalculation formula it is as follows:
The fault detection rate FDR of system entiretyallCalculation formula it is as follows:
False alarm rate FAR is approximate with the calculation formula of fault detection rate FDR, and the testing time calculation formula of expense TC isThe evolution target and constraint condition of each testability index is arranged according to the actual situation.With this
For embodiment, need fault detection rate FDR as high as possible, false alarm rate FAR and testing time expense TC are as low as possible, and therefore
Barrier verification and measurement ratio FDR needs to be more than or equal to preset minimum threshold, so as to obtain evolution target and the constraint in the present embodiment
Condition is expressed as follows:
Wherein,Indicate the fault detection rate threshold value of m-th of mission mode,The failure of expression system totality
Verification and measurement ratio threshold value.
S102: the individual task mode test based on multi-target evolution is preferred:
Each mission mode is obtained testing preferred elite individual collections Y using multi-objective Evolutionary Algorithmm, wherein more mesh
Marking the individual in evolution algorithm is that testing scheme selects vector X=[b1,b2,…,bQ], evolution target is selected in step S101
The N number of testability index F selectednThe constraint condition of evolution target under the mission mode, each evolution target refers to for testability
The constraint condition being marked under the mission mode tests each individual according to multitask and relies on matrix according under the mission mode
Testability index calculation formulaCalculate its evolution target value.
Several multi-objective Evolutionary Algorithms are had existed in industry at present, specific algorithm can carry out according to actual needs
Selection, by performance test, the present embodiment uses the preferable multi-objective Genetic evolution algorithm of effect.Fig. 2 is institute in the present embodiment
The flow chart of the multi-objective Genetic evolution algorithm of use.It is calculated as shown in Fig. 2, multi-objective Genetic employed in the present embodiment is evolved
The specific steps of method include:
S201: initialization of population:
It is random to generate Q individual composition initial population A in the value range of multi-target evolution individual, each individual
Score score=α initializes elite individual collectionsThe number of iterations t=1.Obviously, preferred in the test of individual task mode
When, the individual in population is that the testing scheme that element is different valued combinations selects vector.The quantity Q of initial population can root
It is determined according to actual needs.
S202: new individual is generated:
It is new that two individual progress cross and variations generation Q are randomly choosed according to score score individual in current population A
Individual, score score is bigger, bigger by select probability.
S203: individual sequence:
The new individual generated in step S202 and current population A are merged into new population B, by individual each in new population B
Score score be reduced to initial value α, then according to it is following rule carry out ranking: calculate population B in each individual evolution
Target value (individual task mode test it is preferred when according toCalculated), if evolution target value meets in advance
If constraint condition, then it is corresponding individual be known as feasible solution;Then two individuals are not repeatedly selected from population B, if two
Individual is feasible solution, then according to regular ranking is dominated, enables the side score=0.9score dominated, a side of domination
Any operation is not made;If only a side is feasible solution, one side of the feasible solution does not make any operation, the score=of another party
0.9score;If both sides are not feasible solution, the farthest side score=0.9score of target value distance constraints, separately
One side does not make any operation;All individuals have been traversed to rear, descending arrangement is carried out to individual according to score score.
S204: selection elite individual:
The individual that score score=α is selected from population B, if the evolution target value of the individual is unsatisfactory for constraint condition,
Or repeated with the individual in elite individual collections Y, or dominated by the individual in elite individual collections Y, then do not make any operation, it is no
Then the individual is added in elite individual collections Y;It is propped up if there are original individuals in elite individual collections Y by the new individual that is added
Match, then deletes these original individuals, otherwise do not make any operation.
S205: judging whether that the number of iterations reaches threshold value, i.e. t < T, and T indicates preset the number of iterations threshold value, if so,
S206 is entered step, otherwise multi-objective Genetic evolution algorithm terminates.
S206: next-generation population is selected:
The individual in population B is carried out since the individual amount in population B is twice of population A, and in step S203
Sequence, therefore the first half individual collections in selected population B enable t=t+1, return step S202 as new population A.
According to step S201 to step S206, multi-target evolution genetic algorithm passes through T iteration, obtains an elite individual
Gather, the individual in the elite individual collections is non-dominant mutually, so being a Pareto optimal solution set.
S103: the multitask integration test based on multi-target evolution is preferred:
According to the elite individual collections Y of the obtained each mission mode of step S102m, multi-objective Evolutionary Algorithm is set
Individual is X '=[x1′,x2′,…,x′M], wherein x 'mIndicate that the elite individual serial number of m-th of mission mode, value range are
1≤x′m≤|Ym|, | Ym| indicate elite individual collections YmIn elite individual amount;Evolution target is selected in step S101
N number of testability index FnIn the evolution target of system totality, the constraint condition of each evolution target is that testability index is being
The constraint condition for totality of uniting;Each individual is tested according to multitask and relies on matrix according to the meter of system overall test index
Calculate formulaIts evolution target value is calculated, wherein Xall=y (x1′)|y(x2′)|…|y(x′M), y (x 'm) table
Show elite individual collections YmMiddle xth 'mA elite individual, | expression is asked every dimension element of vector or operation.
The multi-objective Evolutionary Algorithm is run according to arrangement above, obtains elite individual collections Z, wherein each elite individual Zd
=[zd1,zd2,…,zdM] correspond to a multitask testing scheme selection vectorI.e.zdmThe m task mould in d-th of elite individual in expression elite individual collections Z
The elite individual serial number of formula, y (zdm) indicate elite individual collections YmIn zdmA elite individual, d=1,2 ..., | Z |, | Z | table
Show elite individual amount in elite individual collections Z;To obtain | Z | a multitask testing scheme selects vector, as multitask
Test preferred non-domination solution.
Assuming that there are 3 mission modes, 4 testing schemes, the elite individual collections Y of the 1st mission mode for certain system1=
{ (0,1,0,1), (0,1,1,0) }, the elite individual collections Y of the 2nd mission mode2={ (1,1,0,1) }, the 3rd task mould
The elite individual collections Y of formula3={ (0,1,0,1), (0,1,1,0), (1,1,0,0) }, it is seen that the elite individual of 3 mission modes
Quantity be respectively 2,1,3.So when the multitask integration test based on multi-target evolution is preferred, it is assumed that multi-target evolution is calculated
The individual of method is X '=[1,1,3], then its corresponding testing scheme selects vector Xall=(0,1,0,1) | (1,1,0,1) |
(1,1,0,0)=(1,1,0,1).As it can be seen that in the multitask integration test the present invention is based on multi-target evolution is preferred, by a
The setting of body contacts the elite individual of individual task mode and the individual of system totality, thus by grouping-synthesis come
Realize that multitask test is preferred.
Similarly, the multi-objective Evolutionary Algorithm in this step can also select according to actual needs, same in the present embodiment
Sample uses multi-objective Genetic evolution algorithm as shown in Figure 2, tests preferred phase with the individual task mode based on multi-target evolution
Than being individual difference, calculating and the constraint condition difference of evolution target except difference.
Technical effect in order to better illustrate the present invention, by taking a practical multitask test optimal selection problem as an example, directly
Whole ask is carried out using multi-objective Evolutionary Algorithm (global optimization target and single task model-based optimization targeted integration are as target)
Solution, and being solved using the method for exhaustion, being solved time and result and use solution time of the invention and result carries out pair
Than.Table 2 is the solution time and Comparative result of three kinds of algorithms.
Table 2
As shown in table 2, the solution time spent using the present invention is minimum, although the quantity of obtained non-domination solution can not
Reach the non-domination solution quantity of the method for exhaustion, but more than the non-domination solution for directlying adopt multi-objective Evolutionary Algorithm progress integrated solution
Quantity, and obtained non-domination solution is global non-domination solution, therefore from the point of view of comprehensive solution time and solving result, the present invention
Optimal for the preferred solution effect of multitask test.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art
For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these
Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.
Claims (2)
1. a kind of test preferred method based on the multitask of grouping-synthesis multi-target evolution, which is characterized in that including following step
It is rapid:
S1: matrix is relied on according to the multitask test that system information obtains system, selects multitask test excellent according to actual needs
The testability index of reference, including fault detection rate or Percent Isolated are selected, determines that system is each totally and under each mission mode
The evolution target of a testability index and the constraint condition of fault detection rate or Percent Isolated;Define testing scheme selection
Vector X=[b1,b2,…,bQ], q=1,2 ..., Q, Q indicate testing scheme quantity, bqValue is 0 or 1, bq=0 indicates q-th
Testing scheme is not chosen, bq=1 indicates that q-th of testing scheme is chosen, and remembers testability index F under m-th of mission modenCalculating
Formula isM=1,2 ..., M, M indicate mission mode quantity, remember the testability index F of system totalitynMeter
Calculating formula is
S2: multi-objective Evolutionary Algorithm is respectively adopted to each mission mode and obtains testing preferred elite individual collections Ym, wherein more
Individual in target evolution algorithm is that testing scheme selects vector X=[b1,b2,…,bQ], evolution target is selected in step S1
The N number of testability index F selectednThe constraint condition of evolution target under the mission mode, each evolution target refers to for testability
The constraint condition being marked under the mission mode tests each individual according to multitask and relies on matrix according under the mission mode
Testability index calculation formulaCalculate its evolution target value;
S3: according to the elite individual collections Y of the obtained each mission mode of step S2m, the individual of multi-objective Evolutionary Algorithm is set
For X '=[x '1,x′2,…,x′M], wherein x 'mIndicating the elite individual serial number of m-th of mission mode, value range is 1≤
x′m≤|Ym|, | Ym| indicate elite individual collections YmIn elite individual amount;Evolution target is N number of selected in step S1
Testability index FnIn the evolution target of system totality, the constraint condition of each evolution target is testability index in system totality
Constraint condition;Each individual is tested according to multitask and relies on matrix according to the calculation formula of system overall test indexIts evolution target value is calculated, wherein Xall=y (x '1)|y(x′2)|…|y(x′M), y (x 'm) indicate elite
Individual collections YmMiddle xth 'mA elite individual;
The multi-objective Evolutionary Algorithm is run according to arrangement above, obtains elite individual collections Z, wherein each elite individual Zd=
[zd1,zd2,…,zdM] correspond to a multitask testing scheme selection vectorI.e.zdmM-th of task mould in d-th of elite individual in expression elite individual collections Z
The elite individual serial number of formula, y (zdm) indicate elite individual collections YmIn zdmA elite individual, d=1,2 ..., | Z |, | Z | table
Show elite individual amount in elite individual collections Z;To obtain | Z | a multitask testing scheme selects vector, as multitask
Test preferred non-domination solution.
2. preferred method is tested in multitask according to claim 1, which is characterized in that the multi-objective Evolutionary Algorithm uses
Multi-objective Genetic evolution algorithm, specific steps include:
S2.1: it is random to generate Q individual composition initial population A in the value range of multi-target evolution individual, each individual
Score score=α initializes elite individual collectionsThe number of iterations t=1;
S2.2: it is new that two individual progress cross and variations generation Q are randomly choosed according to score score individual in current population A
Individual, score score is bigger, bigger by select probability;
S2.3: the new individual generated in step S2.2 and current population A are merged into new population B, by individual each in new population B
Score score be reduced to initial value α, then according to it is following rule carry out ranking: calculate population B in each individual evolution mesh
Scale value, if evolution target value meets preset constraint condition, corresponding individual is known as feasible solution;Then from population B not
Two individuals are repeatedly selected, if two individuals are feasible solutions, according to regular ranking is dominated, enable the side dominated
A side of score=0.9score, domination do not make any operation;If only a side is feasible solution, one side of the feasible solution is not
Make any operation, the score=0.9score of another party;If both sides are not feasible solution, target value distance constraints are most
A remote side score=0.9score, another party do not make any operation;All individuals have been traversed to rear, according to score pairs of score
Individual carries out descending arrangement;
S2.4: selecting the individual of score score=α from population B, if the evolution target value of the individual is unsatisfactory for constraint item
Part, or repeated with the individual in elite individual collections Y, or dominated by the individual in elite individual collections Y, then do not make any behaviour
Make, otherwise the individual is added in elite individual collections Y;If there are original individuals by new addition in elite individual collections Y
Body is dominated, then deletes these original individuals, otherwise do not make any operation;
S2.5: if the preset the number of iterations threshold value of the number of iterations t < T, T expression, enters step S2.6, otherwise multi-objective Genetic
Evolution algorithm terminates;
S2.6: the first half individual collections in selected population B enable t=t+1, return step S2.2 as new population A.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710102212.4A CN106886467B (en) | 2017-02-24 | 2017-02-24 | Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710102212.4A CN106886467B (en) | 2017-02-24 | 2017-02-24 | Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106886467A CN106886467A (en) | 2017-06-23 |
CN106886467B true CN106886467B (en) | 2019-02-01 |
Family
ID=59179439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710102212.4A Expired - Fee Related CN106886467B (en) | 2017-02-24 | 2017-02-24 | Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106886467B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107657311B (en) * | 2017-11-03 | 2019-10-29 | 电子科技大学 | Test preferred method based on multi-objective particle swarm algorithm |
CN107909194B (en) * | 2017-11-07 | 2021-07-13 | 电子科技大学 | System-level testability design multi-objective optimization method |
CN110346748A (en) * | 2019-07-05 | 2019-10-18 | 国网福建省电力有限公司电力科学研究院 | A kind of electric power metering system fault testing optimization method |
CN111563549B (en) * | 2020-04-30 | 2023-07-28 | 广东工业大学 | Medical image clustering method based on multitasking evolutionary algorithm |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103838820A (en) * | 2013-12-24 | 2014-06-04 | 西安电子科技大学 | Evolutionary multi-objective optimization community detection method based on affinity propagation |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103294595B (en) * | 2013-05-23 | 2016-06-22 | 西安电子科技大学 | A kind of software repair based on genetic algorithm |
CN104063593B (en) * | 2014-06-18 | 2017-01-18 | 电子科技大学 | Board-level circuit testability index calculation method |
CN104698365B (en) * | 2015-03-23 | 2017-10-31 | 电子科技大学 | Unreliable test method for optimizing based on block encoding genetic algorithm |
CN105629156B (en) * | 2016-03-10 | 2018-04-13 | 电子科技大学 | Analog circuit fault based on genetic planning tests optimal sequential search method |
CN105842607A (en) * | 2016-03-21 | 2016-08-10 | 北京航空航天大学 | Test point quantitative selection method and device in testing design |
CN105955927B (en) * | 2016-04-28 | 2018-11-09 | 电子科技大学 | Constrained optimization method based on decomposition-paralleling genetic algorithm |
CN105929690B (en) * | 2016-04-29 | 2018-07-24 | 南京信息工程大学 | A kind of Flexible Workshop Robust Scheduling method based on decomposition multi-objective Evolutionary Algorithm |
-
2017
- 2017-02-24 CN CN201710102212.4A patent/CN106886467B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103838820A (en) * | 2013-12-24 | 2014-06-04 | 西安电子科技大学 | Evolutionary multi-objective optimization community detection method based on affinity propagation |
Also Published As
Publication number | Publication date |
---|---|
CN106886467A (en) | 2017-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106886467B (en) | Preferred method is tested based on the multitask of grouping-synthesis multi-target evolution | |
Ge et al. | Nonlinear process monitoring based on linear subspace and Bayesian inference | |
US8489247B1 (en) | Agent-based chaotic control of wind turbines | |
CN108563806B (en) | Engine air passage parameter long-range forecast method and system based on similitude | |
Zhang et al. | Cloud-based in-situ battery life prediction and classification using machine learning | |
CN105320727A (en) | Method for detecting anomalies in real time series | |
CN104156615A (en) | Sensor test data point anomaly detection method based on LS-SVM | |
CN107133695A (en) | A kind of wind power forecasting method and system | |
Villez et al. | Optimal flow sensor placement on wastewater treatment plants | |
CN110352389A (en) | Information processing unit and information processing method | |
JP7067234B2 (en) | Data discrimination program, data discrimination device and data discrimination method | |
CN114266289A (en) | Complex equipment health state assessment method | |
Uyun et al. | Feature selection for multiple water quality status: Integrated bootstrapping and SMOTE approach in imbalance classes | |
Tabakhpour et al. | Neural network model for false data detection in power system state estimation | |
CN107909194A (en) | System level testing designs Multipurpose Optimal Method | |
Ahmed | Short-term electrical load demand forecasting based on lstm and rnn deep neural networks | |
CN105203327B (en) | A kind of gas circuit measurement parameter selection method applied to engine air passage analysis | |
CN108830407B (en) | Sensor distribution optimization method in structure health monitoring under multi-working condition | |
Sanz et al. | Leakage localization in water networks using fuzzy logic | |
US20220215144A1 (en) | Learning Apparatus, Learning Method and Learning Program | |
Di et al. | Multisensor data fusion in testability evaluation of equipment | |
Wang et al. | Feature selection methods in the framework of mrmr | |
CN112184037B (en) | Multi-modal process fault detection method based on weighted SVDD | |
CN114821322A (en) | Small sample remote sensing image classification method and system based on attention mechanism | |
Zhang et al. | A hybrid adversarial domain adaptation network for bearing fault diagnosis under varying working conditions |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for 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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190201 |