CN110363275A - Immune algorithm and data fusion for Damage Assessment Method - Google Patents
Immune algorithm and data fusion for Damage Assessment Method Download PDFInfo
- Publication number
- CN110363275A CN110363275A CN201910485654.0A CN201910485654A CN110363275A CN 110363275 A CN110363275 A CN 110363275A CN 201910485654 A CN201910485654 A CN 201910485654A CN 110363275 A CN110363275 A CN 110363275A
- Authority
- CN
- China
- Prior art keywords
- antibody
- population
- affinity
- damage
- follows
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/257—Belief theory, e.g. Dempster-Shafer
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
For the immune algorithm and data fusion of Damage Assessment Method, it is related to Damage Assessment Method technical field.Immune algorithm for Damage Assessment Method, which is characterized in that it comprises the steps of: that random generation scale is the antibody initial population of N, and the various parameters in algorithm are arranged；Antibody is divided into each sub- population of m according to the equiblibrium mass distribution rule of proposition by the affinity for calculating all antibody, and each sub- population from selected at random in data base τ non-self antibody realize populations between interaction.After by adopting the above technical scheme, the invention has the following beneficial effects: reducing the population scale of single calculation, greatly reduce algorithm search space, the influence of single-sensor measurement error, sensor to factors such as damage sensitivity deficiencies to recognition result itself is reduced to a certain extent, improves identification accuracy.
Description
Technical field
The present invention relates to Damage Assessment Method technical fields, and in particular to the immune algorithm sum number for Damage Assessment Method
According to fusion.
Background technique
Infrastructure always is the object of focusing the consruction in China as the backing and necessary condition of socio-economic development, and
The large scale structures such as bridge naturally become the project of giving special assistance to as the important component in traffic infrastructure.
As the service life of Structural Engineering increases, inevitably will appear different degrees of damage, intelligent algorithm by
In having the characteristics that search capability is strong, degree of well adapting to is widely used, lot of domestic and international scholar is to it in structural damage
The application in field is studied.Chen Yuzhou etc. detects the structural unit damaged first with wavelet analysis, uses later
Immune algorithm carries out the identification of degree of injury to the structural unit damaged.Guo.H.Y etc. carries out multiple sensing datas
Bayesian Fusion processing after obtaining significantly more efficient damage factor, using immune algorithm identification of damage information, improves identification knot
The reliability of fruit.Guo.T etc. is searched in order to reduce noise jamming when damage check, using multiscale space theory from multi-angle
Damage characteristic, and by its fusion treatment, obtain accurate damage information.Liu Jian etc. is examined using the immune algorithm based on favorable selection
The exception for surveying Wavelet Energy Spectrum, improves the speed of non-destructive tests.
At this stage when identifying the damage problem of large scale structure, mainly there is following problem to need to solve: the biography of large scale structure
There are many sensor number, and the data volume of processing is huge, cause to identify that the faulted condition speed of structure is slower；Single-sensor is due to surveying
Measuring data, there are the accuracy of identification of error, sensor itself is insufficient, it is difficult to make accurate judgement to On Damage State.
Summary of the invention
In view of the defects and deficiencies of the prior art, the present invention intends to provide the immune calculations for Damage Assessment Method
Method and data fusion reduce the population scale of single calculation, greatly reduce algorithm search space, improve the anti-of data base
Body performance reduces algorithm in the blindness of search process, which is quickly found when handling a large amount of sensing datas
Optimal solution reduces search time, to improve the speed of identification；Identification using D-S evidence theory to multisensor parameter
As a result fusion treatment is carried out, reduces single-sensor measurement error, sensor itself to a certain extent to damage sensitivity not
The influences of the factors to recognition result such as foot, improve identification accuracy.
To achieve the above object, the present invention is using following technical scheme:
For the immune algorithm of Damage Assessment Method, it is comprised the steps of:
Step 1: generating the antibody initial population that scale is N at random, the various parameters in algorithm are set；
Step 2: calculating the affinity of all antibody, antibody is divided into each son kind of m according to the equiblibrium mass distribution rule of proposition
Group, and each sub- population from selected at random in data base the non-self antibody of τ realize populations between interaction；
Step 3: preceding δ % outstanding antibody for selecting every sub- population are cloned, the quantity and antibody affinity of clone
Directly proportional, i.e., more outstanding antibody accounting is bigger；
Step 4: carrying out affine mutation to clonal antibody, the probability of mutation is as follows:
In formula, Fmax, Fmin are the minimum and maximum affine angle value of clonal antibody；
Step 5: calculating the clonal antibody similarity after variation, the high antibody of similarity is rejected, guarantees the more of clonal plant population
Sample, similarity calculation are as follows:
Step 6: the affinity of clonal antibody is recalculated, by the classic antibody of affinity compared with parent antibody, if
It is more outstanding than parent, then parent antibody is replaced with outstanding antibody；
Step 7: selecting outstanding antibody in every sub- population is added data base, affinity threshold values A is utilized latertvAnd concentration
Threshold values CtvData base antibody is updated, affinity difference and highly concentrated antibody are rejected, guarantees data base with the overall superior of population
Property constantly changes, and antibody concentration is obtained by antibody similarity:
In formula, n is the antibody levels in current population；
Step 8: judging whether to reach defined maximum number of iterations, stop search, is exported in data base if reaching
Continue as a result, being transferred to step 2 if not up to.
The step 1 specifically: assuming that scale is the antibody population X={ x1, x2, x3 ..., xN } of N, antibody and antigen
Between affinity indicated with A (xi), each antibody calculate with the affinity of antigen, the affinity of population is obtained
All antibody is arranged successively, then using affinity size as standard by A={ A1, A2, A3 ..., AN } according to affinity size
Population is divided into n class, successively an antibody is taken out at random from every one kind later and is added to a Small Population, by n × m times
After extract operation, population is divided into the Small Population of m affinity equilibrium.
The step 2 specifically: each Small Population must be intersected intervention to it using other populations when evolving and constantly be adjusted
Itself the whole direction of search deviates general direction to avoid the direction of search of oneself, due to summarizing each sub- population in data base
Outstanding antibody, therefore use data base herein as intermediate vector to realize the interaction between population, it is assumed that antibody in data base
Scale be ι, for each Small Population, reject belong to the antibody of population first, it is random from remaining antibody later
τ antibody of selection is replicated, and the antibody of the smallest τ antibody duplication of affinity in this population is replaced, to all small
Population all implements this operation, that is, completes a population interaction.
Affinity threshold values A in the step 7tvCalculating are as follows:
Atv=kA×Amax
Antibody affinity all in population are arranged from big to small first, wherein maximum affinity is denoted as Amax, take a ratio
Number of cases kA；
Concentration threshold values CtvCalculating are as follows:
Ctv=kC×Cmin
Antibody by affinity in population in this section (Atv, Amax) is elected, later by antibody by concentration index from
Minispread is arrived greatly, wherein Cmin is denoted as Cmin, takes a proportional numbers kc, the antibody in this section (Cmin, Ctv) is retained,
Other are removed, and form data base.
For the data fusion of Damage Assessment Method, it is comprised the steps of:
Using D-S evidence theory to the non-destructive tests result of acceleration and displacement parameter, the non-destructive tests knot of stress parameters
Fruit carries out data fusion, determines with the following method to the basic probability assignment function m (A) in D-S evidence theory, uses vectorThe degree of injury for indicating position j is α %, it is assumed that the recognition result of damage criterion i is ri, and damage refers to
The result for marking i identification position j is rij, then the position j confidence level of index i identification is expressed as follows:
Determine index i to the mi (A) of faulted condition A using following formula:
The rule of combination of D-S evidence theory is as follows:
K is known as the inconsistent factor in formula, calculates as follows:
The non-destructive tests result of the acceleration and displacement parameter: the eigenfrequncies and vibration models of structure are combined
To obtain accurate structural damage information, the eigenfrequncies and vibration models of structure can be obtained by acceleration and displacement sensor,
It is assumed that α is stiffness injury's coefficient, it will be between the eigenfrequncies and vibration models after the eigenfrequncies and vibration models that calculated by α and actual measurement damage
Difference as objective function, then constantly adjustment α make objective function be intended to 0, finally can be obtained damage after each α
Value, so that it is determined that damage position and degree.Objective function is expressed as follows:
In formula, C ω, C φ are weight coefficient；ω c i and ω t i is the calculated value and reality of intrinsic frequency after the damage of the i-th rank
Measured value；φ c ij, φ t ij are the calculated value and measured value of the vibration shape after i-th rank of j-th of node is damaged.
The non-destructive tests result of the stress parameters: the strain mode change of each node after damage is obtained by strain gauge
Change amount can also identify the damage position and degree of structure, acquire node using central difference method to node vibration shape φ (i) first
Camber mode ρ (i), be shown below:
In formula, △ is the distance of adjacent node,
Strain mode amount is calculated by node curvature and structure height later, the node vibration shape can be by stiffness injury's factor alpha
It is calculated, so the strain mode difference of structure can also be expressed as the objective function of α, it is as follows:
In formula, ε c ij and ε t ij is respectively the strain mode value for calculating and surveying after i-th rank of j-th of node is damaged.
After adopting the above technical scheme, greatly subtracting the invention has the following beneficial effects: reduce the population scale of single calculation
Small algorithm search space improves the antibody performance of data base, reduces algorithm in the blindness of search process, so that the algorithm exists
Optimal solution can be quickly found when handling a large amount of sensing datas, reduces search time, to improve the speed of identification；Utilize D-
S evidence theory carries out fusion treatment to the recognition result of multisensor parameter, reduces single-sensor measurement to a certain extent
The influence of error, sensor to factors such as damage sensitivity deficiencies to recognition result itself, improves identification accuracy.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention without any creative labor, may be used also for those of ordinary skill in the art
To obtain other drawings based on these drawings.
Fig. 1 is the schematic process flow diagram of immune algorithm in the present invention；
Fig. 2 is the structural schematic diagram of the main bridge model of cable-stayed bridge in the present invention；
Fig. 3 is the preceding 10 rank intrinsic frequency figure after damaging in the present invention；
Fig. 4 is the relevant parameter figure of optimization for ant algorism in the present invention；
Fig. 5 is the mean value functions variation diagram of two kinds of algorithm search objective functions 1 in the present invention；
Fig. 6 is the mean value functions variation diagram of two kinds of algorithm search objective functions 2 in the present invention；
Fig. 7 is the ASSOCIATE STATISTICS value figure of two kinds of algorithm search in the present invention；
Fig. 8 is three kinds of recognition result figures compared with actual value in the present invention.
Specific embodiment
Referring to shown in Fig. 1-Fig. 8, present embodiment the technical solution adopted is that: for the immune of Damage Assessment Method
Algorithm, it is comprised the steps of:
Step 1: generating the antibody initial population that scale is N at random, the various parameters in algorithm are set；
Step 2: calculating the affinity of all antibody, antibody is divided into each son kind of m according to the equiblibrium mass distribution rule of proposition
Group, and each sub- population from selected at random in data base the non-self antibody of τ realize populations between interaction；
Step 3: preceding δ % outstanding antibody for selecting every sub- population are cloned, the quantity and antibody affinity of clone
Directly proportional, i.e., more outstanding antibody accounting is bigger；
Step 4: carrying out affine mutation to clonal antibody, the probability of mutation is as follows:
In formula, Fmax, Fmin are the minimum and maximum affine angle value of clonal antibody；
Step 5: calculating the clonal antibody similarity after variation, the high antibody of similarity is rejected, guarantees the more of clonal plant population
Sample, similarity calculation are as follows:
Step 6: the affinity of clonal antibody is recalculated, by the classic antibody of affinity compared with parent antibody, if
It is more outstanding than parent, then parent antibody is replaced with outstanding antibody；
Step 7: selecting outstanding antibody in every sub- population is added data base, affinity threshold values A is utilized latertvAnd concentration
Threshold values CtvData base antibody is updated, affinity difference and highly concentrated antibody are rejected, guarantees data base with the overall superior of population
Property constantly changes, and antibody concentration is obtained by antibody similarity:
In formula, n is the antibody levels in current population；
Step 8: judging whether to reach defined maximum number of iterations, stop search, is exported in data base if reaching
Continue as a result, being transferred to step 2 if not up to.
The step 1 specifically: assuming that scale is the antibody population X={ x1, x2, x3 ..., xN } of N, antibody and antigen
Between affinity indicated with A (xi), each antibody calculate with the affinity of antigen, the affinity of population is obtained
All antibody is arranged successively, then using affinity size as standard by A={ A1, A2, A3 ..., AN } according to affinity size
Population is divided into n class, successively an antibody is taken out at random from every one kind later and is added to a Small Population, by n × m times
After extract operation, population is divided into the Small Population of m affinity equilibrium.
The step 2 specifically: each Small Population must be intersected intervention to it using other populations when evolving and constantly be adjusted
Itself the whole direction of search deviates general direction to avoid the direction of search of oneself, due to summarizing each sub- population in data base
Outstanding antibody, therefore use data base herein as intermediate vector to realize the interaction between population, it is assumed that antibody in data base
Scale be ι, for each Small Population, reject belong to the antibody of population first, it is random from remaining antibody later
τ antibody of selection is replicated, and the antibody of the smallest τ antibody duplication of affinity in this population is replaced, to all small
Population all implements this operation, that is, completes a population interaction.
Affinity threshold values A in the step 7tvCalculating are as follows:
Atv=kA×Amax
Antibody affinity all in population are arranged from big to small first, wherein maximum affinity is denoted as Amax, take a ratio
Number of cases kA；
Concentration threshold values CtvCalculating are as follows:
Ctv=kC×Cmin
Antibody by affinity in population in this section (Atv, Amax) is elected, later by antibody by concentration index from
Minispread is arrived greatly, wherein Cmin is denoted as Cmin, takes a proportional numbers kc, the antibody in this section (Cmin, Ctv) is retained,
Other are removed, and form data base.Assuming that concentration of the antibody in population is indicated with C (xi), the parent between antibody and antigen before
It is indicated with degree with A (xi), has used a kind of stage threshold values division mode to obtain the mode of data base, so that in data base
Antibody is always made of that a part of antibody closest to optimal solution, and guarantees the otherness between antibody with concentration index, so that note
Recall library and real-time adjustment is made according to population antibody Optimality.
For the data fusion of Damage Assessment Method, it is comprised the steps of:
Using D-S evidence theory to the non-destructive tests result of acceleration and displacement parameter, the non-destructive tests knot of stress parameters
Fruit carries out data fusion, determines with the following method to the basic probability assignment function m (A) in D-S evidence theory, uses vectorThe degree of injury for indicating position j is α %, it is assumed that the recognition result of damage criterion i is ri, and damage refers to
The result for marking i identification position j is rij, then the position j confidence level of index i identification is expressed as follows:
Determine index i to the mi (A) of faulted condition A using following formula:
The rule of combination of D-S evidence theory is as follows:
K is known as the inconsistent factor in formula, calculates as follows:
The non-destructive tests result of the acceleration and displacement parameter: the eigenfrequncies and vibration models of structure are combined
To obtain accurate structural damage information, the eigenfrequncies and vibration models of structure can be obtained by acceleration and displacement sensor,
It is assumed that α is stiffness injury's coefficient, it will be between the eigenfrequncies and vibration models after the eigenfrequncies and vibration models that calculated by α and actual measurement damage
Difference as objective function, then constantly adjustment α make objective function be intended to 0, finally can be obtained damage after each α
Value, so that it is determined that damage position and degree.Objective function is expressed as follows:
In formula, C ω, C φ are weight coefficient；ω c i and ω t i is the calculated value and reality of intrinsic frequency after the damage of the i-th rank
Measured value；φ c ij, φ t ij are the calculated value and measured value of the vibration shape after i-th rank of j-th of node is damaged.The a certain position of structure
Damage meeting occurs so that the parameter changes such as rigidity of structure, so as to cause the whole intrinsic frequency of structure and the vibration shape of each node
It changes, intrinsic frequency reflects the overall permanence of structure, but the damage information precision for including is not high；Office of the vibration shape to structure
Portion's variation is more sensitive, includes more damage informations.
The non-destructive tests result of the stress parameters: the strain mode change of each node after damage is obtained by strain gauge
Change amount can also identify the damage position and degree of structure, acquire node using central difference method to node vibration shape φ (i) first
Camber mode ρ (i), be shown below:
In formula, △ is the distance of adjacent node,
Strain mode amount is calculated by node curvature and structure height later, the node vibration shape can be by stiffness injury's factor alpha
It is calculated, so the strain mode difference of structure can also be expressed as the objective function of α, it is as follows:
In formula, ε c ij and ε t ij is respectively the strain mode value for calculating and surveying after i-th rank of j-th of node is damaged.
Stiffness variation when structural damage can cause the stress changes of injury region simultaneously, so that the strain mode of each node also becomes
Change.
Using an extra-large bridge as research object, 2478 meters of bridge overall length, wherein main bridge is the double ropes of double tower of one or three straddle types
Face cable-stayed bridge, it is 1074 meters of girder overall length (282+510+282), 40 meters wide, 109 sections are divided into, using C50 concrete material, phase
Close parameter are as follows: V=0.2, E=325Gpa, ρ=2600kg/m3.Overall structure is divided into multiple regions and carries out non-destructive tests, this
Text analyzes this 20 sections degree of impairment of its 30# -50#, and the finite element model of structure is established using ANSYS software, will select
Fixed 20 sections are divided into 300 structural units, the rigidity of the 40th, 100,150,210,270 this five units is reduced by 30% respectively,
50%, 75%, 45%, 20%,.Model analysis is carried out to the structure after damage, is taken out the preceding intrinsic frequency of 10 rank of structure
Preceding 3 first order mode and strain mode of rate and each node are as the measured data value in objective function.
In order to verify the superiority and feasibility of modified immune algorithm, has chosen the traditional immunization based on Immune Clone Selection and calculate
Method and the modified immune algorithm of proposition respectively scan for two objective functions in MATLAB environment.In an experiment, two
The population scale of kind of algorithm is 100, and clone sizes take 60%, it is specified that maximum number of iterations is 400.Affinity is by objective function
Value determines that functional value is smaller, and the affinity of antibody is higher.In order to express easily, the function based on eigenfrequncies and vibration models is denoted as
Objective function 1, strain mode function are denoted as objective function 2.
Two kinds of algorithms can all reach convergence state, but traditional algorithm target in search process in defined the number of iterations
Functional value fluctuating range is larger, and the direction of search is easy to appear deviation, and there is also fuctuation within a narrow range when will finally restrain；Change
Into algorithm, fluctuating range is significantly smaller in search process, can restrain rapidly in final stage.This shows that innovatory algorithm is compared
Reduce in the blindness of traditional algorithm search, avoids many meaningless calculating.
Improved algorithm can reach convergence state compared to traditional algorithm faster, search for time-consuming less.This expression changes
Search efficiency into type algorithm is higher, and search speed faster, can quickly identify the faulted condition of structure.
Objective function is searched for using optimization for ant algorism, obtains the identification of two kinds of damage criterions known to the α value of each node
As a result, basis later
Two kinds of indexs are constructed to the basic probability assignment function value of each damage unit, two kinds of damage results are merged
Processing.In the recognition result of eigenfrequncies and vibration models damage criterion, it is consistent at preset five damages unit with actual value, accidentally
Difference is within 4%, but to hurt identification error larger to unit is closed on, and multiple do not damage is all identified at unit and had damage；It is answering
Become in the recognition result of modal damage index, although closing on identification error very little at unit, at default damage unit with
Actual value difference is larger.After being merged two kinds of damage criterion recognition results using D-S evidence theory, damage is only in default five knots
Larger in structure unit, the Error of damage of remaining position is within 5%, compared to two kinds single injury indexs, not only in injury region
Recognition result it is more accurate, and to the identification error of unit is closed on also within zone of reasonableness, this shows at D-S evidence fusion
Reason eliminates the recognition result error of single index to a certain extent, by respective mutual supplement with each other's advantages, so that result is with higher
Accuracy.
After adopting the above technical scheme, the invention has the following beneficial effects: immune algorithm improves the non-destructive tests speed of structure,
Antibody population is divided into multiple Small Population parallel searches, and enhances data base to the guided bone of search process, is searched with shortening algorithm
The rope time；Carry out the fusion of data, the redundancy between elimination to the recognition result of multiple sensors with D-S evidence theory later
Property, error, achieve the purpose that accurately identify On Damage State.
The above is only used to illustrate the technical scheme of the present invention and not to limit it, and those of ordinary skill in the art are to this hair
The other modifications or equivalent replacement that bright technical solution is made, as long as it does not depart from the spirit and scope of the technical scheme of the present invention,
It is intended to be within the scope of the claims of the invention.
Claims (7)
1. being used for the immune algorithm of Damage Assessment Method, which is characterized in that it is comprised the steps of:
Step 1: generating the antibody initial population that scale is N at random, the various parameters in algorithm are set；
Step 2: calculating the affinity of all antibody, antibody is divided into each sub- population of m according to the equiblibrium mass distribution rule of proposition, and
And each sub- population from selected at random in data base the non-self antibody of τ realize populations between interaction；
Step 3: preceding δ % outstanding antibody for selecting every sub- population are cloned, the quantity and antibody affinity of clone is at just
Than that is, more outstanding antibody accounting is bigger；
Step 4: carrying out affine mutation to clonal antibody, the probability of mutation is as follows:
In formula, Fmax, Fmin are the minimum and maximum affine angle value of clonal antibody；
Step 5: calculating the clonal antibody similarity after variation, the high antibody of similarity is rejected, guarantees the multiplicity of clonal plant population
Property, similarity calculation are as follows:
Step 6: the affinity of clonal antibody is recalculated, by the classic antibody of affinity compared with parent antibody, if comparing father
For outstanding, then parent antibody is replaced with outstanding antibody；
Step 7: selecting outstanding antibody in every sub- population is added data base, affinity threshold values A is utilized latertvWith concentration threshold values
CtvUpdate data base antibody, reject affinity difference and highly concentrated antibody, guarantee data base with population overall superior not
Disconnected variation, antibody concentration are obtained by antibody similarity:
In formula, n is the antibody levels in current population；
Step 8: judging whether to reach defined maximum number of iterations, stop search if reaching, exports the knot in data base
Fruit is transferred to step 2 if not up to and continues.
2. the immune algorithm according to claim 1 for Damage Assessment Method, it is characterised in that: the step 1 is specific
Are as follows: assuming that scale is the antibody population X={ x1, x2, x3 ..., xN } of N, affinity A (xi) table between antibody and antigen
Show, each antibody calculate with the affinity of antigen, the affinity A={ A1, A2, A3 ..., AN } of population is obtained,
All antibody is arranged successively according to affinity size, population is then divided into n class, Zhi Houyi using affinity size as standard
It is secondary to take out an antibody at random from every one kind and be added to a Small Population, after by n × m extract operation, population just by
It is divided into the Small Population of m affinity equilibrium.
3. the immune algorithm according to claim 1 for Damage Assessment Method, it is characterised in that: the step 2 is specific
Are as follows: each Small Population must intersect the direction of search intervened and constantly adjust itself when evolving using other populations to it, to keep away
Exempt from the direction of search deviation general direction of oneself makes herein due to summarizing the outstanding antibody of each sub- population in data base
Data base is used as intermediate vector to realize the interaction between population, it is assumed that the scale of antibody is ι in data base, for each microspecies
For group, the antibody for belonging to population is rejected first, is randomly choosed τ antibody from remaining antibody later and is replicated, it will
The antibody of the smallest τ antibody duplication of affinity replaces in this population, all implements this operation to all Small Populations, i.e., complete
At a population interaction.
4. the immune algorithm according to claim 1 for Damage Assessment Method, it is characterised in that: close in the step 7
With bottom valve value AtvCalculating are as follows:
Atv=kA×Amax
Antibody affinity all in population are arranged from big to small first, wherein maximum affinity is denoted as Amax, take a proportional numbers
kA；
Concentration threshold values CtvCalculating are as follows:
Ctv=kC×Cmin
Antibody by affinity in population in this section (Atv, Amax) is elected, later by antibody by concentration index from greatly to
Minispread, wherein Cmin is denoted as Cmin, takes a proportional numbers kc, and the antibody in this section (Cmin, Ctv) is retained, other
It removes, forms data base.
5. being used for the data fusion of Damage Assessment Method, which is characterized in that it is comprised the steps of:
Using D-S evidence theory to the non-destructive tests result of acceleration and displacement parameter, the non-destructive tests result of stress parameters into
Row data fusion determines the basic probability assignment function m (A) in D-S evidence theory with the following method, uses vectorThe degree of injury for indicating position j is α %, it is assumed that the recognition result of damage criterion i is ri, and damage refers to
The result for marking i identification position j is rij, then the position j confidence level of index i identification is expressed as follows:
Determine index i to the mi (A) of faulted condition A using following formula:
The rule of combination of D-S evidence theory is as follows:
K is known as the inconsistent factor in formula, calculates as follows:
6. the data fusion according to claim 5 for Damage Assessment Method, it is characterised in that: the acceleration and position
The non-destructive tests result of shifting parameter: the eigenfrequncies and vibration models of structure, which are combined, can obtain accurate structure damage
Hurt information, the eigenfrequncies and vibration models of structure can be obtained by acceleration and displacement sensor, it is assumed that α is stiffness injury's coefficient, will
By the difference between the eigenfrequncies and vibration models after the α eigenfrequncies and vibration models calculated and actual measurement damage as objective function, so
Afterwards constantly adjustment α make objective function be intended to 0, finally can be obtained damage after each α value, so that it is determined that damage position and
Degree.Objective function is expressed as follows:
In formula, C ω, C φ are weight coefficient；ω ci and ω ti are the calculated value and measured value of intrinsic frequency after the damage of the i-th rank；φ
Cij, φ tij are the calculated value and measured value of the vibration shape after i-th rank of j-th of node is damaged.
7. the data fusion according to claim 5 for Damage Assessment Method, it is characterised in that: the stress parameters
Non-destructive tests result: the strain mode variable quantity that each node after damaging is obtained by strain gauge can also identify the damage of structure
Hurt position and degree, acquire the camber mode ρ (i) of node using central difference method to node vibration shape φ (i) first, such as following formula institute
Show:
In formula, △ is the distance of adjacent node,
Strain mode amount is calculated by node curvature and structure height later, the node vibration shape can be calculated by stiffness injury's factor alpha
It obtains, so the strain mode difference of structure can also be expressed as the objective function of α, it is as follows:
In formula, ε cij and ε tij are respectively the strain mode value for calculating and surveying after i-th rank of j-th of node is damaged.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910485654.0A CN110363275A (en) | 2019-06-05 | 2019-06-05 | Immune algorithm and data fusion for Damage Assessment Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910485654.0A CN110363275A (en) | 2019-06-05 | 2019-06-05 | Immune algorithm and data fusion for Damage Assessment Method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110363275A true CN110363275A (en) | 2019-10-22 |
Family
ID=68215602
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910485654.0A Pending CN110363275A (en) | 2019-06-05 | 2019-06-05 | Immune algorithm and data fusion for Damage Assessment Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110363275A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056163A (en) * | 2016-06-08 | 2016-10-26 | 重庆邮电大学 | Multi-sensor information fusion object identification method |
| CN106126774A (en) * | 2016-06-14 | 2016-11-16 | 中山大学 | Structural Damage Identification based on BMO algorithm |
-
2019
- 2019-06-05 CN CN201910485654.0A patent/CN110363275A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106056163A (en) * | 2016-06-08 | 2016-10-26 | 重庆邮电大学 | Multi-sensor information fusion object identification method |
| CN106126774A (en) * | 2016-06-14 | 2016-11-16 | 中山大学 | Structural Damage Identification based on BMO algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 曹豪等: "免疫算法和数据融合在结构损伤识别中的应用", 《计算机应用与软件》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110728411B (en) | High-low altitude area combined rainfall prediction method based on convolutional neural network | |
| CN106055918A (en) | Power system load data identification and recovery method | |
| CN110210681B (en) | Prediction method of PM2.5 value of monitoring station based on distance | |
| CN109145464B (en) | Structural damage identification method integrating multi-target ant lion optimization and trace sparse regularization | |
| CN106056136A (en) | Data clustering method for rapidly determining clustering center | |
| CN113468968B (en) | Remote sensing image rotating target detection method based on non-anchor frame | |
| CN107909027A (en) | It is a kind of that there is the quick human body target detection method for blocking processing | |
| CN113313145B (en) | Expressway traffic event detection method based on mixed kernel correlation vector machine | |
| CN104361142A (en) | Detection method for rapid change in multi-source navigation electronic map vector road network | |
| CN105930900B (en) | The Forecasting Methodology and system of a kind of hybrid wind power generation | |
| CN102521527B (en) | Method for predicting space epitope of protein antigen according to antibody species classification | |
| CN110852243B (en) | Road intersection detection method and device based on improved YOLOv3 | |
| Xiaoyu et al. | Short-term traffic flow forecasting based on two-tier k-nearest neighbor algorithm | |
| CN102163300A (en) | Method for optimizing fault diagnosis rules based on ant colony optimization algorithm | |
| CN108416373A (en) | A kind of unbalanced data categorizing system based on regularization Fisher threshold value selection strategies | |
| CN110263945A (en) | Based on markovian Line Loss of Distribution Network System processing system | |
| CN109272491A (en) | The recognition methods of crack tip, device and equipment under experimental enviroment | |
| CN108804635A (en) | A kind of method for measuring similarity based on Attributions selection | |
| CN109446614A (en) | A kind of discrimination method and device of static reactive model parameter | |
| CN102708246A (en) | Multi-vortex ring microburst model parameter selection method based on nested differential evolution algorithm | |
| CN110363275A (en) | Immune algorithm and data fusion for Damage Assessment Method | |
| CN110378433A (en) | The classifying identification method of bridge cable surface defect based on PSO-SVM | |
| Fang et al. | An improved decision tree algorithm based on mutual information | |
| CN114821322A (en) | Small sample remote sensing image classification method and system based on attention mechanism | |
| Boniewicz et al. | Review of selected algorithms in the method energy evening algorithm in wireless sensor network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191022 |