CN106682448A - Sequential test optimization method based on multi-objective genetic programming algorithm - Google Patents
Sequential test optimization method based on multi-objective genetic programming algorithm Download PDFInfo
- Publication number
- CN106682448A CN106682448A CN201710103734.6A CN201710103734A CN106682448A CN 106682448 A CN106682448 A CN 106682448A CN 201710103734 A CN201710103734 A CN 201710103734A CN 106682448 A CN106682448 A CN 106682448A
- Authority
- CN
- China
- Prior art keywords
- node
- individuality
- fault
- measuring point
- test
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Abstract
The invention discloses a sequential test optimization method based on a multi-objective genetic programming algorithm. The method comprises the steps of firstly initializing to obtain a fault diagnostic tree population, using the multi-objective genetic programming method to select, cross and mutate a fault diagnostic tree, during the iteration of each generation, grouping the fault diagnosis tree individuals, wherein, the adaptability of each individual is calculated by using grouping adaptability and a crowding distance; after multiple iterations, selecting non-domination individuals from a final generation population as a non-domination fault diagnosis tree of the sequential test of the system. The sequential test optimization method based on the multi-objective genetic programming algorithm can be used for acquiring a Pareto optimum solution of the fault diagnosis tree of the sequential test for which multiple test indexes can be used as optimized targets, the optimum solutions can be selected by testers to provide guidance for system testers.
Description
Technical field
The invention belongs to Fault Diagnosis for Electronic System technical field, more specifically, is related to a kind of based on multiple target something lost
Pass the sequential test optimization method of planning algorithm.
Background technology
In Fault Diagnosis for Electronic System technology, sequential test problem be defined as a five-tuple problem (S, P, T, C,
D).Wherein, S={ s0,s1,s2,…,sMRepresent system fault condition finite aggregate, wherein s0Expression system does not have failure to occur
State, s1To sMThere is the state of different faults in expression system.P={ p0,p1,p2,…,pMIt is that each system mode occurs
Priori probability of malfunction vector.Hypothesis system can be only in certain malfunction or unfaulty conditions, need to priori probability of malfunction
Vectorial P is normalized.T={ t1,t2,…,tNIt is the N number of available test set of system.C={ c1,c2,…,cNRepresent right
The test cost vector answered, wherein test cost is weighed using testing time, testing expense and other correlative factors.D is
The 0-1 matrixes of one N × (M+1) rank, each of which element represents the relation of test and the system failure, is defined as failure-test
Rely on matrix.For test tjIf, tjMalfunction s can be diagnosed to bei, then dij=1, otherwise dij=0, wherein i=0,
1 ..., M, j=1,2 ..., N.Obviously for unfaulty conditions s0, to any j, there is d0j=0.
In the sequential test problem of single goal, the related parameter of the test such as testing time, testing expense is all normalized to
Test cost C, but in actual production, part system (such as testing time, test fee on for different test indexs
With) demand it is not necessarily consistent, i.e., exist inconsistent in Different Optimization target and probably produce conflict, it is therefore desirable to find out
Qualified one group of optimal solution can take into account demand of the system for different test indexs, be mesh more than on the question essence
Mark optimization problem.
In actual applications, people are frequently encountered the problem of objective design and decision-making.In order to solve multiple-objection optimization
Problem, to it establishes a general Mathematical Modeling.First have to determine its decision variable, generally, can be decision-making
Variable X regards n dimension Euclidean space E asnA point, namely：
X=(x1,x2,...,xn)∈En
The decision variable meets following constraints：
gi(x)≤0, i=1,2 ..., P
hj(x)=0, j=1,2 ..., Q
Wherein, giX ()≤0 represents i-th inequality constraints, P represents the quantity of inequality constraints, hjX ()=0 represents jth
Individual equality constraint, Q represents the quantity of equality constraint.
Hypothesis has R optimization aim, and each optimization aim fr(X) collide with each other between, r=1,2 ..., R, it is overall excellent
Change target to be represented by：
SeekMakeIn meet the constraint gi(x)≤0 and hjReach most while (x)=0
It is excellent.
In most of the cases, due to each target of multi-objective optimization question be it is conflicting, sub-goal
Improvement be possible to that the reduction of other sub-goal performances can be caused, want so that multiple targets at the same be optimal be it is impossible,
Thus each sub-goal can only be coordinated when multi-objective optimization question is solved and compromise process, make each specific item scalar functions
All it is optimal as much as possible.Multi-objective optimization question is essentially different with single-object problem, in multiple-objection optimization
Optimal solution, normally referred to as Pareto (Pareto) optimal solution.
If XpAnd XqIt is any two meet the constraint g simultaneouslyi(x)≤0 and hjX the different solutions of ()=0, claim XpDomination Xq, then
It must is fulfilled for following two conditions：
(1) to all of sub-goal, XpUnlike XqDifference, i.e. fr(Xp)≤fr(Xq)；
(2) at least there is a sub-goal, make XpCompare XqIt is good, i.e.,So that fr′(Xp) ＜ fr(Xq)。
In all feasible solutions of multi-objective optimization question, if there is certain solution, the solution is not by other solution dominations, this solution
Referred to as Pareto optimal solutions (non-domination solution).In theory, multi-objective optimization algorithm can after a Pareto optimal solution is found
To stop, but a problem generally has multiple Pareto optimal solutions, and policymaker's needs are selected most suitable according to actual conditions
Pareto optimal solutions, therefore the target of multi-objective optimization question is exactly to find out all of Pareto optimal solutions.
Genetic algorithm is the random search algorithm of analoglike biology natural selection and natural evolution, with implicit parallel
Property and the ability scanned in global solution space, it once runs and just can obtain one group of optimal solution, and genetic planning is hereditary calculation
The extension of method, is a kind of Optimum search technology, mainly by imitating biological evolution and heredity, according to the principle of the survival of the fittest,
Problem to be solved is set to approach optimal solution a step by a step from initial solution, main evolutionary process is initialization of population, selects, intersects
And variation.Individuality in genetic planning normally behaves as tree, has flexible processing structure and size not to predefine and asks
The advantage of topic, is typically usually used in solving the problems, such as Symbolic Regression.Fault diagnosis tree and genetic planning in sequential test problem is individual
Expression is all represented using binary tree structure, therefore the method that sequential test problem can be solved with genetic planning.
In the patent of Publication No. " CN105629156A ", " analog circuit fault based on genetic planning is tested most to disclose one kind
Excellent sequential search method ", the fault diagnosis tree of analog circuit is regarded as in the method an individual for genetic planning, is then adopted
Evolved with the algorithm of genetic planning and obtain optimum individual solution.But one optimization mesh of test cost is only considered in the method
Mark, it is impossible to suitable for there is system of multiple test indexs as optimization aim, needs further improvement.
The content of the invention
It is an object of the invention to overcome the deficiencies in the prior art, there is provided a kind of sequence based on multi-objective Genetic planning algorithm
Test optimization method being passed through, for there is system of multiple test indexs as optimization aim, the fault diagnosis of sequential test is asked for
The Pareto optimal solutions of tree, for system testers guidance is provided.
For achieving the above object, the present invention is included based on the sequential test optimization method of multi-objective Genetic planning algorithm
Following steps：
S1：Matrix is relied on according to the test of system and generates H fault diagnosis tree at random, the size of H determines as needed；
S2：Make iterations w=1；
S3：Each individual test index value F in population is calculated respectivelyhr, h=1,2 ..., H, r=1,2 ..., R, R represents
The quantity of test index；According to test index value FhrIndividuality in population is grouped, the concrete steps of packet include：
S3.1：Initialization set A is equal to current population, makes grouping serial number v=1；
S3.2：Each test index value F is searched for from set AhrThe corresponding individuality of extreme value, if the optimization mesh of test index
Mark is as big as possible, then select the corresponding individuality of maximum, if the optimization aim of test index is as little as possible, is selected most
It is little to be worth corresponding individuality；
S3.3：Not tested desired value F is filtered out from set AhrThe individuality of extreme value domination；
S3.4：V-th packet of individual composition that step S3.2 and step S3.3 are obtained during epicycle is screened, and this is grouped
In individuality delete from set A；
S3.5：Judge whether that set A is sky, if it is, packet terminates, otherwise into step S3.6；
S3.6：Make v=v+1, return to step S3.2；
S4：Calculate each individual fitness value λ in current populationh=αh-0.5βh, wherein αh=0.5 (V-vh+ 1), vh
The grouping serial number belonging to individuality h is represented,GhRepresent that other are individual in addition to the individuality in packet belonging to individuality h
The set of composition, γhh′Crowding distance is represented, its computing formula is as follows：
Wherein, δsharePredeterminable range threshold value is represented,h′∈Gh；
S5：According to each individual fitness that step S4 is calculated, selection obtains father's individual collections；
S6：Individuality in father's individual collections is randomly divided into into H/2 groups, then in units of group, relatively per group in two
Individuality whether there is fault set identical node, if there is no fault set identical node in two individualities in certain group, should
Group does not carry out crossover operation, if two individualities in certain group have fault set identical node, size is deleted first and is equal to M-1
Or the same fault collection equal to 1, M represents number of faults in fault set S, in arbitrarily one event of selection of remaining same fault concentration
Barrier collection, exchanges the failure assembly place and its subtree；
S7：Individuality to obtaining after intersection carries out mutation operation, and its variation method is：It is individual for each in population,
One node of random selection, if the node only includes a failure, reselects, if the node is comprising several failures and has
The available measuring point of correspondence fault set can be split, then one is randomly choosed from the available measuring point of the node with used survey before
The measuring point that point is different from, the subtree with the node as root node is deleted from fault diagnosis tree, with the measuring point weight newly selected
Newly-generated corresponding subtree；
S8：Judge whether that w ＜ W, W represent maximum iteration time, if it is, into step S9, otherwise into step S10；
S9：Make w=w+1, return to step S3；
S10：From finally for searching for each test index value F in populationhrThe corresponding individuality of extreme value, then filter out not tested
Desired value FhrThe individuality of extreme value domination, as the non-dominant fault diagnosis tree of the sequential test of the system.
Sequential test optimization method of the present invention based on multi-objective Genetic planning algorithm, first initialization obtain fault diagnosis
Seeds group, the method planned using multi-objective Genetic is selected fault diagnosis tree, intersected and is made a variation, in per generation iteration,
Fault diagnosis tree individuality is grouped, each individual fitness is passed through by fitness is grouped and crowding distance is calculated
After successive ignition from finally in population select non-dominant individuality as the sequential test of the system non-dominant fault diagnosis tree.
Experiments verify that and understand, the present invention can obtain there are multiple test indexs as the sequential test of optimization aim
The Pareto optimal solutions of fault diagnosis tree, select, so as to provide guidance for system testers for system testers.
Description of the drawings
Fig. 1 is specific embodiment flow process of the present invention based on the sequential test optimization method of multi-objective Genetic planning algorithm
Figure；
Fig. 2 is the generation method of fault diagnosis tree in the present embodiment；
Fig. 3 is population at individual group technology flow chart in the present invention；
Fig. 4 is to be grouped fitness in the present invention to determine schematic diagram；
Fig. 5 is the exemplary plot of fault diagnosis tree individual intersection.
Specific embodiment
The specific embodiment of the present invention is described below in conjunction with the accompanying drawings, so as to those skilled in the art preferably
Understand the present invention.Requiring particular attention is that, in the following description, when known function and design detailed description perhaps
When can desalinate the main contents of the present invention, these descriptions will be ignored here.
Fig. 1 is specific embodiment flow process of the present invention based on the sequential test optimization method of multi-objective Genetic planning algorithm
Figure.As shown in figure 1, the present invention is included based on the concrete steps of the sequential test optimization method of multi-objective Genetic planning algorithm：
S101：Initialization of population：
Initialization of population is to rely on matrix according to the test of system to generate H fault diagnosis tree at random, and the size of H is according to need
Determine.The generation method of fault diagnosis tree can be selected as needed, and Fig. 2 is the generation of fault diagnosis tree in the present embodiment
Method.As shown in Fig. 2 the fault diagnosis tree generation method employed in the present embodiment is comprised the following steps：
S201：Initiation parameter：
Make root node p0For fault set, i.e. p0=S, makes number of plies initial value k=0.
S202：Statistics kth layer fault set number of nodes Dk。
S203：Judge whether Dk=0, if it is not, into step S204, otherwise fault diagnosis tree is generated and terminated.
S204：Make fault set node ID d=1.
S205：Select measuring point：
Determine that the available measuring point collection X=T-T ' of current measuring point node, T ' represent current measuring point father of node node first
Test points set, that is to say, that current measuring point can not repeat with any one measuring point in its upper layer node.From available measuring point collection φ
Arbitrarily selecting one can split fault set SkdMeasuring point t, SkdRepresent the failure corresponding to kth d-th fault set node of layer
Collection, makes measuring point node tkd=t, according to the segmentation result of measuring point t fault set S is determinedkdSub- fault set Skd-leftAnd Skd-right,
If there is no fault set S can be splitkdMeasuring point, then do not do any operation.
Measuring point t is to fault set SkdSegmentation can detect and be out of order to realize according to test, by fault set SkdMiddle energy
The failure for enough being detected by measuring point t divides a sub- fault set, it is impossible to which the failure detected by measuring point t is divided into another height event
Barrier collection, if one of them sub- fault set is sky, illustrates that measuring point t can not be completed to fault set SkdSegmentation.
S206：Judge whether d ＜ Dk, if it is, into step S207, otherwise into step S208.
S207：Make d=d+1, return to step S205.
S208：Make k=k+1, return to step S202.
Fault diagnosis tree generation method according to proposing in the present embodiment there may be the mould that two or several failures are constituted
Paste group finally cannot divided situation, such case may be modified by subsequent operation, it is also possible to due to measuring point
Measuring point is not complete enough in collection T, finally still there may be ambiguity group, therefore as much as possible should all add the measuring point that can be used
Enter measuring point collection T.
S102：Make iterations w=1.
S103：Individual packets：
Each individual test index value F in population is calculated respectivelyhr, h=1,2 ..., H, r=1,2 ..., R, R represents survey
The quantity of examination index.
In the present embodiment, test index is using testing time Jt and testing expense Jc, and computing formula is as follows respectively：
Wherein, p (si) represent failure siThe probability for being occurred, i=1,2 ..., m, m represents number of faults, aij=1 represents
When certain failure siDuring generation, t is testedjThe failure, a can be detectedij=0 represents when failure siCan not tested t during generationjInspection
Measure；ctjTo test tjCorresponding testing time, ccjTo test tjCorresponding test cost.
Then according to test index value FhrIndividuality in population is grouped.Fig. 3 is population at individual packet in the present invention
Method flow diagram.As shown in figure 3, population at individual group technology is comprised the following steps in the present invention：
S301：Initiation parameter：
Initialization set A is equal to current population, makes grouping serial number v=1.
S302：The optimum non-dominant of screening is individual：
Each test index value F is searched for from set AhrThe corresponding individuality of extreme value, if the optimization aim of test index is
It is as big as possible, then the corresponding individuality of maximum is selected, if the optimization aim of test index is as little as possible, select minimum of a value
Corresponding individuality, these individualities possess respectively extreme value of the test index value in current collection A, as current collection A's
Non-domination solution.
By testing expense and as a example by the testing time, the optimization aim of the two test indexs is all as little as possible, therefore is needed
Testing expense is selected to be minimum of a value CbestIndividuality and the testing time be minimum of a value TbestIndividuality.
S303：Screening non-dominant is individual：
Not tested desired value F is filtered out from set AhrThe individuality of extreme value domination.It is with testing expense and testing time
Example, the individual testing expense that is of non-dominant is more than CbestAnd the testing time is less than CbestIndividual and testing cost be less than Tbest
And the testing time is more than TbestIndividuality.
S304：Obtain v-th packet：
V-th packet of individual composition that step S302 and step S303 are obtained during epicycle is screened, and during this is grouped
Individuality is deleted from set A.
It can be seen that, after every time screening obtains v-th packet, remaining individuality is all by the 1st to v-th packet in set A
The individuality of domination, correspondingly, is grouped into the 1st to the v-1 packet domination for v-th, therefore can claim v-th to be grouped into v levels
Non-domination solution, the v-1 packet by before is arranged.
S305：Judge whether that set A is sky, if it is, packet terminates, otherwise into step S306.
S306：Make v=v+1, return to step S302.
S104：Calculate individual fitness：
Population is grouped in the present invention, is defined each packet and be there is a packet fitness, then adopted and gather around
Crowded distance maintaining the diversity of population distribution, therefore each individual fitness be combine its packet fitness and it is crowded away from
From calculating.
The number of packet that note step S103 is obtained is V, the packet fitness α of individuality h in current populationh=0.5 (V-vh+
1), h=1,2 ..., H, vhRepresent the grouping serial number belonging to individuality h.Fig. 4 is to be grouped fitness in the present invention to determine schematic diagram.Such as
Shown in Fig. 4, the 1st packet is not by the packet (the 1st grade of non-domination solution) of other individuality dominations, wherein all individual packets are fitted
Response is 0.5V, and the 2nd packet is to be arranged packet (the 2nd grade of non-domination solution) once, wherein all individual packets are adapted to
Spend for 0.5 (V-1), by that analogy, last packet, i.e., by the packet (V level non-domination solutions) of other all individual dominations
Individual packet fitness be 0.5.
Individual crowding distance can pass through range difference sum of two individualities of calculating same level on each sub-goal
To ask for.The big individual chance for participating in breeding and evolving of crowding distance is more, so as to maintain the diversity of population.Calculate first
Individual h with its be grouped in other individual h ' crowding distance γhh′：
Wherein, δsharePredeterminable range threshold value is represented,h′∈Gh, GhRepresent belonging to individuality h
The set that other individualities are constituted in addition to the individuality in packet, Fh′rRepresent the test index value of individuality h '.
The accumulation crowding distance β of individual hhComputing formula it is as follows：
According to the packet fitness α of individual hhWith accumulation crowding distance βh, it is possible to it is calculated the fitness value of individual h
λh=αh-0.5βh。
It can be seen from the computational methods of individual adaptation degree in the present invention, the bigger individuality of individual adaptation degree, its adaptability
It is stronger, then more may to be retained in new population generating process.
S105：Select father individual：
According to each individual fitness that step S104 is calculated, selection obtains father's individual collections.In father's individual collections
Individual amount remains as H, and this is indicated that, if some fault diagnosis tree individualities are not selected, then just there is other failure to examine
Disconnected tree is chosen number of times and exceedes once.The system of selection of father's individual collections generally adopts wheel disc bet method, it would however also be possible to employ heredity
Other father's individual choice methods in algorithm.
The detailed process of wheel disc bet method is：All individual fitness are cumulative in current population sues for peace, and each is individual
Fitness it is cumulative divided by fitness and obtain the share that the individual adaptation degree accounts for the cumulative sum of fitness, and by the suitable of each individuality
Response share is lined up from small to large.A number between 0 to 1 is generated at random, and this random number is entered with individual share
Row compares, and first possesses fitness share individual selected bigger than current random number and be put into father's individual collections.Such one
H selection is carried out altogether, and all individualities being selected may make up a new population.
S106：Individual intersection：
Individuality in the father's individual collections selected step S105 carries out crossover operation.Intersection in traditional genetic planning
Operation, usually exchanges in two individualities under any two node and node each subtree of institute's band.But the present invention is targeted
Fault diagnosis tree have more conditions to limit compared to general binary tree, main reason is that each subtree of this tree
All it is to be determined by the failure set content and selected measuring point in higher level tree, whole tree contacts from top to bottom very closely, needs
Some special restrictions are carried out to crossover operation.Therefore the concrete grammar of individual intersection is as follows in the present invention：
First the individuality in father's individual collections is randomly divided into into H/2 groups, then in units of group, relatively per group in two
Individuality whether there is fault set identical node, if there is no fault set identical node in two individualities in certain group, should
Group does not carry out crossover operation, if two individualities in certain group have fault set identical node, in order to prevent crossover operation from not having
It is meaningful, first exclude size and be equal to M-1 or the same fault collection equal to 1, M represents number of faults in system failure collection S, then
Remaining same fault is concentrated and arbitrarily select a fault set, exchanges the failure assembly place and its subtree.In order to improve intersection
Meaning is at most only exchanged once with the diversity for ensureing population, per group of individuality.
Fig. 5 is the exemplary plot of fault diagnosis tree individual intersection.As shown in figure 5, having in the two fault diagnosis tree individualities
One includes failure s1、s2、s5And s12Fault set, the node that the two fault sets should be located during crossover operation and
The subtree of each of which is exchanged next in the lump.
S107：Individual variation：
Individuality to obtaining after intersection carries out mutation operation, and its variation method is：It is individual for each in population, at random
A node is selected, if the node only includes a failure, is reselected, if the node is comprising several failures and has and can divide
The available measuring point of correspondence fault set is cut, then one is randomly choosed from the available measuring point of the node with used measuring point before all
The measuring point for differing, the subtree with the node as root node is deleted from fault diagnosis tree, is given birth to again with the measuring point newly selected
Into corresponding subtree.
S108：Judge whether that w ＜ W, W represent maximum iteration time, if it is, into step S109, otherwise into step
S110。
S109：Make w=w+1, return to step S103.
S110：Screen final non-dominant individual：
From finally for searching for each test index value F in populationhrThe corresponding individuality of extreme value, then filter out not tested index
Value FhrThe individuality of extreme value domination, that is, the final non-dominant individuality in population is filtered out, as the sequential test of the system
Non-dominant fault diagnosis tree.
Embodiment
In order to the technique effect of the present invention is better described, simulating, verifying is carried out using a specific embodiment.Table 1 is this
Failure of the electronic system containing probability of malfunction and test cost-test in example relies on matrix.
Table 1
The condition of this experimental verification is as follows：CPU:Pentium G3250；Operating system：Windows 7；Programming language：
JAVA.Genetic planning relevant parameter：Population Size：100；Evolutionary generation：100 crossing-over rates：0.8；Aberration rate 0.05.Adopted
Test index is testing expense and testing time, threshold value δ of crowding distanceshare=0.3.
Sequential test optimization is carried out using the present invention, 12 Pareto optimal solutions are obtained.Table 2 is Pareto in the present embodiment
The testing expense of optimal solution and testing time list.
| Testing time | Testing expense |
| 3.303 | 12.487 |
| 3.784 | 12.118 |
| 4.615 | 11.866 |
| 4.708 | 11.744 |
| 4.724 | 11.275 |
| 4.81 | 9.478 |
| 4.875 | 8.015 |
| 5.035 | 7.994 |
| 5.198 | 7.615 |
| 5.232 | 7.22 |
| 5.311 | 6.601 |
| 5.548 | 4.764 |
Table 2
From table 2 it can be seen that the minimum testing expense of the system is 4.764, now its testing time is 5.548, system
The minimum testing time be 3.303, its corresponding testing expense be 12.487, system cannot obtain simultaneously minimum testing expense
With the time.In Pareto optimal solutions, the testing expense bigger individual testing time is fewer, can be as needed in practical operation
Therefrom select the solution for being best suitable for system.It can be seen that, can obtain there are multiple test indexs as optimization aim using the present invention
The Pareto optimal solutions of the fault diagnosis tree of sequential test, for system testers guidance is provided.
Although being described to illustrative specific embodiment of the invention above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the invention is not restricted to the scope of specific embodiment, the common skill to the art
For art personnel, as long as various change is in the spirit and scope of the present invention of appended claim restriction and determination, these
Change is it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.
Claims (3)
1. a kind of sequential test optimization method based on multi-objective Genetic planning algorithm, it is characterised in that comprise the following steps：
S1：Matrix is relied on according to the test of system and generates H fault diagnosis tree at random, the size of H determines as needed；
S2：Make iterations w=1；
S3：Each individual test index value F in population is calculated respectivelyhr, h=1,2 ..., H, r=1,2 ..., R, R represents test
The quantity of index；According to test index value FhrIndividuality in population is grouped, the concrete steps of packet include：
S3.1：Initialization set A is equal to current population, makes grouping serial number v=1；
S3.2：Each test index value F is searched for from set AhrThe corresponding individuality of extreme value, if the optimization aim of test index is
It is as big as possible, then the corresponding individuality of maximum is selected, if the optimization aim of test index is as little as possible, select minimum of a value
Corresponding individuality；
S3.3：Not tested desired value F is filtered out from set AhrThe individuality of extreme value domination；
S3.4：V-th packet of individual composition that step S3.2 and step S3.3 are obtained during epicycle is screened, and during this is grouped
Individuality is deleted from set A；
S3.5：Judge whether that set A is sky, if it is, packet terminates, otherwise into step S3.6；
S3.6：Make v=v+1, return to step S3.2；
S4：Calculate each individual fitness value λ in current populationh=αh-0.5βh, wherein αh=0.5 (V-vh+ 1), vhRepresent
Grouping serial number belonging to individual h,GhRepresent that other individualities are constituted in addition to the individuality in packet belonging to individuality h
Set, γhh′Crowding distance is represented, its computing formula is as follows：
Wherein, δsharePredeterminable range threshold value is represented,h′∈Gh；
S5：According to each individual fitness that step S4 is calculated, selection obtains father's individual collections；
S6：Individuality in father's individual collections is randomly divided into into H/2 groups, then in units of group, relatively per group in two individualities
With the presence or absence of fault set identical node, if two individualities in certain group do not have fault set identical node, the group is not
Crossover operation is carried out, if two individualities in certain group have fault set identical node, size is deleted first and is equal to M-1 or is waited
In 1 same fault collection, M represents number of faults in system failure collection S, concentrates in remaining same fault and arbitrarily select an event
Barrier collection, exchanges the failure assembly place and its subtree；
S7：Individuality to obtaining after intersection carries out mutation operation, and its variation method is：It is individual for each in population, at random
A node is selected, if the node only includes a failure, is reselected, if the node is comprising several failures and has and can divide
The available measuring point of correspondence fault set is cut, then one is randomly choosed from the available measuring point of the node with used measuring point before all
The measuring point for differing, the subtree with the node as root node is deleted from fault diagnosis tree, is given birth to again with the measuring point newly selected
Into corresponding subtree；
S8：Judge whether that w ＜ W, W represent maximum iteration time, if it is, into step S9, otherwise into step S10；
S9：Make w=w+1, return to step S3；
S10：From finally for searching for each test index value F in populationhrThe corresponding individuality of extreme value, then filter out not tested index
Value FhrThe individuality of extreme value domination, as the non-dominant fault diagnosis tree of the sequential test of the system.
2. sequential test optimization method according to claim 1, it is characterised in that in described step S1, fault diagnosis
The generation method of tree is：
S2.1：Make root node p0For fault set, i.e. p0=S, makes number of plies initial value k=0；
S2.2：Statistics kth layer fault set number of nodes Dk；
S2.3：Judge whether Dk=0, if it is not, into step S2.4, otherwise fault diagnosis tree is generated and terminated；
S2.4：Make fault set node ID d=1；
S2.5：Determine that the available measuring point collection φ=T-T ' of current measuring point node, T ' represent current measuring point father of node node first
Test points set；Arbitrarily selecting one from available measuring point collection φ can split fault set SkdMeasuring point t, SkdRepresent kth layer
Fault set corresponding to d-th fault set node, makes measuring point node tkd=t, according to the segmentation result of measuring point t fault set is determined
SkdSub- fault set Skd-leftAnd Skd-right, if there is no fault set S can be splitkdMeasuring point, then do not do any operation；
S2.6：Judge whether d ＜ Dk, if it is, into step S2.7, otherwise into step S2.8；
S2.7：Make d=d+1, return to step S2.5；
S2.8：Make k=k+1, return to step S2.2.
3. sequential survey optimization method according to claim 1, it is characterised in that test index includes surveying in step S3
Examination time and testing expense.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710103734.6A CN106682448A (en) | 2017-02-24 | 2017-02-24 | Sequential test optimization method based on multi-objective genetic programming algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710103734.6A CN106682448A (en) | 2017-02-24 | 2017-02-24 | Sequential test optimization method based on multi-objective genetic programming algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106682448A true CN106682448A (en) | 2017-05-17 |
Family
ID=58861812
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710103734.6A Pending CN106682448A (en) | 2017-02-24 | 2017-02-24 | Sequential test optimization method based on multi-objective genetic programming algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106682448A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909194A (en) * | 2017-11-07 | 2018-04-13 | 电子科技大学 | System level testing designs Multipurpose Optimal Method |
| CN108090566A (en) * | 2017-12-13 | 2018-05-29 | 电子科技大学 | Multiple target test preferred method based on connection in series-parallel genetic algorithm |
| CN108563875A (en) * | 2018-04-18 | 2018-09-21 | 电子科技大学 | Analog circuit measuring point and frequency based on multiple-objection optimization combine preferred method |
| CN112464551A (en) * | 2020-09-27 | 2021-03-09 | 电子科技大学 | Analog circuit fault parameter range determining method based on non-dominance |
| CN114968828A (en) * | 2022-08-02 | 2022-08-30 | 树优(宁波)科技有限公司 | Performance test method, platform, equipment and storage medium for optimization algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104698365A (en) * | 2015-03-23 | 2015-06-10 | 电子科技大学 | Unreliability test optimizing method based on grouping genetic algorithm |
| CN104899101A (en) * | 2015-06-10 | 2015-09-09 | 合肥工业大学 | Dynamic distributing method of software testing resources based on multi-object difference evolutionary algorithm |
| CN105629156A (en) * | 2016-03-10 | 2016-06-01 | 电子科技大学 | Genetic programming-based analog circuit fault test optimal sequential search method |
| CN106095608A (en) * | 2016-06-17 | 2016-11-09 | 电子科技大学 | Sequential test dynamic adjusting method based on AO* algorithm |
-
2017
- 2017-02-24 CN CN201710103734.6A patent/CN106682448A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104698365A (en) * | 2015-03-23 | 2015-06-10 | 电子科技大学 | Unreliability test optimizing method based on grouping genetic algorithm |
| CN104899101A (en) * | 2015-06-10 | 2015-09-09 | 合肥工业大学 | Dynamic distributing method of software testing resources based on multi-object difference evolutionary algorithm |
| CN105629156A (en) * | 2016-03-10 | 2016-06-01 | 电子科技大学 | Genetic programming-based analog circuit fault test optimal sequential search method |
| CN106095608A (en) * | 2016-06-17 | 2016-11-09 | 电子科技大学 | Sequential test dynamic adjusting method based on AO* algorithm |
Non-Patent Citations (9)
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909194A (en) * | 2017-11-07 | 2018-04-13 | 电子科技大学 | System level testing designs Multipurpose Optimal Method |
| CN108090566A (en) * | 2017-12-13 | 2018-05-29 | 电子科技大学 | Multiple target test preferred method based on connection in series-parallel genetic algorithm |
| CN108563875A (en) * | 2018-04-18 | 2018-09-21 | 电子科技大学 | Analog circuit measuring point and frequency based on multiple-objection optimization combine preferred method |
| CN108563875B (en) * | 2018-04-18 | 2021-05-14 | 电子科技大学 | Multi-objective optimization-based combined optimization method for measuring points and frequencies of analog circuit |
| CN112464551A (en) * | 2020-09-27 | 2021-03-09 | 电子科技大学 | Analog circuit fault parameter range determining method based on non-dominance |
| CN112464551B (en) * | 2020-09-27 | 2023-06-02 | 电子科技大学 | Non-dominant analog circuit fault parameter range determining method |
| CN114968828A (en) * | 2022-08-02 | 2022-08-30 | 树优(宁波)科技有限公司 | Performance test method, platform, equipment and storage medium for optimization algorithm |
| CN114968828B (en) * | 2022-08-02 | 2022-11-04 | 树优(宁波)科技有限公司 | Performance test method, platform, equipment and storage medium for optimization algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106682448A (en) | Sequential test optimization method based on multi-objective genetic programming algorithm | |
| CN105629156B (en) | Analog circuit fault based on genetic planning tests optimal sequential search method | |
| CN106250667A (en) | The monitoring method of a kind of landslide transition between states of paddling and device | |
| CN106874963B (en) | A kind of Fault Diagnosis Method for Distribution Networks and system based on big data technology | |
| CN110473592A (en) | The multi-angle of view mankind for having supervision based on figure convolutional network cooperate with lethal gene prediction technique | |
| CN112330050A (en) | Power system load prediction method considering multiple features based on double-layer XGboost | |
| CN107169871B (en) | Multi-relationship community discovery method based on relationship combination optimization and seed expansion | |
| CN105471647B (en) | A kind of power communication network fault positioning method | |
| CN108682457A (en) | Patient's long-term prognosis quantitative forecast and interfering system and method | |
| CN102609612A (en) | Data fusion method for calibration of multi-parameter instruments | |
| CN106569030A (en) | Alarm threshold optimizing method and device in electric energy metering abnormity diagnosis | |
| CN112580902B (en) | Object data processing method and device, computer equipment and storage medium | |
| CN104834987A (en) | Quantitative decision-making method based on layer analysis, and quantitative decision-making system | |
| CN108563875A (en) | Analog circuit measuring point and frequency based on multiple-objection optimization combine preferred method | |
| CN110111606A (en) | A kind of vessel traffic flow prediction technique based on EEMD-IAGA-BP neural network | |
| CN106685749B (en) | The method of inspection and device of network flow | |
| CN104102694B (en) | Tree node sort method and tree node collator | |
| CN112232526A (en) | Geological disaster susceptibility evaluation method and system based on integration strategy | |
| CN108446712A (en) | ODN nets Intelligent planning method, apparatus and system | |
| CN113192556B (en) | Genotype and phenotype association analysis method in multigroup chemical data based on small sample | |
| CN110334085A (en) | Power distribution network data monitoring and modification method, device, computer and storage medium | |
| CN110109005B (en) | Analog circuit fault testing method based on sequential testing | |
| CN112926664B (en) | Feature selection and CART forest short-time strong precipitation prediction method based on evolutionary algorithm | |
| CN108830407A (en) | Sensor distribution optimization method under the conditions of multi-state in monitoring structural health conditions | |
| CN115146702A (en) | Transformer fault diagnosis method, medium and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170517 |
|
| RJ01 | Rejection of invention patent application after publication |