CN108341342B - The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA - Google Patents
The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA Download PDFInfo
- Publication number
- CN108341342B CN108341342B CN201810062458.8A CN201810062458A CN108341342B CN 108341342 B CN108341342 B CN 108341342B CN 201810062458 A CN201810062458 A CN 201810062458A CN 108341342 B CN108341342 B CN 108341342B
- Authority
- CN
- China
- Prior art keywords
- rna
- prominent ring
- vector machine
- supporting vector
- overhead crane
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000010276 construction Methods 0.000 title claims abstract description 14
- 238000002474 experimental method Methods 0.000 claims abstract description 5
- 108091028043 Nucleic acid sequence Proteins 0.000 claims description 49
- 230000035772 mutation Effects 0.000 claims description 15
- 238000012706 support-vector machine Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 9
- 230000007935 neutral effect Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000000295 complement effect Effects 0.000 claims description 4
- 238000013461 design Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 230000017105 transposition Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 238000005094 computer simulation Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 18
- 230000002068 genetic effect Effects 0.000 description 9
- 238000012360 testing method Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 229910002056 binary alloy Inorganic materials 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B66—HOISTING; LIFTING; HAULING
- B66C—CRANES; LOAD-ENGAGING ELEMENTS OR DEVICES FOR CRANES, CAPSTANS, WINCHES, OR TACKLES
- B66C17/00—Overhead travelling cranes comprising one or more substantially horizontal girders the ends of which are directly supported by wheels or rollers running on tracks carried by spaced supports
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Mechanical Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses the overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA a kind of, belongs to intelligent modeling field.The following steps are included: 1) obtain the sampled data of bridge type crane system output and input by experiment or collection in worksite, and using the error sum of squares of desired output and reality output as the objective function of RNA-GA；2) by RNA molecule dash forward ring structure inspire, take out the overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA a kind of；3) operating parameter of set algorithm；4) it runs prominent ring intersection RNA-GA and obtains overhead crane position least square model supports vector machine and pivot angle least square model supports vector machine model parameter Estimation optimal value respectively, and substitute into optimal value in least square model supports vector machine model, form overhead crane model.The model construction of SVM method has many advantages, such as that fast convergence rate, precision are high, is also applied for other complex system modelings.
Description
Technical field
The present invention relates to intelligent optimization modeling method more particularly to a kind of overhead crane branch of prominent ring crossover operation RNA-GA
Hold vector machine modeling method.
Background technique
Overhead crane is a kind of heavy construction haulage equipment, is widely used in many fields such as industrial enterprise, building, harbour
Cargo carry, very important status is occupied in the development of the national economy.Overwhelming majority crane operation is by professional people at present
Member's manipulation is completed, and the greatest problem that this control mode faces is the hunting of load problem of trolley movement and external interference initiation,
This can not only reduce the working efficiency of overhead crane, can also be because of safety accident caused by transporting cargo and people or Environment Collision.With
The high speed development of science and technology, overhead crane is towards automation, intelligent direction development, therefore there is an urgent need to set up
High-precision overhead crane model.Common traditional mechanisms modeling method is not only complicated, but also modeling accuracy is not high, it is difficult to adapt to
The control requirement of this kind of large scale industry robot of overhead crane.For this purpose, using genetic algorithm as the intelligent optimization modeling method of representative
It attracts attention.
Genetic algorithm is to be put forward by professor Holland for 1975, has global optimizing ability outstanding, applicability
Extensively, the every field such as Combinatorial Optimization, machine learning, signal processing are widely used in.But basic genetic algorithm utilizes binary system
It is encoded, there are problems that hamming steep cliff, be easy to fall into local optimum in searching process, phenomena such as precocious occur.
By the inspiration that RNA molecule operates, RNA molecule operation and coding mode are introduced into genetic algorithm by Tao Jili et al.,
Improve performance of genetic algorithms.Not only low optimization accuracy is high for RNA genetic algorithm (RNA-GA), while also maintaining the more of population
Sample.
The present invention is inspired by RNA Double-chain two stage structure, it is proposed that a kind of prominent ring crossover operation RNA genetic algorithm (prominent ring friendship
Fork operation RNA-GA), the position least square method supporting vector machine model and pivot angle least square for being used for overhead crane are supported
The parameter optimization of vector machine model obtains high-precision overhead crane supporting vector machine model.
Summary of the invention
The deficiency that the present invention is complicated for traditional overhead crane modelling by mechanism method, precision is not high, it is proposed that one kind is prominent
The overhead crane model construction of SVM method of ring crossover operation RNA-GA is used for prominent ring crossover operation RNA-GA to obtain minimum
Two multiply support vector machines (LSSVM) model parameter estimation value, establish the position least square method supporting vector machine of overhead crane
(LSSVM) model and pivot angle least square method supporting vector machine (LSSVM) model, it is real that this method model built can be good at reflection
The nonlinear characteristic of border system.
The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA, comprising the following steps:
Step 1: overhead crane horizontal control force f is obtained by actual measurement or experimentxInput data；And obtain water
Prosposition sets x and pivot angle θxExport sampled data；
Step 2: establishing the position least square method supporting vector machine model and pivot angle least square supporting vector of overhead crane
Machine model；The wherein total n input of position least square method supporting vector machine model and an output Pout, it is as follows to input representation:
[x(t-1),x(t-2),...x(t-n1),fx(t-1),fx(t-2),...,fx(t-n2)]
Wherein t indicates sampling instant, n=n1+n2；n1For the number of horizontal location data input；n2For horizontal control force number
According to the number of input；The horizontal position sampled data at x (t-1) expression (t-1) moment；fx(t-1) level at (t-1) moment is indicated
Control force sampled data；
Pivot angle least square method supporting vector machine model possesses m input and an output Rout, it is as follows to input representation:
[θx(t-1),θx(t-2),...,θx(t-m1),fx(t-1),fx(t-2),...,fx(t-m2)]
Wherein m=m1+m2；m1For the number of pivot angle data input；m2For the number of horizontal control force data input；θx(t-
1) the pivot angle sampled data at (t-1) moment is indicated；
Step 3: being supported using Radial basis kernel function as position least square method supporting vector machine model and pivot angle least square
The kernel function of vector machine model；
Step 4: sample data is normalized；
Step 5: the parameter to be estimated of predeterminated position supporting vector machine model and pivot angle supporting vector machine model；
Step 6: being based on RNA Double-chain two stage structure, design prominent ring and intersect RNA-GA；
Step 7: the sample data obtained in step 1 being input to position supporting vector machine model, obtains desired output
PEout, by the reality output P in step 2outWith PEoutError sum of squares intersect the objective function of RNA-GA as prominent ring, pass through
RNA-GA optimizing obtains position least square method supporting vector machine and pivot angle least square method supporting vector machine parameter to be estimated respectively
Optimal value；
Step 8: the parameter optimal value obtained after step 7 is executed substitutes into position least square method supporting vector machine model respectively
With pivot angle least square method supporting vector machine model, for overhead crane horizontal position and pivot angle estimate.
In above scheme, each step can specifically in the following way:
Radial basis kernel function form in the step 3 is as follows:
Wherein δ is the width parameter of gaussian kernel function；| | | | it is norm；X ', y ' are respectively one group of feature.
It is normalized to zero-mean and unit variance in the step 4, normalizes formula are as follows:
Wherein xk,jIt is the data after normalization,Indicate j-th of component of k-th of sample, ujIt is the jth of all samples
The mean value of a component, δjFor the standard deviation of j-th of component of all samples.
Dashing forward in the step 6, steps are as follows for the prominent ring crossover operation that ring crossover operation RNA-GA is used:
Two RNA sequences are randomly choosed, the head end of two RNA sequences is aligned to form duplex structure, respectively to two RNA
Sequence executes traversing operation, checks whether it can form prominent ring structure；It is random to generate first to wherein any one RNA sequence
One position Pos1, then traverse [the 1, Pos of the RNA sequence1- 2] and [Pos1+ 2, L] position searching position Pos2, wherein L be
RNA sequence length, if Pos1And Pos2The base of two positions is capable of forming A-T G-C form complementary base pair, then should
RNA sequence can form prominent ring structure, and no person regenerates Pos1Position can form the Pos of prominent ring structure until finding1、Pos2,
Or the whole RNA sequence of traversal also fails to find the Pos that can form prominent ring structure1、Pos2Until；Then to another RNA sequence
Also traversing operation is executed；After having executed traversing operation to two RNA sequences, if there is prominent ring structure in two RNA sequences,
It is respectively formed prominent ring structure and transposition in two RNA sequences, the tail portion of a wherein longer RNA sequence is exceeded into former sequence
The part of column length L moves to the head end of another chain；If only having 1 in two RNA sequences has prominent ring structure, then at this
Prominent ring structure is formed on RNA sequence and prominent ring structure is dissociated into (prominent ring structure is detached from from former sequence and is bound in another chain)
The tail portion of a wherein longer RNA sequence is exceeded former sequence by the position corresponding with prominent ring structure root base into another chain
The part of column length L moves to the head end of another chain；If prominent ring structure is not present in two RNA sequences, then retain two chains extremely
It is next-generation.
Ring of dashing forward in the step 6, which intersects RNA-GA step, can be described as follows:
Step 6.1: the operating parameter of RNA-GA is set, including population scale N, maximum evolutionary generation G, individual lengths L, prominent
Ring crossover probability pb, replacement operator Probability piAnd termination rules, wherein termination rules are to reach the low optimization accuracy of setting or reach
To the maximum evolutionary generation of algorithm；
Step 6.2: the parameter of position or pivot angle least square method supporting vector machine model being encoded, N is randomly generated
The initial population Pop of a RNA sequence, each RNA sequence represent the possibility solution of support vector machines unknown parameter, wherein each
Parameter is encoded to the RNA subsequence that a length is L, each supporting vector of overhead crane by character set { 0,1,2,3 }
Machine model is made of two location parameters, i.e. kernel parameter and penalty factor；The code length of each RNA sequence is 2 × L；
Step 6.3: sample data is inputted to the square-error of the output and reality output that obtain after supporting vector machine model
With the objective function for intersecting RNA-GA as prominent ring；
Step 6.4: it uses ratio selection strategy, by population dividing is two parts according to fitness value size, i.e., " neutral
Body "" harmful individual "Two parts；
Step 6.5: to neutral populationA) and b) two kinds of crossover operations are executed respectively, and steps are as follows:
A) fromIt randomly selects two sequences and forms duplex structure, with Probability pbThe prominent ring is executed to duplex structure to hand over
Fork operation, generates the new population of N/2
B) same rightIndividual is with Probability piReplacement operator is executed, the new population of N/2 is generated
Population after crossover operation executes is represented byShared 3N/2 individual；
Step 6.6: to populationEach of individual respectively with self-adaptive mutation pmh、pmlTo its high position [L/2+
1, L] and low level [1, L/2] executes TSP question operation, pmh、pmlCalculation formula it is as follows:
Wherein a1For pmh、pmlInitial mutation probability, b1It is the variation range of mutation probability, g is current evolutionary generation, g0
It is turning point, aa is change rate；
Mutation operation is using the conversion of RNA sequence, transversion and exchange operation；
Step 6.7: if the optimal value precision of current population meets, termination is required or evolutionary generation satisfaction is terminated and required, then
The optimizing parameter of position supporting vector machine model and pivot angle support vector machines is obtained, otherwise return step 6.3.
The present invention devises the prominent ring Crossover Operator of RNA, proposes prominent ring crossover operation RNA-GA crossover operation, will dash forward
Ring crossover operation RNA-GA for overhead crane supporting vector machine model parameter Estimation in, the result shows that the support established to
Amount machine model can be good at the nonlinear characteristic for reflecting real system.
Detailed description of the invention
Fig. 1 is prominent ring crossover operation RNA-GA overhead crane model construction of SVM method flow diagram；
Fig. 2 is bridge type crane system schematic diagram；
Fig. 3 is prominent ring crossover operation schematic diagram；
Fig. 4 is least square method supporting vector machine model；
Fig. 5 is that the prominent ring in position intersects the output of behaviour's RNA-GA LSSVM model and overhead crane physical location exports and compares；
Fig. 6 is that the prominent ring in position intersects behaviour's RNA-GA LSSVM model output error figure；
Fig. 7 is that the prominent ring of pivot angle intersects the output of behaviour's RNA-GA LSSVM model and the practical pivot angle of overhead crane exports and compares；
Fig. 8 is that the prominent ring of pivot angle intersects behaviour's RNA-GA LSSVM model output error figure；
Specific embodiment
The overhead crane modeling of prominent ring crossover operation RNA-GA is described in detail and is discussed below in conjunction with attached drawing.
The step of prominent ring crossover operation RNA-GA overhead crane genetic algorithm, is as follows:
Step 1: overhead crane horizontal control force f is obtained by actual measurement or experimentxInput data；And obtain water
Prosposition sets x and pivot angle θxExport sampled data；
Step 2: establishing the position least square method supporting vector machine model and pivot angle least square supporting vector of overhead crane
Machine model；The wherein total n input of position least square method supporting vector machine model and an output Pout, it is as follows to input representation:
[x(t-1),x(t-2),...x(t-n1),fx(t-1),fx(t-2),...,fx(t-n2)]
Wherein t indicates sampling instant, n=n1+n2；n1For the number of horizontal location data input；n2For horizontal control force number
According to the number of input；The horizontal position sampled data at x (t-1) expression (t-1) moment；fx(t-1) level at (t-1) moment is indicated
Control force sampled data；
Pivot angle least square method supporting vector machine model possesses m input and an output Rout, it is as follows to input representation:
[θx(t-1),θx(t-2),...,θx(t-m1),fx(t-1),fx(t-2),...,fx(t-m2)]
Wherein m=m1+m2；m1For the number of pivot angle data input；m2For the number of control force data input；θx(t-1) table
Show the pivot angle sampled data at (t-1) moment；
Step 3: being supported using Radial basis kernel function as position least square method supporting vector machine model and pivot angle least square
The kernel function of vector machine model；
Radial basis kernel function form is as follows:
Wherein δ is the width parameter of gaussian kernel function；| | | | it is norm；X ', y ' are respectively one group of feature.
Step 4: sample data is normalized；
It is normalized to zero-mean and unit variance in the present invention, normalizes formula are as follows:
Wherein xk,jIt is the data after normalization,Indicate j-th of component of k-th of sample, ujIt is j-th of all samples
The mean value of component, δjFor the standard deviation of j-th of component of all samples.
Step 5: the parameter to be estimated of predeterminated position supporting vector machine model and pivot angle supporting vector machine model；
Step 6: being based on RNA Double-chain two stage structure, design prominent ring and intersect RNA-GA；
Steps are as follows for the prominent ring crossover operation that prominent ring crossover operation RNA-GA is used:
Two RNA sequences are randomly choosed, the head end of two RNA sequences is aligned to form duplex structure, respectively to two RNA
Sequence executes traversing operation, checks whether it can form prominent ring structure；It is random to generate first to wherein any one RNA sequence
One position Pos1, then traverse [the 1, Pos of the RNA sequence1- 2] and [Pos1+ 2, L] position searching position Pos2, wherein L be
RNA sequence length, if Pos1And Pos2The base of two positions is capable of forming A-T G-C form complementary base pair, then should
RNA sequence can form prominent ring structure, and no person regenerates Pos1Position can form the Pos of prominent ring structure until finding1、Pos2,
Or the whole RNA sequence of traversal also fails to find the Pos that can form prominent ring structure1、Pos2Until；Then to another RNA sequence
Also traversing operation is executed；After having executed traversing operation to two RNA sequences, if there is prominent ring structure in two RNA sequences,
It is respectively formed prominent ring structure and transposition in two RNA sequences, the tail portion of a wherein longer RNA sequence is exceeded into former sequence
The part of column length L moves to the head end of another chain；If only having 1 in two RNA sequences has prominent ring structure, then at this
Prominent ring structure is formed on RNA sequence and prominent ring structure is dissociated into (prominent ring structure is detached from from former sequence and is bound in another chain)
The tail portion of a wherein longer RNA sequence is exceeded former sequence by the position corresponding with prominent ring structure root base into another chain
The part of column length L moves to the head end of another chain；If prominent ring structure is not present in two RNA sequences, then retain two chains extremely
It is next-generation.
Prominent ring, which intersects RNA-GA step, can be described as follows:
Step 6.1: the operating parameter of RNA-GA is set, including population scale N, maximum evolutionary generation G, individual lengths L, prominent
Ring crossover probability pb, replacement operator Probability piAnd termination rules, wherein termination rules are to reach the low optimization accuracy of setting or reach
To the maximum evolutionary generation of algorithm；
Step 6.2: the parameter of position or pivot angle least square method supporting vector machine model being encoded, N is randomly generated
The initial population Pop of a RNA sequence, each RNA sequence represent the possibility solution of support vector machines unknown parameter, wherein each
Parameter is encoded to the RNA subsequence that a length is L, each supporting vector of overhead crane by character set { 0,1,2,3 }
Machine model is made of two location parameters, i.e. kernel parameter and penalty factor；The code length of each RNA sequence is 2 × L；
Step 6.3: sample data is inputted to the square-error of the output and reality output that obtain after supporting vector machine model
With the objective function for intersecting RNA-GA as prominent ring；
Step 6.4: it uses ratio selection strategy, by population dividing is two parts according to fitness value size, i.e., " neutral
Body "" harmful individual "Two parts；
Step 6.5: to neutral populationA) and b) two kinds of crossover operations are executed respectively, and steps are as follows:
A) fromIt randomly selects two sequences and forms duplex structure, with Probability pbThe prominent ring is executed to duplex structure to hand over
Fork operation, generates the new population of N/2
B) same rightIndividual is with Probability piReplacement operator is executed, the new population of N/2 is generated
Population after crossover operation executes is represented byShared 3N/2 individual；
Step 6.6: to populationEach of individual respectively with self-adaptive mutation pmh、pmlTo its high position [L/2+
1, L] and low level [1, L/2] executes TSP question operation, pmh、pmlCalculation formula it is as follows:
Wherein a1For pmh、pmlInitial mutation probability, b1It is the variation range of mutation probability, g is current evolutionary generation, g0
It is turning point, aa is change rate；
Mutation operation is using the conversion of RNA sequence, transversion and exchange operation；
Step 6.7: if the optimal value precision of current population meets, termination is required or evolutionary generation satisfaction is terminated and required, then
The optimizing parameter of position supporting vector machine model and pivot angle support vector machines is obtained, otherwise return step 6.3.
Step 7: the sample data obtained in step 1 being input to position supporting vector machine model, obtains desired output
PEout, by the reality output P in step 2outWith PEoutError sum of squares intersect the objective function of RNA-GA as prominent ring, pass through
RNA-GA optimizing obtains position least square method supporting vector machine and pivot angle least square method supporting vector machine parameter to be estimated respectively
Optimal value；
Step 8: the parameter optimal value obtained after step 7 is executed substitutes into position least square method supporting vector machine model respectively
With pivot angle least square method supporting vector machine model, for overhead crane horizontal position and pivot angle estimate.
Below based on this method, using " the three-dimensional overhead crane experiment porch " of certain National Key Laboratory, university as implementation
Example, is further explained the present invention.
Embodiment
In the present embodiment, which can be reduced to a two-dimentional bridge type crane system, see Fig. 1 crane system illustraton of model, dash forward
The step of ring crossover operation RNA-GA overhead crane genetic algorithm, is as follows:
Step 1: two-dimentional bridge type crane system horizontal direction control input f is obtained by platformx, position in horizontal direction
X and pivot angle θxExport sampled data.The parameter of overhead crane is set as trolley mass M=24.6kg, load quality m=5.4kg,
Lifting rope length fixes l=0.7m, gravity acceleration g=9.8m/s2.In carrying out experimental data collection process, open loop shape is kept
State, sampling period 1ms acquire 5 groups of data altogether, and every group has 11000 data, each random pumping 1100 in 5 groups of data
Data, wherein 550 experimental datas are as training sample, remaining is test sample.
Step 2: establishing the position least square method supporting vector machine model and pivot angle least square supporting vector of overhead crane
Machine model；Wherein totally 10 inputs and an output, input representation are as follows for position least square method supporting vector machine model:
[x(t-1),x(t-2),x(t-3),x(t-4),x(t-5),fx(t-1),fx(t-2),fx(t-3),fx(t-4),fx
(t-5)]
Wherein t indicates sampling instant；Similarly, pivot angle least square method supporting vector machine model possesses 8 inputs and one defeated
Out, input representation is as follows:
[θx(t-1),θx(t-2),θx(t-3),θx(t-4),fx(t-1),fx(t-2),fx(t-3),fx(t-4)]
Step 3: selecting core for position least square method supporting vector machine model and pivot angle least square method supporting vector machine model
Function selects Radial basis kernel function here, and kernel function can be expressed as follows:
Wherein δ is the width parameter of gaussian kernel function；| | | | it is norm；X ', y ' are respectively one group of feature.
Step 4: sample data being normalized, zero-mean and unit variance are normalized to, normalizes expression formula
It is as follows:
Wherein xk,jIt is the data after normalization,Indicate j-th of component of k-th of sample, ujIt is the jth of all samples
The mean value of a component, δjFor the standard deviation of j-th of component of all samples.
Step 5: the parameter that setting position supporting vector machine model and pivot angle supporting vector machine model are estimated；
Step 6: being inspired by RNA Double-chain two stage structure, design prominent ring and intersect RNA-GA；Prominent ring crossover operation step can be made such as
Lower description:
Prominent ring crossover operation: two RNA sequences of random selection are aligned the head end of two RNA sequences to form duplex structure,
Traversing operation is executed to two chains respectively, checks whether it can form prominent ring structure；By taking a wherein chain as an example, one is generated at random
A position Pos1, then traverse [the 1, Pos of the RNA sequence1- 2] and [Pos1+ 2, L] position searching Pos2, wherein L is RNA sequence
Length, if the base of two positions is capable of forming A-T G-C form complementary base pair, then the RNA sequence forms prominent ring knot
Structure, no person regenerate Pos1Position to form the Pos of prominent ring structure until finding1、Pos2, or the whole RNA sequence of traversal
Fail to find Pos1、Pos2Until；Then aforesaid operations are executed to another RNA sequence；If existing in two RNA sequences prominent
Ring structure is then respectively formed prominent ring structure and transposition, by the tail of a wherein longer RNA sequence in two RNA sequences
The part that portion exceeds former sequence length L moves to the head end of another chain；If only having 1 in two RNA sequences has prominent ring structure,
Prominent ring structure is then formed on this RNA sequence and prominent ring structure dissociates into another chain and prominent ring structure root base-pair
The part that the tail portion of a wherein longer RNA sequence exceeds former sequence length L is moved to the head end of another chain by the position answered；
If prominent ring structure is not present in two RNA sequences, then retain two chains to the next generation.
Step 7: the sample data obtained in step 1 is input to position supporting vector machine model, obtains model output, it will
The error sum of squares that reality output in step 2 is exported with model intersects the objective function of RNA-GA as prominent ring, passes through RNA-
The optimal value of GA optimizing acquisition position least square method supporting vector machine parameter to be estimated；Similarly, pivot angle is obtained according to above step
The optimal value of the parameter to be estimated of least square method supporting vector machine；
Step 8: the parameter optimal value obtained after step 7 is executed substitutes into position least square method supporting vector machine model respectively
With pivot angle least square method supporting vector machine model, and test sample is input in model, model is tested.
Wherein, above-mentioned prominent ring intersects RNA-GA step and can be described as follows:
Step 6.1: the operating parameter of RNA-GA is set, including population scale N=40, maximum evolutionary generation G=500, a
Body length L=20, prominent ring crossover probability pb=0.8, replacement operator Probability pi=1 and termination rules, wherein termination rules be up to
To low optimization accuracy Δ=0.0001 of setting, or reach the maximum evolutionary generation of algorithm.
Step 6.2: the parameter of position or pivot angle least square method supporting vector machine model being encoded, N is randomly generated
The initial population Pop of a RNA sequence, each RNA sequence represent the possibility solution of support vector machines unknown parameter, wherein each
Parameter is encoded to the RNA subsequence that a length is L, each supporting vector of overhead crane by character set { 0,1,2,3 }
Machine model is made of two location parameters, i.e. kernel parameter and penalty factor；Therefore, the code length of each RNA sequence be 2 ×
L。
Step 6.3: sample data is inputted to the square-error of the output and reality output that obtain after supporting vector machine model
With the objective function for intersecting RNA-GA as prominent ring.
Step 6.4: it uses ratio selection strategy, by population dividing is two parts according to fitness value size, i.e., " neutral
Body "" harmful individual "
Step 6.5: to neutral populationTwo kinds of crossover operations are executed respectively, and steps are as follows:
A) fromIt randomly selects two sequences and forms duplex structure, with Probability pbProminent ring is executed to duplex structure and intersects behaviour
Make, generates the new population of N/2
B) same rightIndividual is with Probability piReplacement operator is executed, the new population of N/2 is generated
Population after crossover operation executes is represented byShared 3N/2 individual；
Step 6.6: to populationEach of individual respectively with self-adaptive mutation pmh、pmlTo its high position [L/2+
1, L] and low level [1, L/2] executes TSP question operation, pmh、pmlCalculation formula it is as follows:
Wherein a1=0.05 is pmh、pmlInitial mutation probability, b1=0.2 is the variation range of mutation probability, and g is current
Evolutionary generation, g0=G/2 is turning point, and aa=20/G is change rate；
Variation is using the conversion of RNA sequence, transversion and exchange operation；
Step 6.7: if the optimal value precision of current population meets, termination is required or evolutionary generation satisfaction is terminated and required, then
The estimation parameter for obtaining position supporting vector machine model or pivot angle support vector machines, otherwise re-execute the steps 6.3.Test result
As shown in figures 5-8.
It can be seen that position LSSVM model from the test result of Fig. 5~8 and physical location coincide substantially, pivot angle LSSVM
Model output and practical pivot angle coincide substantially, and error is only fluctuated in very a small range.Therefore height may be implemented in the method for the present invention
Accuracy modeling.
Claims (5)
1. a kind of overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA, it is characterised in that including following step
It is rapid:
Step 1: overhead crane horizontal control force f is obtained by actual measurement or experimentxInput data；And obtain horizontal position
Set x and pivot angle θxExport sampled data；
Step 2: establishing the position least square method supporting vector machine model and pivot angle least square method supporting vector machine mould of overhead crane
Type；The wherein total n input of position least square method supporting vector machine model and an output Pout, it is as follows to input representation:
[x(t-1),x(t-2),...x(t-n1),fx(t-1),fx(t-2),...,fx(t-n2)]
Wherein t indicates sampling instant, n=n1+n2；n1For the number of horizontal location data input；n2It is defeated for horizontal control force data
The number entered；The horizontal position sampled data at x (t-1) expression (t-1) moment；fx(t-1) the horizontal control at (t-1) moment is indicated
Power sampled data；
Pivot angle least square method supporting vector machine model possesses m input and an output Rout, it is as follows to input representation:
[θx(t-1),θx(t-2),...,θx(t-m1),fx(t-1),fx(t-2),...,fx(t-m2)]
Wherein m=m1+m2；m1For the number of pivot angle data input；m2For the number of control force data input；θx(t-1) (t- is indicated
1) the pivot angle sampled data at moment；
Step 3: using Radial basis kernel function as position least square method supporting vector machine model and pivot angle least square supporting vector
The kernel function of machine model；
Step 4: sample data is normalized；
Step 5: the parameter to be estimated of predeterminated position supporting vector machine model and pivot angle supporting vector machine model；
Step 6: being based on RNA Double-chain two stage structure, design prominent ring and intersect RNA-GA；
Step 7: the sample data obtained in step 1 being input to position supporting vector machine model, obtains desired output PEout, will
Reality output P in step 2outWith PEoutError sum of squares intersect the objective function of RNA-GA as prominent ring, pass through RNA-GA
Optimizing obtains the optimal value of position least square method supporting vector machine and pivot angle least square method supporting vector machine parameter to be estimated respectively；
Step 8: the parameter optimal value obtained after step 7 is executed substitutes into position least square method supporting vector machine model and pendulum respectively
Angle least square method supporting vector machine model, for overhead crane horizontal position and pivot angle estimate.
2. the overhead crane model construction of SVM method for the ring crossover operation RNA-GA that dashes forward according to claim 1, feature
It is that the Radial basis kernel function form in the step 3 is as follows:
Wherein δ is the width parameter of gaussian kernel function；| | | | it is norm；X ', y ' are respectively one group of feature.
3. the overhead crane model construction of SVM method for the ring crossover operation RNA-GA that dashes forward according to claim 1, feature
It is to be normalized to zero-mean and unit variance in the step 4, normalizes formula are as follows:
Wherein xk,jIt is the data after normalization,Indicate j-th of component of k-th of sample, ujIt is j-th of component of all samples
Mean value, δjFor the standard deviation of j-th of component of all samples.
4. the overhead crane model construction of SVM method for the ring crossover operation RNA-GA that dashes forward according to claim 1, feature
Be in the step 6 dash forward ring crossover operation RNA-GA use prominent ring crossover operation steps are as follows:
Two RNA sequences are randomly choosed, the head end of two RNA sequences is aligned to form duplex structure, respectively to two RNA sequences
Traversing operation is executed, checks whether it can form prominent ring structure；It is random to generate one first to wherein any one RNA sequence
Position Pos1, then traverse [the 1, Pos of the RNA sequence1- 2] and [Pos1+ 2, L] position searching position Pos2, wherein L is RNA sequence
Column length, if Pos1And Pos2The base of two positions is capable of forming A-T G-C form complementary base pair, then the RNA sequence
Column can form prominent ring structure, and no person regenerates Pos1Position can form the Pos of prominent ring structure until finding1、Pos2, or time
Whole RNA sequence is gone through to also fail to find the Pos that can form prominent ring structure1、Pos2Until；Then another RNA sequence is also executed
Traversing operation；After having executed traversing operation to two RNA sequences, if there is prominent ring structure in two RNA sequences, at two
It is respectively formed prominent ring structure and transposition in RNA sequence, the tail portion of a wherein longer RNA sequence is exceeded into former sequence length
The part of L moves to the head end of another chain；If only having 1 in two RNA sequences has prominent ring structure, then in this RNA sequence
It is upper to form prominent ring structure and prominent ring structure dissociates into another chain position corresponding with prominent ring structure root base, it will wherein
The part that the tail portion of one longer RNA sequence exceeds former sequence length L moves to the head end of another chain；If two RNA sequences are equal
There is no prominent ring structures, then retain two chains to the next generation.
5. the overhead crane model construction of SVM method for the ring crossover operation RNA-GA that dashes forward according to claim 1, feature
It is in the step 6 that prominent ring intersects RNA-GA step and can be described as follows:
Step 6.1: the operating parameter of RNA-GA is set, including population scale N, maximum evolutionary generation G, individual lengths L, prominent ring are handed over
Pitch Probability pb, replacement operator Probability piAnd termination rules, wherein termination rules are to reach the low optimization accuracy of setting or reach calculation
The maximum evolutionary generation of method；
Step 6.2: the parameter of position or pivot angle least square method supporting vector machine model being encoded, N number of RNA is randomly generated
The initial population Pop of sequence, each RNA sequence represent the possibility solution of support vector machines unknown parameter, wherein each parameter
The RNA subsequence that one length is L, each support vector machines mould of overhead crane are encoded to by character set { 0,1,2,3 }
Type is made of two location parameters, i.e. kernel parameter and penalty factor；The code length of each RNA sequence is 2 × L；
Step 6.3: the error sum of squares that sample data inputs the output and reality output that obtain after supporting vector machine model is made
Intersect the objective function of RNA-GA for ring of dashing forward；
Step 6.4: using ratio selection strategy, by population dividing be two parts, i.e. " neutral individual " according to fitness value size
Popb" harmful individual " PopwTwo parts；
Step 6.5: to neutral individual PopbA) and b) two kinds of crossover operations are executed respectively, and steps are as follows:
A) from PopbIt randomly selects two sequences and forms duplex structure, with Probability pbThe prominent ring is executed to duplex structure and intersects behaviour
Make, generates the new population Pop of N/21；
B) equally to PopbIndividual is with Probability piReplacement operator is executed, the new population Pop of N/2 is generated2；
Population after crossover operation executes is represented by PopNew=[Pop1；Pop2；Popw], share 3N/2 individual；
Step 6.6: to population PopNewEach of individual respectively with self-adaptive mutation pmh、pmlTo its high-order [L/2+1, L]
TSP question operation, p are executed with low level [1, L/2]mh、pmlCalculation formula it is as follows:
Wherein a1For pmh、pmlInitial mutation probability, b1It is the variation range of mutation probability, g is current evolutionary generation, g0It is to turn
Break, aa are change rate；
Mutation operation is using the conversion of RNA sequence, transversion and exchange operation；
Step 6.7: if the optimal value precision of current population meets, termination is required or evolutionary generation satisfaction is terminated and required, then is obtained
The optimizing parameter of position supporting vector machine model and pivot angle support vector machines, otherwise return step 6.3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810062458.8A CN108341342B (en) | 2018-01-23 | 2018-01-23 | The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810062458.8A CN108341342B (en) | 2018-01-23 | 2018-01-23 | The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108341342A CN108341342A (en) | 2018-07-31 |
| CN108341342B true CN108341342B (en) | 2019-04-12 |
Family
ID=62960802
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810062458.8A Active CN108341342B (en) | 2018-01-23 | 2018-01-23 | The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108341342B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109871953B (en) * | 2019-01-25 | 2020-09-15 | 浙江大学 | Wavelet neural network modeling method for heavy oil cracking process of fpRNA genetic algorithm |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1062614A1 (en) * | 1999-01-19 | 2000-12-27 | Maxygen, Inc. | Methods for making character strings, polynucleotides and polypeptides |
| CN103177291A (en) * | 2013-04-19 | 2013-06-26 | 浙江大学 | Variable-search-space ribonucleic acid (RNA) genetic algorithm modeling method for continuous stirred tank reactor |
| CN103279659A (en) * | 2013-05-22 | 2013-09-04 | 浙江大学 | Fuel cell optimization modeling method of double loop crossover operation RNA genetic algorithm |
| CN105302966A (en) * | 2015-11-02 | 2016-02-03 | 浙江大学 | Overhead crane neural network modeling method based on RNA genetic algorithm of hairpin mutation operation |
| CN105354379A (en) * | 2015-11-02 | 2016-02-24 | 浙江大学 | Modeling method for bridge crane support vector machine of hairpin crossover operation RNA genetic algorithm |
| CN107463995A (en) * | 2017-08-01 | 2017-12-12 | 浙江大学 | Fuel cell Optimization Modeling method with Adaptive Genetic strategy RNA GA |
-
2018
- 2018-01-23 CN CN201810062458.8A patent/CN108341342B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1062614A1 (en) * | 1999-01-19 | 2000-12-27 | Maxygen, Inc. | Methods for making character strings, polynucleotides and polypeptides |
| CN103177291A (en) * | 2013-04-19 | 2013-06-26 | 浙江大学 | Variable-search-space ribonucleic acid (RNA) genetic algorithm modeling method for continuous stirred tank reactor |
| CN103279659A (en) * | 2013-05-22 | 2013-09-04 | 浙江大学 | Fuel cell optimization modeling method of double loop crossover operation RNA genetic algorithm |
| CN105302966A (en) * | 2015-11-02 | 2016-02-03 | 浙江大学 | Overhead crane neural network modeling method based on RNA genetic algorithm of hairpin mutation operation |
| CN105354379A (en) * | 2015-11-02 | 2016-02-24 | 浙江大学 | Modeling method for bridge crane support vector machine of hairpin crossover operation RNA genetic algorithm |
| CN107463995A (en) * | 2017-08-01 | 2017-12-12 | 浙江大学 | Fuel cell Optimization Modeling method with Adaptive Genetic strategy RNA GA |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108341342A (en) | 2018-07-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022012265A1 (en) | Robot learning from demonstration via meta-imitation learning | |
| Guerra et al. | Fuzzy control turns 50: 10 years later | |
| Zamuda et al. | Differential evolution and underwater glider path planning applied to the short-term opportunistic sampling of dynamic mesoscale ocean structures | |
| Cruz et al. | Path planning of multi-agent systems in unknown environment with neural kernel smoothing and reinforcement learning | |
| Auerbach et al. | Evolving complete robots with CPPN-NEAT: the utility of recurrent connections | |
| Sharma et al. | Fuzzy coding of genetic algorithms | |
| CN105976029B (en) | A kind of overhead crane neural network modeling approach of cuckoo behavior RNA-GA | |
| Guo et al. | A fusion method of local path planning for mobile robots based on LSTM neural network and reinforcement learning | |
| CN107633105B (en) | Improved hybrid frog-leaping algorithm-based quad-rotor unmanned aerial vehicle parameter identification method | |
| Chen et al. | An effective path planning of intelligent mobile robot using improved genetic algorithm | |
| CN108341342B (en) | The overhead crane model construction of SVM method of prominent ring crossover operation RNA-GA | |
| Ou et al. | GPU-based global path planning using genetic algorithm with near corner initialization | |
| Huang et al. | Dtpp: Differentiable joint conditional prediction and cost evaluation for tree policy planning in autonomous driving | |
| CN105302966A (en) | Overhead crane neural network modeling method based on RNA genetic algorithm of hairpin mutation operation | |
| Kumra et al. | Learning robotic manipulation tasks via task progress based Gaussian reward and loss adjusted exploration | |
| CN105354379B (en) | The overhead crane model construction of SVM method of hair clip crossover operation RNA genetic algorithm | |
| CN105260556B (en) | The overhead crane modeling method of hair clip mutation operation RNA genetic algorithm | |
| Pradhan et al. | Intelligent navigation of multiple coordinated robots | |
| Vianna et al. | Neural Network Based Model Predictive Control for an Autonomous Vehicle | |
| Zhang et al. | Deep reinforcement learning with domain randomization for overhead crane control with payload mass variations | |
| CN113239484B (en) | IGBO-based modeling method for RBF neural network of two-stage pendulum two-dimensional bridge crane | |
| Wu et al. | A rough set GA-based hybrid method for robot path planning | |
| Nakazono et al. | Load swing suppression in jib crane systems using a genetic algorithm-trained neuro-controller | |
| Krejsa et al. | Determination of optimal local path for mobile robot | |
| Tan et al. | Real-time global optimal path planning of mobile robots based on modified ant system algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |