CN106776088A - Diagnosis method for system fault based on Malek models - Google Patents
Diagnosis method for system fault based on Malek models Download PDFInfo
- Publication number
- CN106776088A CN106776088A CN201611037746.5A CN201611037746A CN106776088A CN 106776088 A CN106776088 A CN 106776088A CN 201611037746 A CN201611037746 A CN 201611037746A CN 106776088 A CN106776088 A CN 106776088A
- Authority
- CN
- China
- Prior art keywords
- node
- individuality
- population
- fitness
- individual
- 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
- 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
- G06F11/079—Root cause analysis, i.e. error or fault diagnosis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention discloses a kind of diagnosis method for system fault based on Malek models, comprise the following steps:Fault-free node method generation initial population is specified under Malek models;Calculate the fitness of individuality in population and judge whether contain the individuality that fitness value is 1 in population;Selection operation, optimal save strategy;Mutation operation;Crossover operation simultaneously judges whether to meet t diagnosable systems;Calculate the fitness of individuality in new population and judge whether there is the individuality that fitness is 1 in population.Diagnostic method of the invention improves the efficiency of positioning fault set, in combination with Malek comparison models, also superior to traditional pmc model in terms of the accuracy of target faults collection is judged by the characteristic such as parallel, ability of searching optimum is strong of genetic algorithm.The present invention is applied in system fault diagnosis problem, can more accurately and rapidly find out target faults collection.
Description
Technical field
The present invention relates to intelligent trouble diagnosis algorithm, in particular, provide a kind of system failure based on Malek models and examine
Disconnected method.
Background technology
The arrival in big data epoch, allows multi-computer system to carry substantial amounts of data, information, algorithm etc., but, cover a large amount of
Once breaking down, how accurately and rapidly to find out failure node is the current significant problem for facing to the multi-computer system of PC.
The thinking of Methods for Diagnosing System Level Malfunctions is to set up suitable diagnostic model by the communication capacity of node itself, and combination is effectively examined
Disconnected algorithm finds out failure collection.At present have six kinds of diagnostic models, they be respectively PMC, BGM, Chwa&Hakimi,
Malek, MM and MM* model.Based on test or compare and be divided into two major classes according between node:Test model and comparison model.Survey
The principle of die trial type is to allow between node mutually test, and whether the expection according to tested node feedack and test node
It is identical to judge node state.And in comparison model, allow neighborhood of nodes to complete identical test assignment, by comparing their knot
Fruit judges the state of node.
Malek diagnostic models, i.e., based on the Asymmetric Model for comparing, its basic thought is:Two processors are allowed to be held first
Then acquired results are compared by the same item task of row.If result is identical, two processors all think that other side is fault-free
Machine;Otherwise (i.e. result is different), this two processors all think that other side is failure machine.The a collection of survey for obtaining is tested in this way
Examination report, determines a n rank symmetrical matrix S=(sij)n×n,sij∈ { ± 1,0 }, on the contrary the matrix also determines this group test
Report.
It is random complete that genetic algorithm (GeneticAlgorithm, GA) is that natural imitation circle biological evolution mechanism grows up
Office's search and optimization method, its essence are a kind of efficient, parallel, methods of global search, and it can automatically be obtained in search procedure
The knowledge about search space is taken and accumulates, adaptively command deployment process is in the hope of optimal solution.The operation of genetic algorithm is right
As if a group binary string (individuality), i.e. population.Each individuality, from initial population, is used corresponding to a solution of problem
Selection strategy based on adaptive value (being produced by fitness function) ratio selects individuality in current population, uses hybridization and variation
To produce population of future generation.Evolution so generation upon generation of is gone down, until meeting desired end condition.System is solved using genetic algorithm
The research of system troubleshooting issue has a lot, and this also indicates that genetic algorithm shows good effect in terms of system fault diagnosis
Really.But the system diagnosability algorithm based on Malek diagnostic models yet there are no.
The content of the invention
It is an object of the invention to be directed to above mentioned problem of the prior art, based on Malek diagnostic models, node shape is designed
The state constraint equation compatible with disease is compared, there is provided a kind of diagnosis method for system fault based on Malek models, the method application
Onto system fault diagnosis problem, target faults collection can be more accurately and rapidly found out.
For achieving the above object, present invention employs following technical scheme:
A kind of diagnosis method for system fault based on Malek models, comprises the following steps:
Step 1:Fault-free node method generation initial population is specified under Malek models, in the initial population for being generated
I.e. one binary string of each individuality corresponds to multi-computer system, and each in individuality is bit corresponding in system
Node;
Step 2:According to the constraint equation of Malek modelling node states, suitable fitness function, Ran Houji are designed
Individual fitness in population is calculated, judges in population whether containing fitness value to be 1 individuality according to fitness, if not having,
Perform step 3;If so, then direct output system failure collection, diagnosis terminates;
Step 3:Following genetic manipulation is carried out to population:
3.1 selection operations:Using roulette selection method, the cumulative select probability of each individuality is calculated, fitness is high
Individuality is preferentially selected as follow-on individuality;Meanwhile, excellent operation is deposited in introducing, and the next generation is substituted with fitness highest is individual
The corresponding individuality of middle fitness minimum value;
3.2 mutation operations:Made a variation using binary system, according to the fitness and mutation probability p of individual each node in populationm
Certain node in random selection population enters row variation, will be changed between 0 and 1;
3.3 crossover operations:According to fitness individual in population, p is randomly choosed from populationcPopsize is individual,
Wherein pcCrossover probability is represented, popsize represents number individual included in population;Again this is randomly choosed from remaining population
The pairing of a few bodies is individual, takes single-point cross method, randomly chooses a crosspoint, and crossover operation is carried out to two father's strings,
Obtain new heredity individual;
3.4 judge that state for whether the number of network nodes t of failure meets t- diagnosable systems, that is, is sentenced in individuality after crossover operation
Whether disconnected state meets 2 × t+1≤n conditions for the number of network nodes t of failure, if meeting, performs step 4;If it is not satisfied, then weighing
It is new to perform above-mentioned steps 3.1~3.3;
Step 4:Each individual fitness in new population is calculated, whether is judged in population containing the individuality that fitness is 1,
If so, then direct output system failure collection, diagnosis terminates;If no, continuing to allow population loop iteration to perform step 3, directly
To finding out the individuality that fitness is 1.
Comprised the following steps using specified fault-free node method generation initial population under Malek models in above-mentioned steps 1:
(1) in comprising the n multi-computer system of node, a state of node k is randomly assigned for fault-free;
(2) according to the degree of node k, the node j adjacent with node k is found, if S (k, j)=0, then it represents that node k and node
The test result of j is identical, and the state of node j is fault-free;If S (k, j)=1, then it represents that the test result of node k and node j
Differ, the state of node j is faulty, and wherein S (k, j) is represented and the feelings of same test task are being distributed to node k and node j
Under condition, the comparative result between two nodes;
(3) each state for obtaining step (2) carries out step 2 for trouble-free node;
(4) node to unknown state is randomly assigned state;
(5) judge whether state meets t- diagnosable systems for the number of network nodes t of failure, if it is not satisfied, then return performing
Step (1);If meeting, step (6) is performed;
(6) according to step (1)~(5) loop iteration popsize times, you can obtain the initial population that size is popsize.
Above-mentioned steps 2 comprise the following steps:
State f (the u of node under definition Malek modelsk,uj) with compare side ekjMeet following constraint equation:
Design fitness function:Dd (v [k])=| { ekj∈ E and meet f (uk,uj) |, wherein, dd (v [k]) represent with
Node k adjoining node and meet constraint equation f (uk,uj) node number and;
Calculate the fitness of node j in individuality i:Wherein dd (v [j]) is represented and node j adjoinings
Node j1And meet constraint equationNode number and, d (v [j]) represents the degree of node j;
Calculate each individual fitness in population:Wherein FT (v [i]) represents individuality i's
Fitness, includes n node in n expression systems.
The method of the selection operation described in above-mentioned steps 3 is:
Fitness highest individuality in population is stored in Elite, wherein Elite represents adaptive value in every generation population
Highest is individual;
Calculate individual adaptation degree summation in population:Wherein FT (v [i]) represents individual
The fitness of i;
Calculate probability of each individuality in wheel disc:Wherein piRepresent individuality i in wheel disc
Probability;
Calculate the cumulative select probability of each individuality:Wherein qiRepresent individuality i's
Cumulative select probability;
Fitness individual in population after selecting is calculated, and the corresponding individuality of minimum adaptive value is replaced with Elite.
The method of the mutation operation described in above-mentioned steps 3 is:All each individual nodes in searching loop population
Fitness value, if fitness value fv (i, j)<Rand (), then select for node j in individual i to enter row variation, wherein rand ()
Represent a function for being used to generate the random number between 0~1.
Compared to prior art, advantage of the invention is that:
The method that the present invention is provided is applied in extensive processor system, and the Malek models based on distributed diagnostics can
Greatly reduce the load of center machine, and the overall situation of genetic algorithm, parallel search capabilities improve search efficiency, make algorithm higher
Search system failure collection to effect.The method provided with the present invention can solve the problems, such as system fault diagnosis, shorten judgement system
Time needed for fault set, reduce the loss that the system failure is brought.
Brief description of the drawings
Fig. 1 opens up complement for multi-computer system node of the present invention.
Fig. 2 is selection opertor operation chart of the present invention.
Fig. 3 is mutation operator operation chart of the present invention.
Fig. 4 is the single-point interior extrapolation method flow chart of individuality in population of the present invention.
Fig. 5 is the flow chart of diagnosis algorithm of the present invention.
Specific embodiment
With reference to embodiments and its accompanying drawing is further non-limitingly described in detail to technical solution of the present invention.
According to any two processors are faulty or trouble-free various situations, the test result of Malek models can be by table 1
To describe, wherein 1 represents failure, 0 represents fault-free.
Numbering | The test result of node k | The test result of node j | Comparative result S (k, j) |
1 | 0 | 0 | 0 |
2 | 0 | 1 | 1 |
3 | 1 | 0 | 1 |
4 | 1 | 1 | 1 |
The test result of the Malek models of table 1
As shown in figure 5, the diagnosis method for system fault based on Malek models, comprises the following steps:
Step 1:Fault-free node method generation initial population is specified under Malek models, in the initial population for being generated
I.e. one binary string of each individuality corresponds to multi-computer system, and each in individuality is bit corresponding in system
Node;Specific method comprises the following steps:
(1) in comprising the n multi-computer system of node, a state of node k is randomly assigned for fault-free;
(2) according to the degree of node k, the node j adjacent with node k is found, if S (k, j)=0, then it represents that node k and node
The test result of j is identical, and the state of node j is fault-free;If S (k, j)=1, then it represents that the test result of node k and node j
Differ, the state of node j is faulty, and wherein S (k, j) is represented and the feelings of same test task are being distributed to node k and node j
Under condition, the comparative result between two nodes;
(3) each state for obtaining step (2) carries out step 2 for trouble-free node;
(4) node to unknown state is randomly assigned state;
(5) judge whether state meets t- diagnosable systems for the number of network nodes t of failure, if it is not satisfied, then return performing
Step (1);If meeting, step (6) is performed;
(6) according to step (1)~(5) loop iteration popsize times, you can obtain the initial population that size is popsize;
Step 2:According to the constraint equation of Malek modelling node states, suitable fitness function, Ran Houji are designed
Individual fitness in population is calculated, judges in population whether containing fitness value to be 1 individuality according to fitness, if not having,
Perform step 3;If so, then direct output system failure collection, diagnosis terminates;Specific method comprises the following steps:
State f (the u of node under definition Malek modelsk,uj) with compare side ekjMeet following constraint equation:
Design fitness function:Dd (v [k])=| { ekj∈ E and meet f (uk,uj) |, wherein, dd (v [k]) represent with
Node k adjoining node and meet constraint equation f (uk,uj) node number and;
Calculate the fitness of node j in individuality i:Wherein dd (v [j]) is represented and node j adjoinings
Node j1And meet constraint equationNode number and, d (v [j]) represents the degree of node j;
Calculate each individual fitness in population:Wherein FT (v [i]) represents individuality i's
Fitness, includes n node in n expression systems;
Step 3:Following genetic manipulation is carried out to population:
3.1 selection operations:Using roulette selection method, the cumulative select probability of each individuality is calculated, fitness is high
Individuality is preferentially selected as follow-on individuality;Meanwhile, excellent operation is deposited in introducing, and the next generation is substituted with fitness highest is individual
The corresponding individuality of middle fitness minimum value;Specific method comprises the following steps:
Fitness highest individuality in population is stored in Elite, wherein Elite represents adaptive value in every generation population
Highest is individual;
Calculate individual adaptation degree summation in population:Wherein FT (v [i]) represents individual
The fitness of i;
Calculate probability of each individuality in wheel disc:Wherein piRepresent individuality i in wheel disc
Probability;
Calculate the cumulative select probability of each individuality:Wherein qiRepresent individuality i's
Cumulative select probability;
Fitness individual in population after selecting is calculated, and the corresponding individuality of minimum adaptive value is replaced with Elite;
3.2 mutation operations:Made a variation using binary system, according to the fitness and mutation probability p of individual each node in populationm
Certain node in random selection population enters row variation, will be changed between 0 and 1;
The fitness value of all each individual nodes in searching loop population, if fitness value fv (i, j)<Rand (),
Then select for node j in individual i to enter row variation, wherein rand () is a function in MATLAB softwares, for generating one
Random number between individual 0~1;
3.3 crossover operations:According to fitness individual in population, p is randomly choosed from populationcPopsize is individual,
Wherein pcCrossover probability is represented, popsize represents number individual included in population;Again this is randomly choosed from remaining population
The pairing of a few bodies is individual, takes single-point cross method, randomly chooses a crosspoint, and crossover operation is carried out to two father's strings,
Obtain new heredity individual;
3.4 judge that state for whether the number of network nodes t of failure meets t- diagnosable systems, that is, is sentenced in individuality after crossover operation
Whether disconnected state meets 2 × t+1≤n conditions for the number of network nodes t of failure, if meeting, performs step 4;If it is not satisfied, then weighing
It is new to perform above-mentioned steps 3.1~3.3;
Step 4:Each individual fitness in new population is calculated, whether is judged in population containing the individuality that fitness is 1,
If so, then direct output system failure collection, diagnosis terminates;If no, continuing to allow population loop iteration to perform step 3, directly
To finding out the individuality that fitness is 1.
As shown in figure 1, the system for describing node a~h8 node open up complement (more node is had in actual conditions,
Merely just illustrated with 8 nodes), each node represents a processor.Represent that numbering is 1~8 in circle respectively
Node, represents node state outside circle, 1 represents failure, and 0 represents fault-free.Each system is constituted with a binary coding
Individuality represents, such as 1:10000100, the 1st node and the 6th node are failure node in expression system, other nodes without
Failure.
As shown in Fig. 2 represent the wheel disc of individual A~F6 individuality (i.e. chromosome), each individual cartographic represenation of area its quilt
The probability of selection, area is bigger, and probability is bigger, then selected probability is bigger, conversely, the probability being eliminated is bigger.And each
Individual selected probability is determined by its fitness value.
As shown in figure 3, mutation process individual in population, randomly selects father individuality a1 as variation in parent population
Body, randomly selects the 4th position and enters row variation, and variation becomes 0 by 1, obtains sub- individuality m1.
As shown in figure 4, crossover operation individual in population, randomly selects father individuality a1, then random choosing from population first
Take father's individuality a2 to intersect therewith, in Fig. 4 since the 6th position, new son individuality m1 and son individuality is produced after two individual intersections
m2。
The method for generating initial population:(1) majority vote method generation initial population;(2) initial population is generated at random;(3)
Specify fault-free node method generation initial population.The essence of majority vote method is if its nothing is thought on most test sides of node
Failure, then the node state is fault-free.Random generation initial population method is relatively common, for example, the work that genetic algorithm is carried
Tool case can at random be quickly generated initial population.And specify fault-free node method to mainly use the communication of node itself
Ability simultaneously judges node state with reference to corresponding diagnostic model, and correlative study shows, the method judges target faults collection
Probability is higher.
Selection operation:Genetic algorithm is selected by the individuality with stronger vitality and adaptive capacity to environment,
Make what colony developed towards more preferable state, the operation of selection is exactly by outstanding individual replicate to follow-on evolution colony
Go, the outstanding degree of individuality is judged by fitness value, more outstanding selected probability is higher, conversely, lower.Selection operation
It is important leverage that algorithm is smoothed out, it is ensured that the global convergence and computational efficiency of algorithm are because it can will be useful
Hereditary information is selected and allows it to enter in follow-on evolution.The good and bad direct relation of selection opertor the quality of solving result,
Influence of the operator to genetic algorithm is very important.Selection needs prudent, otherwise may result in similar individuality in colony
Growth drastically and cause algorithm stagnation, search efficiency is low.Or the generation of super individual is may result in, make colony
Developing direction occur deviateing optimal, have to local optimum, whole algorithm is precocious because diversity glides.More typical selection
Operation has:Roulette selection (roulette wheel selection), elitist selection (elitist selection), sequence
Selection (rank selection), random ergodic sampling (stochastic universal sampling), algorithm of tournament selection
(tournament selection) etc..
What is used in the invention is roulette selection method, and the method is a kind of conventional random selection method, and individuality exists
Fitness in population is proportionally converted to selected probability, and the ratio shared by individuality carries out ratio on a disk and draws
Divide, treat that the corresponding individuality in disk stopping backpointer stop sector is the individuality chosen after disk is rotated every time.Individual fitness is got over
Greatly, selected probability is bigger.
Mutation operation:In the biological evolutionary process of nature, due to some genes caused by some accidental factors
Change produce new biomorph be referred to as variation.Genetic algorithm is changed by less probability to genes of individuals code
To simulate this variation.In the mutation operator of genetic algorithm, in the coded strings of chromosome, with the equipotential base on the locus
Because replacing existing gene, so that it may reach the effect of variation.Relative to the ability of searching optimum that crossover operator is embodied, variation
Operator is directed to the local search ability of algorithm, and new individual that mutation is obtained is there is a possibility that algorithm enters into new sky
Between, jump out current limitations.Fig. 3 describes the mutation process of individuality.For binary coding, variation is exactly by ' 0 ' by genic value
It is changed into ' 1 ', ' 0 ' is changed into from ' 1 '.
For mutation operator, according to the individual fitness and mutation probability p of each in populationmIn random selection population
Row variation is entered in a certain position, and detailed process is the fitness value of all each individual nodes in searching loop population, if fitness
Value fv (i, j)<Rand (), then select the node j of individuality i to enter row variation.Failure nodal point number is during individuality need not be judged after variation
It is no to meet t- diagnosable systems, but after three evolution operations are finished, that is, after having performed crossover operation, then judge individuality
Whether middle failure nodal point number meets t- diagnosable systems.Judge in the final step of each iteration, reduce the time of algorithm
Complexity.
Crossover operation:Intersection is important link in biological heredity and evolutionary process, and the biology in nature is not by
Disconnected evolution and obtain new individuality and species, this evolution is to select two chromosomes to carry out pairing intersection, enables gene
Recombinate and form new chromosome.In order to imitate this bioprocess, genetic algorithm constructs crossover operator to realize individuality
Pairing intersects, and crossover operator is that two individualities are selected from parent population by certain probability, by certain section of gene of the two
Value is swapped, so as to obtain two new individualities.Mainly there are arithmetic crossover operator, uniformity crossover, multiple-spot detection at present
Operator, single-point crossover operator and two point crossover operator etc..
Single-point, two point, the crossover operator of multiple spot are much like crossover operations, and their main distinction is to intersect site
Difference, the mode that single-point intersects generation new individual is to determine a site at random, with this site as basic point, individuality is divided into
Two parts, one is first half, and one is latter half.Then, a crossover probability is generated by certain way, by two
Individual first half or latter half is interchangeable.It is then two sites of random selection that two point intersects, and individuality is divided into three
Individual part, individuality is exchanged with each other the part between two sites to produce the individuality of filial generation by certain crossover probability.By single-point
With the concept of two point it can be concluded that multiple-spot detection operator is exactly multiple sites, the genic value between site is carried out respectively
Exchange obtains new individuality, and this mode has greatly destructive and is rarely used to the mode configuration of chromosome.It is uniform to hand over
The fork difference maximum with other interleaved modes is the homogeny for having crossover probability, and this homogeny refers to the gene on locus
Crossover probability is identical, and locus is directed to each locus on two individualities that will be intersected.Usual arithmetic crossover
It is the crossover operation taken when coded system is floating-point encoding, is to produce two by two linear combinations of individuality
It is individual new individual.It is well known that every kind of crossover operator has respective characteristic, according to the different suitable sides of problem selection
Formula, plays its advantage, avoids its inferior position.
Embodiment 1
To verify the performance of algorithm of the present invention, algorithm is write using Matlab language, and in an internal memory
4.00GB, CPU are to be tested on the computer of Core (TM) i52.5GHz.
Based on the diagnosis method for system fault of Malek models, comprise the following steps:
Step 1:Fault-free node method generation initial population is specified under Malek models, in the initial population for being generated
I.e. one binary string of each individuality corresponds to multi-computer system, and each in individuality is bit corresponding in system
Node;Specific method comprises the following steps:
(1) in comprising the n multi-computer system of node, a state of node k is randomly assigned for fault-free;
(2) according to the degree of node k, the node j adjacent with node k is found, if S (k, j)=0, then it represents that node k and node
The test result of j is identical, and the state of node j is fault-free;If S (k, j)=1, then it represents that the test result of node k and node j
Differ, the state of node j is faulty, and wherein S (k, j) is represented and the feelings of same test task are being distributed to node k and node j
Under condition, the comparative result between two nodes;
(3) each state for obtaining step (2) carries out step 2 for trouble-free node;
(4) node to unknown state is randomly assigned state;
(5) judge whether state meets t- diagnosable systems for the number of network nodes t of failure, if it is not satisfied, then return performing
Step (1);If meeting, step (6) is performed;
(6) according to step (1)~(5) loop iteration popsize times, you can obtain the initial population that size is popsize;
Emulation experiment is carried out to fault-free node method specified above generation initial population method, first for Setup Experiments are closed
Suitable parameter, it is 20 to set Population Size popsize, and number of network nodes n increases to 300 from 10, compares pmc model and Malek moulds
Initial population is generated by specifying fault-free node method under type, the probability of target faults collection is produced.The simulation experiment result shows,
Averagely correct individuality probability is for averagely correctly individuality probability is 0.9933 under 0.8533, Malek models under pmc model.Illustrate
Specifying fault-free node method combination Malek models can play more preferable effect.
Step 2:According to the constraint equation of Malek modelling node states, suitable fitness function, Ran Houji are designed
Individual fitness in population is calculated, judges in population whether containing fitness value to be 1 individuality according to fitness, if not having,
Perform step 3;If so, then direct output system failure collection, diagnosis terminates;Specific method comprises the following steps:
State f (the u of node under definition Malek modelsk,uj) with compare side ekjMeet following constraint equation:
Design fitness function:Dd (v [k])=| { ekj∈ E and meet f (uk,uj) |, wherein, dd (v [k]) represent with
Node k adjoining node and meet constraint equation f (uk,uj) node number and;
Calculate the fitness of node j in individuality i:Wherein dd (v [j]) is represented and node j adjoinings
Node j1And meet constraint equationNode number and, d (v [j]) represents the degree of node j;
Calculate each individual fitness in population:Wherein FT (v [i]) represents individuality i's
Fitness, includes n node in n expression systems;
After devising new fitness function, the calculating fitness method to the present embodiment is briefly described.
Input:All individualities in population
Output:Individual fitness in population
Step 3:Following genetic manipulation is carried out to population:
3.1 selection operations:Using roulette selection method, the cumulative select probability of each individuality is calculated, fitness is high
Individuality is preferentially selected as follow-on individuality;Meanwhile, excellent operation is deposited in introducing, and the next generation is substituted with fitness highest is individual
The corresponding individuality of middle fitness minimum value;Specific method comprises the following steps:
Fitness highest individuality in population is stored in Elite, wherein Elite represents adaptive value in every generation population
Highest is individual;
Calculate individual adaptation degree summation in population:Wherein FT (v [i]) represents individual
The fitness of i;
Calculate probability of each individuality in wheel disc:Wherein piRepresent individuality i in wheel disc
Probability;
Calculate the cumulative select probability of each individuality:Wherein qiRepresent individuality i's
Cumulative select probability;
Fitness individual in population after selecting is calculated, and the corresponding individuality of minimum adaptive value is replaced with Elite;
3.2 mutation operations:Made a variation using binary system, according to the fitness and mutation probability p of individual each node in populationm
Certain node in random selection population enters row variation, will be changed between 0 and 1;
The fitness value of all each individual nodes in searching loop population, if fitness value fv (i, j)<Rand (),
Then select for node j in individual i to enter row variation, wherein rand () is a function in MATLAB softwares, for generating one
Random number between individual 0~1;
Because iteration is needed by three operations each time in genetic algorithm:Selection, variation intersects;The present invention takes friendship
Whether failure judgement nodal point number meets t- diagnosable systems again after fork.Due to judging whether to meet t- diagnosable after variation every time
System, can limit mutation operation, cause population to generate fitness value individuality high slower.This algorithm each iteration last
Judge after step i.e. intersection, the complexity of the time of algorithm can be reduced.
3.3 crossover operations:According to fitness individual in population, p is randomly choosed from populationcPopsize is individual,
Wherein pcCrossover probability is represented, popsize represents number individual included in population;Again this is randomly choosed from remaining population
The pairing of a few bodies is individual, takes single-point cross method, randomly chooses a crosspoint, and crossover operation is carried out to two father's strings,
Obtain new heredity individual;
3.4 judge that state for whether the number of network nodes t of failure meets t- diagnosable systems, that is, is sentenced in individuality after crossover operation
Whether disconnected state meets 2 × t+1≤n conditions for the number of network nodes t of failure, if meeting, performs step 4;If it is not satisfied, then weighing
It is new to perform above-mentioned steps 3.1~3.3;
Step 4:Each individual fitness in new population is calculated, whether is judged in population containing the individuality that fitness is 1,
If so, then direct output system failure collection, diagnosis terminates;If no, continuing to allow population loop iteration to perform step 3, directly
To finding out the individuality that fitness is 1.
For genetic manipulation selects suitable genetic parameter:Such as Population Size popsize, mutation probability pm, crossover probability pc。
The selection of Population Size popsize:
Forefathers' research shows that effect is relatively good when Population Size is between 7~10.By emulation experiment, popsize=7
When, the mean iterative number of time of algorithm is relatively low, therefore, we choose popsize=7;
Mutation probability pmSelection:
In popsize=7, crossover probability p is madec=0.01, number of network nodes n increase to 110, mutation probability p from 5mFrom 0.1
Increase to 0.9, emulation experiment has been carried out to algorithm, compare under Different Variation probability, the mean iterative number of time of algorithm.Experimental result
Show, mutation probability pmWhen=0.7, the mean iterative number of time of algorithm is minimum, therefore, mutation probability pm=0.7 is that optimal variation is general
Rate.
Crossover probability pcSelection:
Because hybrid individual is 2pcpopsize in cross method, so crossover probability pcNo more than 0.5.Order
Popsize=7, pm=0.7, number of network nodes n increase to 200 from 10, and crossover probability changes to 0.49 from 0.01, comparison algorithm
Mean iterative number of time.Test result indicate that, work as pcWhen=0.01 and 0.41, mean iterative number of time is smaller, respectively 6.65 Hes
6.45.But because the two difference is smaller, and the result based on forefathers' research, work as pc=0.01 is optimal, is considered, this hair
Bright middle selection pc=0.01 is more particularly suitable.
To check and evaluating the beneficial effect of algorithm of the present invention, it is right that the present embodiment is carried out using two kinds of variation methods
Than:Variation method one is the mutation operation in above-mentioned steps 3.2 in the present invention.The concrete thought of variation method two is:First by
Each individual fitness forms selection wheel disc, and Probability p is then pressed from populationmIndividuality is randomly selected, further according to every in population
The fitness selection variable position of one performs mutation operation.If t- diagnosable systems are unsatisfactory for after variation, according to individuality
In each fitness randomly choose one enter row variation.
We make the popsize increase to 50, mutation probability p from 2m=0.7, crossover probability pc=0.01, number of network nodes n are from 5
Increase to 120, done one group of emulation experiment, under observation Different Variation method, diagnosis algorithm mean iterative number of time is with Population Size
Change.By to interpretation, under Malek models, the mean iterative number of time of variation method of the present invention is 8.0400,
And the mean iterative number of time of variation method two is 94.4150, variation method is more suitable for Malek models, and this in illustrating the present invention
Iterations needed for algorithm can substantially reduce tracing trouble collection in invention, improves diagnosis efficiency.
Embodiment 2
To check and evaluating beneficial effect of the algorithm of the present invention in terms of system fault diagnosis, compare two kinds of diagnosis and calculate
The average CPU time of method.Diagnosis algorithm one is the system diagnosability algorithm under Malek models of the present invention;Diagnosis is calculated
Method two is comprised the following steps that:
Step 1:Fault-free node method is specified to produce initial population;
Step 2:The fitness of individuality in population is calculated, determines whether that fitness value is 1 individuality, if not having, performed
Step 3;
Step 3:Following genetic manipulation is carried out to population:
3.1 selection operations, with diagnosis algorithm one;
3.2 mutation operations, i.e., variation method two described in embodiment 1;
3.3 crossover operations, with diagnosis algorithm one;
3.4 judge that state for whether the number of network nodes t of failure meets t- diagnosable systems, that is, is sentenced in individuality after crossover operation
Whether disconnected state meets 2 × t+1≤n conditions for the number of network nodes t of failure, if meeting, performs step 4;If it is not satisfied, then
Re-execute above-mentioned steps 3.1~3.3;
Step 4:Each individual fitness in new population is calculated, judges whether there is the individuality that fitness is 1 in population, if
Have, then find the failure collection of system, diagnosis terminates;If no, continuing to allow population loop iteration to perform step 3, until looking for
Go out the individuality that fitness is 1.
In Population Size popsize=7, mutation probability pm=0.7, crossover probability pc=0.01, number of network nodes n increase from 10
It is added under 300 such one group of parameter, compares two kinds of average CPU times of algorithm.Test result indicate that, diagnosis algorithm in the present invention
The average CPU time is 0.6540 second, and the average CPU time of diagnosis algorithm two is 4.3203 seconds, it is seen that under Malek models originally
The efficiency of invention algorithm is apparently higher than diagnosis algorithm two.
Diagnostic method of the invention by genetic algorithm the characteristic such as parallel, ability of searching optimum is strong, improve positioning therefore
Hinder the efficiency of collection, in combination with Malek comparison models, also superior to traditional PMC in terms of the accuracy of target faults collection is judged
Model.The present invention is applied in system fault diagnosis problem, can more accurately and rapidly find out target faults collection.
Claims (5)
1. a kind of diagnosis method for system fault based on Malek models, it is characterised in that comprise the following steps:
Step 1:Fault-free node method generation initial population is specified under Malek models, each in the initial population for being generated
I.e. one binary string of individuality corresponds to multi-computer system, and each in individuality is the knot during bit corresponds to system
Point;
Step 2:According to the constraint equation of Malek modelling node states, suitable fitness function is designed, then calculate and plant
Whether individual fitness in group, judges in population containing fitness value to be 1 individuality according to fitness, if not having, performs
Step 3;If so, then direct output system failure collection, diagnosis terminates;
Step 3:Following genetic manipulation is carried out to population:
3.1 selection operations:Using roulette selection method, the cumulative select probability of each individuality, fitness individuality high are calculated
Preferentially it is selected as follow-on individuality;Meanwhile, excellent operation is deposited in introducing, is fitted with the individual replacement next generation of fitness highest
The corresponding individuality of response minimum value;
3.2 mutation operations:Made a variation using binary system, according to the fitness and mutation probability p of individual each node in populationmAt random
Certain node in selected population enters row variation, will be changed between 0 and 1;
3.3 crossover operations:According to fitness individual in population, p is randomly choosed from populationcPopsize is individual, wherein
pcCrossover probability is represented, popsize represents number individual included in population;This to be randomly choosed from remaining population a few again
The pairing of body is individual, takes single-point cross method, randomly chooses a crosspoint, and crossover operation is carried out to two father's strings, obtains
New heredity is individual;
3.4 judge that state for whether the number of network nodes t of failure meets t- diagnosable systems, that is, judges shape in individuality after crossover operation
Whether state meets 2 × t+1≤n conditions for the number of network nodes t of failure, if meeting, performs step 4;If it is not satisfied, then holding again
Row above-mentioned steps 3.1~3.3;
Step 4:Each individual fitness in new population is calculated, whether is judged in population containing the individuality that fitness is 1, if
Have, then direct output system failure collection, diagnosis terminates;If no, continuing to allow population loop iteration to perform step 3, until
Find out the individuality that fitness is 1.
2. the diagnosis method for system fault based on Malek models according to claim 1, it is characterised in that the step 1
Fault-free node method generation initial population is specified to comprise the following steps under middle use Malek models:
(1) in comprising the n multi-computer system of node, a state of node k is randomly assigned for fault-free;
(2) according to the degree of node k, the node j adjacent with node k is found, if S (k, j)=0, then it represents that node k's and node j
Test result is identical, and the state of node j is fault-free;If S (k, j)=1, then it represents that the test result of node k and node j not phase
Together, the state of node j is faulty, and wherein S (k, j) is represented and the situation of same test task is being distributed to node k and node j
Under, the comparative result between two nodes;
(3) each state for obtaining step (2) carries out step 2 for trouble-free node;
(4) node to unknown state is randomly assigned state;
(5) judge whether state meets t- diagnosable systems for the number of network nodes t of failure, if it is not satisfied, then return performing step
(1);If meeting, step (6) is performed;
(6) according to step (1)~(5) loop iteration popsize times, you can obtain the initial population that size is popsize.
3. the diagnosis method for system fault based on Malek models according to claim 1, it is characterised in that the step 2
Comprise the following steps:
State f (the u of node under definition Malek modelsk,uj) with compare side ekjMeet following constraint equation:
Design fitness function:Dd (v [k])=| { ekj∈ E and meet f (uk,uj) |, wherein, dd (v [k]) is represented and node k
Adjacent node and meet constraint equation f (uk,uj) node number and;
Calculate the fitness of node j in individuality i:Wherein dd (v [j]) represents the knot with node j adjoinings
Point j1And meet constraint equationNode number and, d (v [j]) represents the degree of node j;
Calculate each individual fitness in population:Wherein FT (v [i]) represents the adaptation of individuality i
Degree, includes n node in n expression systems.
4. the diagnosis method for system fault based on Malek models according to claim 1, it is characterised in that institute in step 3
The method of the selection operation stated is:
Fitness highest individuality in population is stored in Elite, wherein Elite represents adaptive value highest in every generation population
Individuality;
Calculate individual adaptation degree summation in population:Wherein FT (v [i]) represents individuality i's
Fitness;
Calculate probability of each individuality in wheel disc:Wherein piRepresent probability of the individuality i in wheel disc;
Calculate the cumulative select probability of each individuality:Wherein qiRepresent that individuality i's is cumulative
Select probability;
Fitness individual in population after selecting is calculated, and the corresponding individuality of minimum adaptive value is replaced with Elite.
5. the diagnosis method for system fault based on Malek models according to claim 1, it is characterised in that institute in step 3
The method of the mutation operation stated is:The fitness value of all each individual nodes in searching loop population, if fitness value fv
(i,j)<Rand (), then select for node j in individual i to enter row variation, wherein rand () represent one be used to generating 0~1 it
Between random number function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611037746.5A CN106776088A (en) | 2016-11-23 | 2016-11-23 | Diagnosis method for system fault based on Malek models |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611037746.5A CN106776088A (en) | 2016-11-23 | 2016-11-23 | Diagnosis method for system fault based on Malek models |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106776088A true CN106776088A (en) | 2017-05-31 |
Family
ID=58974410
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611037746.5A Pending CN106776088A (en) | 2016-11-23 | 2016-11-23 | Diagnosis method for system fault based on Malek models |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106776088A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112217668A (en) * | 2020-09-29 | 2021-01-12 | 福州大学 | Self-adaptive network fault diagnosis method based on comparison model |
CN116924287A (en) * | 2023-09-18 | 2023-10-24 | 临工重机股份有限公司 | Control method, device, equipment and medium of hydraulic compensation leveling mechanism |
CN117349623A (en) * | 2023-10-25 | 2024-01-05 | 广西财经学院 | System-level fault diagnosis method based on double-population Harris eagle algorithm |
-
2016
- 2016-11-23 CN CN201611037746.5A patent/CN106776088A/en active Pending
Non-Patent Citations (3)
Title |
---|
M. EL HADEF: "Efficient Comparison-Based Fault Diagnosis of Multiprocessor Systems Using an Evolutionary Approach", 《PROCEEDINGS 15TH INTERNATIONAL PARALLEL AND DISTRIBUTED PROCESSING SYMPOSIUM. IPDPS 2001》 * |
宣恒农,孟亚: "基于比较的非对称模型下的系统级故障诊断", 《计算机应用与软件》 * |
文学志: "Malek故障模型的方程诊断算法设计", 《计算机工程与应用》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112217668A (en) * | 2020-09-29 | 2021-01-12 | 福州大学 | Self-adaptive network fault diagnosis method based on comparison model |
CN112217668B (en) * | 2020-09-29 | 2021-10-29 | 福州大学 | Self-adaptive network fault diagnosis method based on comparison model |
CN116924287A (en) * | 2023-09-18 | 2023-10-24 | 临工重机股份有限公司 | Control method, device, equipment and medium of hydraulic compensation leveling mechanism |
CN116924287B (en) * | 2023-09-18 | 2023-12-08 | 临工重机股份有限公司 | Control method, device, equipment and medium of hydraulic compensation leveling mechanism |
CN117349623A (en) * | 2023-10-25 | 2024-01-05 | 广西财经学院 | System-level fault diagnosis method based on double-population Harris eagle algorithm |
CN117349623B (en) * | 2023-10-25 | 2024-04-19 | 广西财经学院 | System-level fault diagnosis method based on double-population Harris eagle algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bajpai et al. | Genetic algorithm–an approach to solve global optimization problems | |
Keijzer et al. | Evolving objects: A general purpose evolutionary computation library | |
Eklöf et al. | Secondary extinctions in food webs: a Bayesian network approach | |
Vergidis et al. | Business process improvement using multi-objective optimisation | |
Wessing et al. | Niching by multiobjectivization with neighbor information: Trade-offs and benefits | |
CN106776088A (en) | Diagnosis method for system fault based on Malek models | |
Georgoulakos et al. | Evolutionary multi-objective optimization of business process designs with pre-processing | |
CN104656620A (en) | Comprehensive evaluation system for remanufacturing of heavy-duty machine tool | |
Coelho et al. | Multi-objective design of hierarchical consensus functions for clustering ensembles via genetic programming | |
Galván et al. | Semantics in multi-objective genetic programming | |
Šikulová et al. | Coevolution in cartesian genetic programming | |
Michelakos et al. | A hybrid classification algorithm evaluated on medical data | |
Kim et al. | Structure learning of Bayesian networks by estimation of distribution algorithms with transpose mutation | |
Maashi | A Comprehensive Review of Software Testing Methodologies Based on Search-based Software Engineering | |
Kent et al. | BOP-Elites, a Bayesian Optimisation Approach to Quality Diversity Search with Black-Box descriptor functions | |
Cornford et al. | Optimizing Spacecraft Design-Optimization Engine Development: Progress and Plans | |
Barlow et al. | The impact of connection topology and agent size on cooperation in the iterated prisoner's dilemma | |
CN114819151A (en) | Biochemical path planning method based on improved agent-assisted shuffled frog leaping algorithm | |
CN109726362A (en) | Solve the local search method for solving and system for weighting maximum satisfiability problem | |
Kim et al. | Multileveled symbiotic evolutionary algorithm: Application to FMS loading problems | |
Serdyukov et al. | Using genetic algorithm for generating optimal data sets to automatic testing the program code | |
Eiben et al. | Evolutionary algorithms | |
CN113539517A (en) | Prediction method of time sequence intervention effect | |
Henry | Forming productive student groups using a massively parallel brute-force algorithm | |
Kaufmann et al. | Toward self-adaptive embedded systems: multi-objective hardware evolution |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170531 |