CN107679312A - A kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network - Google Patents
A kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network Download PDFInfo
- Publication number
- CN107679312A CN107679312A CN201710890023.8A CN201710890023A CN107679312A CN 107679312 A CN107679312 A CN 107679312A CN 201710890023 A CN201710890023 A CN 201710890023A CN 107679312 A CN107679312 A CN 107679312A
- Authority
- CN
- China
- Prior art keywords
- contactor
- mrow
- msub
- armature
- neural network
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Hardware Design (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Geometry (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Tests Of Circuit Breakers, Generators, And Electric Motors (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network, its step are as follows：First, the size of each structure of contactor, rated voltage, coil resistance, coil turn, armature quality and each structure material therefor of contactor of contactor are obtained according to the drawing of contactor；2nd, the FEM model of contactor is established；3rd, contactor FEM model is emulated；4th, the radial basis function neural network model of contactor magnetic linkage and the radial basis function neural network model of electromagnetic contactor power are established；5th, the original state that contactor dynamic characteristic calculates is set, it is determined that calculating time step and calculating total time；6th, the dynamic characteristic equation of contactor is solved using Fourth order Runge-Kutta；7th, the obtained data dynamic characteristic for obtaining contactor corresponding with the time will be solved.This method takes into account computational efficiency and computational accuracy, is provided the foundation for contactor optimization design, has good actual application value.
Description
Technical field
The invention belongs to contactor design field, is related to a kind of contactor dynamic characteristic computational methods, and in particular to
A kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network.
Background technology
Contactor plays a part of to control connecting and disconnecting of the circuit in circuit system, and its performance is directly connected to the overall safety of circuit
Property and stability, are important components.The dynamic characteristic of contactor is the important performance indexes of contactor, and it can be to contactor
The electric life of switching circuit speed and contactor, which produces, directly to be influenceed.Recently as correlations such as power system and Aero-Space
The development in field, people propose higher and higher requirement to contactor dynamic characteristic, are necessary when design produces contactor
Design is optimized to the dynamic characteristic of contactor.Dynamic characteristic for optimization design contactor is, it is necessary to repeatedly to contactor
Dynamic characteristic carries out simulation calculation, therefore a kind of method that can quick and precisely calculate contactor dynamic characteristic is necessary.
The contactor dynamic characteristic computational methods commonly used in engineering include two kinds of Magnetic Circuit Method and FInite Element, wherein：Magnetic circuit
The computational efficiency of method is high but precision is low, is only applicable to approximate estimation；The precision of FInite Element is high but computational efficiency is low, is not suitable for
Need the situation of repeated multiple times computing.Therefore both the above dynamic characteristic computational methods are not all suitable for optimization design field.
The content of the invention
It is an object of the invention to provide a kind of quick meter of the contactor dynamic characteristic based on radial basis function neural network
Calculation method, this method take into account computational efficiency and computational accuracy, and contactor is solved by radial basis function neural network approximate model
Dynamic characteristic, asking for computational efficiency and accuracy in computation can not be taken into account simultaneously by solving existing contactor dynamic characteristic computational methods
Topic, provided the foundation for contactor optimization design, there is good actual application value.
The purpose of the present invention is achieved through the following technical solutions：
A kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network, including following step
Suddenly：
First, according to the drawing of contactor obtain the size of each structure of contactor, the rated voltage of contactor, contactor line
Enclose resistance, the coil turn of contactor, contactor armature quality and each structure material therefor of contactor；
The 2nd, the parameter obtained in step 1 is established to the FEM model of contactor in FLUX softwares；
3rd, the different coil currents of n groups, different armature misalignments are uniformly chosen as sampled point (ib, xb), it is soft to FLUX
The contactor FEM model that part is established is emulated, and above n groups coil current and armature displacement, emulation solution are inputted during emulation
Obtain the coil flux linkage ψ corresponding to above n group sampled pointsbAnd electromagnetic force F suffered by armaturebAs sampled data, b=1 ... n；
4th, according to the different coil currents obtained in step 3, magnetic linkage and electromagnetic force hits under different armature displacements
According to the radial basis function neural network model and the radial ba-sis function network of electromagnetic contactor power for establishing contactor magnetic linkage respectively
Network model；
5th, set the original state that contactor dynamic characteristic calculates, including the coil current of initial time, armature displacement,
Coil flux linkage, armature speed, it is determined that calculating time step Δ t and calculating total time tmax；
6th, Fourth order Runge-Kutta is utilized to contactor according to the coil current of contactor initial time and armature displacement
Dynamic characteristic equation is solved；
7th, the dynamic spy of contactor will have been obtained by solve corresponding with the time of the data that runge kutta method solves to obtain
Property.
The invention has the advantages that：
1st, the present invention proposes the method for building up of the radial basis function neural network model of electromagnetic contactor power (magnetic linkage), should
Radial basis function neural network model determines RBF center using Orthogonal Least Square.In the footpath of electromagnetic contactor power
The method that segmentation modeling is employed into basis function neural network modeling process, improve the RBF of electromagnetic force
The accuracy of neural network model, the electromagnetic force (magnetic in the contactor course of work can be fast and accurately calculated by the model
Chain).
2nd, contactor dynamic spy is quick and precisely calculated by radial basis function neural network model and Fourth order Runge-Kutta
Property, as a result error is within 5% compared with FInite Element, and calculating the time only needs 0.2% of FInite Element or so.
3rd, fast algorithm of the invention calculates contactor work using the FEM model commonly used in approximate model substitution engineering
Magnetic linkage and electromagnetic force during work, because the computational efficiency of approximate model is far above FEM model, and the degree of accuracy is with having
It is suitable to limit meta-model, therefore solution effect can be substantially improved in the fast algorithm while contactor dynamic characteristic solving precision is ensured
Rate.
4th, solving a contactor dynamic characteristic by 2D FEM models in engineering needs 15min or so, 3D finite elements
Model then needs a few hours, and solves the dynamic characteristic of a contactor then by dynamic characteristic fast algorithm proposed by the present invention
Only need 1.5s or so.
Brief description of the drawings
Fig. 1 is the flow chart of fast algorithm of the present invention.
Fig. 2 is contactor construction figure, 1- stationary contacts, 2- movable contacts, 3- connecting rods, 4- coils, 5- coil racks, 6- holes yoke
Iron, 7- excess of stroke springs, 8- yokes, 9- return springs, 10- connecting rods, 11- armature, 12- axle sleeves.
Fig. 3 is FEM model.
Fig. 4 is contactor coil electric current.
Fig. 5 is electromagnetic contactor power.
Fig. 6 is contactor magnetic linkage.
Fig. 7 is contactor armature displacement.
Embodiment
Technical scheme is further described below in conjunction with the accompanying drawings, but is not limited thereto, it is every to this
Inventive technique scheme is modified or equivalent substitution, without departing from the spirit and scope of technical solution of the present invention, all should cover
In protection scope of the present invention.
The invention provides a kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network,
As shown in figure 1, specific implementation step is as follows：
First, according to the drawing of contactor obtain the size of each structure of contactor, the rated voltage of contactor, contactor line
Enclose resistance, the coil turn of contactor, contactor armature quality and each structure material therefor of contactor.
2nd, the parameter obtained according to each physical dimension of contactor, material properties etc. in step 1 is in FLUX softwares
The FEM model of contactor is established, is comprised the following steps that：
The first step：The geometry mould of contactor is established according to the physical dimension of contactor by finite element emulation software FLUX
Type；
Second step：Subnetting processing is carried out to the geometrical model of contactor, obtains the contactor model after subnetting；
3rd step：According to the real material attribute of contactor, the physical attribute of each structure of model after contactor subnetting is determined,
Obtain the FEM model of contactor.
3rd, the different coil currents of n groups, different armature misalignments are uniformly chosen as sampled point (ib, xb), it is soft to FLUX
The contactor FEM model that part is established is emulated, and above n groups coil current and armature displacement, emulation solution are inputted during emulation
Obtain the coil flux linkage ψ corresponding to above n group sampled pointsbAnd electromagnetic force F suffered by armaturebAs sampled data, b=1 ... n.
4th, according to the different coil currents obtained in step 3, magnetic linkage and electromagnetic force hits under different armature displacements
According to the radial basis function neural network model and the radial ba-sis function network of electromagnetic contactor power for establishing contactor magnetic linkage respectively
Network model.
In this step, the radial basis function neural network model of electromagnetic contactor power is established according to electromagnetic force sampled data
Step is as follows, here by the mode of segmentation, passes through two radial ba-sis function networks respectively before and after contactor adhesive
Network model describes：
The first step：N group sampled points are normalized, make all sampled points all between 0 to 1.Normalizing used
It is as follows to change formula：
In formula：iminRepresent sample rate current minimum value, imaxRepresent sample rate current maximum, xminRepresent sampling armature displacement
Minimum value, xmaxRepresent sampling armature displacement maximum.
Second step：Calculate sampled point output matrix Φ*, and the individual column vectors of n ' (n=n ') in order matrix are respectively vectorial Pe
=[φ1eφ2e...φne]T, e=1 ... n '.
In formula：C is RBF width, typically takes 0.4.
3rd step：According to formulaF=[F in formula1F2....Fn]TFor n group electromagnetic force hits
According to calculating contribution rate δ E of each column vector to error, choose to error contribution rate highest i.e. | δ E | maximum column vector Pe, order
Qh=Pe, h=1 ... m, m represent the RBF central point number of electromagnetic force after contactor adhesive, corresponding to the column vector
Sampled point (ie, xe) it is chosen for RBF central point (i'h, x'h)。
4th step：The weights λ corresponding to RBF center is asked for by equation belowF, establish RBF mould
Type：
λF=(ΦT*Φ)-1ΦT*F；
In formula：
5th step：The estimate of all sampled points is calculated according to equation below：
And calculate the error sum of squares E of all sampled datas under the model：
In formula：FbRepresent F=[F in the 3rd step1F2....Fn]TB-th of electromagnetic force sampled data of the inside.
If error E meetsThen stop computing, wherein q represents relative error, typically takes 2%.This
When obtain m RBF center (i'h, x'h) and each center of RBF corresponding weights λF, for solving contactor
Electromagnetic force after adhesive.As error is unsatisfactory forThen continue the 6th step.
6th step：According to equation below by Φ*It is replaced：
In formula：QhRepresent in the 3rd step to error contribution rate highest i.e. | δ E | maximum column vector,Represent the row to
The transposition of amount.
Obtain one group of new column vector Pe(e=1 ... n '), the 3rd step is repeated, now need to ensure not repeat to select same
Column vector.
7th step：OrderObtain F'=[F'1F'2....F'n]T, k is index coefficient in formula, typically takes 3.By
F=[F in three steps1F2....Fn]TReplace with F'=[F'1F'2....F'n]TAs sampled data, the error E in the 5th step needs
The condition to be met replaces withThe 3rd step to the 6th step is re-started, is connect until obtaining v
The RBF center of electromagnetic force before tentaculum adhesive(u=1 ... v) and the corresponding weights at each center of RBFFor solving the electromagnetic force before contactor adhesive.To sum up complete electromagnetic contactor power radial basis function neural network mould
The foundation of type.
The radial basis function neural network model that contactor magnetic linkage is established according to magnetic linkage sampled data described in this step
The step of it is similar to establishing the radial basis function neural network model step of electromagnetic force, but staged operation need not be carried out, i.e., not
Need to carry out the 7th step, overall step is as follows：
The first step：N group sampled points are normalized, make all sampled points all between 0 to 1.Normalizing used
It is as follows to change formula：
In formula：iminRepresent sample rate current minimum value, imaxRepresent sample rate current maximum, xminRepresent sampling armature displacement
Minimum value, xmaxRepresent sampling armature displacement maximum.
Second step：Sampled point output matrix Φ * are calculated, and the individual column vectors of n ' in order matrix are respectively vectorial Pe=[φ1e
φ2e...φne]T, e=1 ... n '.
In formula：C is RBF width, typically takes 0.4.
3rd step：According to formulaψ=[ψ in formula1ψ2....ψn]TFor n group magnetic linkage hits
According to calculating contribution rate δ E of each column vector to error, choose to error contribution rate highest i.e. | δ E | maximum column vector Pe, order
Q's=Pe, s=1 ... r (r represents the RBF central point number of contactor magnetic linkage), the sampled point corresponding to the column vector
(ie, xe) it is chosen for RBF central point (is", x "s)。
4th step：The weights λ corresponding to RBF center is asked for by equation belowψ, establish RBF mould
Type：
λψ=(Φ 'T*Φ')-1Φ'T*ψ；
In formula：
5th step：The estimate of all sampled points is calculated according to equation below：
And calculate the error sum of squares E of all sampled datas under the model：
ψbRepresent ψ=[ψ in the 3rd step1ψ2....ψn]TB-th of magnetic linkage sampled data of the inside.
If error E meetsThen stop computing, wherein q represents relative error, typically takes 2%.This
When obtain r RBF center (i "s, x "s) and each center of RBF corresponding weights λψ, for solving contactor
Magnetic linkage, that is, complete contactor magnetic linkage radial basis function neural network model foundation.As error is unsatisfactory forThen continue the 6th step.
6th step：According to equation below by Φ*It is replaced：
Obtain one group of new column vector Pe(e=1 ... n '), the 3rd step is repeated, now need to ensure not repeat to select same
Column vector.
Finally give the common r RBF center (i " corresponding to magnetic linkages,x”s) (s=1 ... r) and corresponding weight
For λψ, for solving the magnetic linkage of contactor.To sum up complete building for the radial basis function neural network model of contactor magnetic linkage
It is vertical.
5th, set the original state that contactor dynamic characteristic calculates, including the coil current of initial time, armature displacement,
Coil flux linkage, armature speed, it is determined that calculating time step Δ t and calculating total time tmax。
6th, Fourth order Runge-Kutta is utilized to contactor according to the coil current of contactor initial time and armature displacement
Dynamic characteristic equation is solved.
Dynamic characteristic equation is：
In formula：ψ represents the magnetic linkage of magnetizing coil；vRankRepresent armature motion speed；ucRepresent the voltage of magnetizing coil；R tables
Show the resistance of magnetizing coil；I represents coil current；F、FfExpression acts on electromagnetic attraction, the reaction force of armature respectively；mRankTable
Show the quality of armature；X represents the displacement of armature.
Fourth order Runge-Kutta is a kind of high-precision single step algorithm wide variety of in engineering, passes through fourth order Runge-Kutta
The canonical form that method solves dynamic characteristic equation is as follows, is iterated and reaches t until calculating the timemaxComplete contact
The calculating of device dynamic characteristic.
In formula：Kj、Lj、MjRepresent respectivelyWithIt can be calculated according to dynamic characteristic equation.
Need during the iterative above equation according to corresponding to solving magnetic linkage and armature displacement coil current and
And pass through electromagnetic force corresponding to armature displacement and coil current solution, and the mechanical counter-force according to corresponding to solving armature displacement.
It is as follows the step of electromagnetic force according to corresponding to solving armature displacement and coil current in this step：
The first step：Coil current i and armature displacement x are determined, and electric current and armature displacement are normalized.
Second step：According to contactor armature displacement judge contactor whether adhesive, i.e., if x=xmax, then according to step 4
It is middle to establish obtained electromagnetic force radial basis function neural network model, pass through formula F=[φ1φ2...φm]*λF, in formulaElectromagnetic force when calculating coil current is i and armature displacement is x；If x ≠ xmax, then public affairs are passed through
FormulaIn formulaWhen calculating coil current is i and armature displacement is x
Electromagnetic force.
It is as follows the step of coil current according to corresponding to solving magnetic linkage and armature displacement in this step：
The first step：Determine coil flux linkage ψ and armature displacement x.
Second step：The table of comparisons on the magnetic linkage of magnetizing coil, electric current and armature displacement is established, is comprised the following steps that：
(1) descending coil current value and armature shift value are set according to armature total displacement and coil maximum current value
Data, and electric current and armature displacement are normalized.
(2) the descending coil current value and armature displacement Value Data that will be set in step (1), according in step 4
The radial basis function neural network model of the magnetic linkage of foundation, pass through formula ψ=[φ1φ2...φr]*λψ, in formulaCalculate the magnetic linkage value of corresponding setting coil current and armature displacement.
(3) table of comparisons is made in coil current value, armature shift value and corresponding magnetic linkage value, wherein, the behavior of the table of comparisons
Coil current, it is classified as armature displacement.
3rd step：According to the coil current at the control table search current time, by inserting if there is no respective value in table
Value determines present coil electric current.
It is as follows the step of mechanical counter-force according to corresponding to solving armature displacement in this step：
The first step：Determine the armature displacement x at current time.
Second step：According to formula Ff=kfX calculates the counter-force at current time, k in formulaf, can be from connecing for device of spring stiffness coefficient
Obtained in tentaculum drawing.
7th, by data such as the coil current for solving to obtain by runge kutta method, armature displacement, coil flux linkage, electromagnetic forces
Solve corresponding with the time has obtained the dynamic characteristic of contactor.
Embodiment：
For certain model Direct Action Type high-power direct current contactor, its dynamic characteristic is calculated, its structure is as shown in Fig. 2 this connects
Tentaculum rated voltage 28V, the Ω of coil resistance 40, coil turn are 2100 circles, and armature quality is 0.0203kg, main permeability magnetic material
For DT4E, armature total kilometres are 2.68e-3m, and coil maximum current is 0.7A.Specific calculation procedure is as follows：
First, the FEM model of contactor is established in FLUX softwares according to contactor drawing, due to contactor height
Symmetrically, 1/4 model therefore is only established.The FEM model established after obtaining subnetting is as shown in Figure 3.
2nd, 120 groups of difference coil currents, different armature misalignments are chosen as sampled point (ib,xb), levels of current is
0.1A, 0.2A ... 0.7A totally 8, level of displacement 0.2e-3m, 0.4e-3m ... 2.6e-3m, 2.68e-3m totally 15.By having
Limit meta-model and solve the coil flux linkage ψ corresponding to 120 groups of sampled points abovebAnd electromagnetic force F suffered by armaturebAs sampled data.
3rd, the radial direction base of electromagnetic contactor power and contactor magnetic linkage is established respectively according to the sampled data provided in step 2
Function Neural Network model.
4th, the original state that contactor dynamic characteristic calculates is set, the coil current of initial time is 0A, armature displacement is
0m, coil flux linkage 0Wb, armature speed are 0m/s2, it is determined that calculate time step be 1e-4s calculate total time be 50ms.
5th, Fourth order Runge-Kutta is utilized to contactor according to the coil current of contactor initial time and armature displacement
Dynamic characteristic equation is solved.
6th, contactor dynamic characteristic is obtained eventually through the computational methods based on radial basis function neural network, and will knot
The contactor dynamic characteristic Comparative result that fruit obtains with Finite element arithmetic, comparing result include contactor coil electric current with the time
Change curve (as shown in Figure 4), electromagnetic contactor power versus time curve (as shown in Figure 5), contactor magnetic linkage at any time
Between change curve (as shown in Figure 6), the armature displacement versus time curve (as shown in Figure 7) of contactor.
From Fig. 4-8, quick result of calculation and finite element result are basically identical in figure, each curve error 5% with
It is interior.15min or so is needed by the dynamic characteristic of Finite element arithmetic contactor, and connecing based on radial basis function neural network
The dynamic characteristic that tentaculum dynamic characteristic fast algorithm solves a contactor only needs 1.5s, it is only necessary to which FInite Element 0.2% is left
The right time, it is seen that method of the invention, which significantly saves, calculates the time.
Claims (8)
1. a kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network, it is characterised in that described
Method and step is as follows：
First, it is electric according to the coil of the size of the drawing of contactor acquisition each structure of contactor, the rated voltage of contactor, contactor
Resistance, the coil turn of contactor, contactor armature quality and each structure material therefor of contactor；
The 2nd, the parameter obtained in step 1 is established to the FEM model of contactor in FLUX softwares；
3rd, the different coil currents of n groups, different armature misalignments are uniformly chosen as sampled point (ib, xb), FLUX softwares are built
Vertical contactor FEM model is emulated, and above n groups coil current and armature displacement are inputted during emulation, is emulated solution and is obtained
Coil flux linkage ψ corresponding to above n group sampled pointsbAnd electromagnetic force F suffered by armaturebAs sampled data, b=1 ... n；
4th, according to the different coil currents obtained in step 3, the magnetic linkage under different armature displacements and electromagnetic force sampled data point
The radial basis function neural network model of contactor magnetic linkage and the radial basis function neural network mould of electromagnetic contactor power are not established
Type；
5th, the original state that contactor dynamic characteristic calculates, including the coil current of initial time, armature displacement, coil are set
Magnetic linkage, armature speed, it is determined that calculating time step Δ t and calculating total time tmax；
6th, the dynamic according to the coil current of contactor initial time and armature displacement using Fourth order Runge-Kutta to contactor
Characteristic equation is solved；
7th, the dynamic characteristic of contactor will have been obtained by solve corresponding with the time of the data that runge kutta method solves to obtain.
2. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that the step 2 comprises the following steps that：
The first step：The geometrical model of contactor is established according to the physical dimension of contactor by finite element emulation software FLUX；
Second step：Subnetting processing is carried out to the geometrical model of contactor, obtains the contactor model after subnetting；
3rd step：According to the real material attribute of contactor, the physical attribute of each structure of model after contactor subnetting is determined, is produced
To the FEM model of contactor.
3. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that in the step 4, the Radial Basis Function neural of electromagnetic contactor power is established according to electromagnetic force sampled data
The step of network model, is as follows：
The first step：N group sampled points are normalized, make all sampled points all between 0 to 1；
Second step：Calculate sampled point output matrix Φ*：
And the individual column vectors of n ' in order matrix are respectively vectorial Pe=[φ1e φ2e ... φne]T；
In formula：C is RBF width e=1 ... n ', n=n '；
3rd step：According to formula δ E=- (FTPe)2/(Pe TPe), F=[F in formula1 F2 .... Fn]TFor n group electromagnetic force hits
According to calculating contribution rate δ E of each column vector to error, choose to error contribution rate highest i.e. | δ E | maximum column vector Pe, order
Qh=Pe, h=1 ... m, m represent the RBF central point number of electromagnetic force after contactor adhesive, corresponding to the column vector
Sampled point (ie, xe) it is chosen for RBF central point (i'h, x'h)；
4th step：The weights λ of output node is asked for by equation belowF, establish RBF model：
λF=(ΦT*Φ)-1ΦT*F；
In formula：
5th step：The estimate of all sampled points is calculated according to equation below：
<mrow>
<msubsup>
<mi>F</mi>
<mi>b</mi>
<mo>*</mo>
</msubsup>
<mo>=</mo>
<mo>&lsqb;</mo>
<mtable>
<mtr>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mn>...</mn>
</mtd>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
<mo>&rsqb;</mo>
<mo>*</mo>
<msub>
<mi>&lambda;</mi>
<mi>F</mi>
</msub>
<mo>;</mo>
</mrow>
And calculate the error sum of squares E of all sampled datas under the model：
<mrow>
<mi>E</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>b</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>F</mi>
<mi>b</mi>
</msub>
<mo>-</mo>
<msubsup>
<mi>F</mi>
<mi>b</mi>
<mo>*</mo>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>;</mo>
</mrow>
In formula：FbRepresent b-th of electromagnetic force sampled data；
If error E meetsThen stop computing, wherein q represents relative error, now obtains m radial direction base
Function center (i'h, x'h) and each center of RBF corresponding weights λF, for solving the electromagnetic force after contactor adhesive；
As error is unsatisfactory forThen continue the 6th step；
6th step：According to equation below by Φ*It is replaced：
<mrow>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
<mo>=</mo>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>Q</mi>
<mi>h</mi>
</msub>
<mo>*</mo>
</msup>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mi>T</mi>
</msubsup>
<mo>*</mo>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
</mrow>
<mrow>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mi>T</mi>
</msubsup>
<mo>*</mo>
<msub>
<mi>Q</mi>
<mi>h</mi>
</msub>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
In formula：QhRepresent in the 3rd step to error contribution rate highest i.e. | δ E | maximum column vector,Represent the column vector
Transposition；
Obtain one group of new column vector Pe(e=1 ... n), the 3rd step is repeated, now need to ensure not repeat to select same column vector；
7th step：OrderObtain F'=[F1' F′2 .... F′n]T, k is index coefficient in formula, by the F in the 3rd step
=[F1 F2 .... Fn]TReplace with F'=[F1' F′2 .... F′n]TAs sampled data, the error E in the 5th step needs
The condition of satisfaction replaces withThe 3rd step to the 6th step is re-started, until obtaining v contact
The RBF center of electromagnetic force before device adhesiveAnd the corresponding weights at each center of RBFWherein u=
1 ... v, for solving the electromagnetic force before contactor adhesive.
4. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that in the step 4, the radial basis function neural network of contactor magnetic linkage is established according to magnetic linkage sampled data
The step of model, is as follows：
The first step：N group sampled points are normalized, make all sampled points all between 0 to 1；
Second step：Calculate sampled point output matrix Φ*：
And the individual column vectors of n ' in order matrix are respectively vectorial Pe=[φ1e φ2e ... φne]T；
In formula：C is RBF width e=1 ... n ', n=n '；
3rd step：According to formula δ E=- (ψTPe)2/(Pe TPe), ψ=[ψ in formula1 ψ2 .... ψn]TFor n group magnetic linkage sampled datas,
Contribution rate δ E of each column vector to error is calculated, is chosen to error contribution rate highest i.e. | δ E | maximum column vector Pe, make Q 's
=Pe, s=1 ... r, r represent the RBF central point number of contactor magnetic linkage, the sampled point (i corresponding to the column vectore,
xe) it is chosen for RBF central point (i "s, x "s)；
4th step：The weights λ corresponding to RBF center is asked for by equation belowψ, establish RBF model：
λψ=(Φ 'T*Φ')-1Φ'T*ψ；
In formula：
5th step：The estimate of all sampled points is calculated according to equation below：
<mrow>
<msubsup>
<mi>&psi;</mi>
<mi>b</mi>
<mo>*</mo>
</msubsup>
<mo>=</mo>
<mo>&lsqb;</mo>
<mtable>
<mtr>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mn>...</mn>
</mtd>
<mtd>
<msub>
<mi>&phi;</mi>
<mrow>
<mi>b</mi>
<mi>r</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
<mo>&rsqb;</mo>
<mo>*</mo>
<msub>
<mi>&lambda;</mi>
<mi>&psi;</mi>
</msub>
<mo>;</mo>
</mrow>
And calculate the error sum of squares E of all sampled datas under the model：
<mrow>
<mi>E</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>b</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>&psi;</mi>
<mi>b</mi>
</msub>
<mo>-</mo>
<msubsup>
<mi>&psi;</mi>
<mi>b</mi>
<mo>*</mo>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>;</mo>
</mrow>
ψbRepresent b-th of magnetic linkage sampled data；
If error E meetsThen stop computing, wherein q represents relative error, now obtains r radial direction base
Function center (i "s, x "s) and each center of RBF corresponding weights λψ, for solving the magnetic linkage of contactor, that is, complete to connect
The foundation of the radial basis function neural network model of tentaculum magnetic linkage；As error is unsatisfactory forThen continue
Six steps；
6th step：According to equation below by Φ*It is replaced：
<mrow>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
<mo>=</mo>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
<mo>-</mo>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mrow>
<mo>&prime;</mo>
<mi>T</mi>
</mrow>
</msubsup>
<mo>*</mo>
<mrow>
<mo>(</mo>
<mfrac>
<mrow>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mrow>
<mo>&prime;</mo>
<mi>T</mi>
</mrow>
</msubsup>
<mo>*</mo>
<msup>
<mi>&Phi;</mi>
<mo>*</mo>
</msup>
</mrow>
<mrow>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mrow>
<mo>&prime;</mo>
<mi>T</mi>
</mrow>
</msubsup>
<mo>*</mo>
<msubsup>
<mi>Q</mi>
<mi>h</mi>
<mo>&prime;</mo>
</msubsup>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
Obtain one group of new column vector Pe(e=1 ... n '), repeat the 3rd step, now need ensure repeat selection same row to
Amount；
Finally give the common r RBF center (i " corresponding to magnetic linkages,x″s) and corresponding weight be λψ, for solving
The magnetic linkage of contactor.
5. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that as follows the step of electromagnetic force according to corresponding to solving armature displacement and coil current in the step 6：
The first step：Coil current i and armature displacement x are determined, and electric current and armature displacement are normalized；
Second step：According to contactor armature displacement judge contactor whether adhesive, i.e., if x=xmax, then built according in step 4
Vertical obtained electromagnetic force radial basis function neural network model, passes through formula F=[φ1 φ2 ... φm]*λF, in formulaElectromagnetic force when calculating coil current is i and armature displacement is x；If x ≠ xmax, then public affairs are passed through
FormulaIn formulaWhen calculating coil current is i and armature displacement is x
Electromagnetic force.
6. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that as follows the step of coil current according to corresponding to solving magnetic linkage and armature displacement in the step 6：
The first step：Determine coil flux linkage ψ and armature displacement x；
Second step：Establish the table of comparisons on the magnetic linkage of magnetizing coil, electric current and armature displacement；
3rd step：It is true by interpolation if there is no respective value in table according to the coil current at the control table search current time
Settled preceding coil current.
7. the quick calculating side of the contactor dynamic characteristic according to claim 6 based on radial basis function neural network
Method, it is characterised in that the second step comprises the following steps that：
(1) descending coil current value and armature shift value number are set according to armature total displacement and coil maximum current value
According to, and electric current and armature displacement are normalized；
(2) by the descending coil current value and armature displacement Value Data of setting in step (1), established according in step 4
Magnetic linkage radial basis function neural network model, pass through formula ψ=[φ1 φ2 ... φr]*λψ, in formulaCalculate the magnetic linkage value of corresponding setting coil current and armature displacement；
(3) table of comparisons is made in coil current value, armature shift value and corresponding magnetic linkage value, wherein, the behavior coil of the table of comparisons
Electric current, it is classified as armature displacement.
8. the quick calculating side of the contactor dynamic characteristic according to claim 1 based on radial basis function neural network
Method, it is characterised in that as follows the step of mechanical counter-force according to corresponding to solving armature displacement in the step 6：
The first step：Determine the armature displacement x at current time；
Second step：According to formula Ff=kfX calculates the counter-force at current time, k in formulaf, can be from contactor for device of spring stiffness coefficient
Obtained in drawing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710890023.8A CN107679312B (en) | 2017-09-27 | 2017-09-27 | Rapid calculation method for contactor dynamic characteristics based on radial basis function neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710890023.8A CN107679312B (en) | 2017-09-27 | 2017-09-27 | Rapid calculation method for contactor dynamic characteristics based on radial basis function neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107679312A true CN107679312A (en) | 2018-02-09 |
| CN107679312B CN107679312B (en) | 2021-01-01 |
Family
ID=61136779
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710890023.8A Active CN107679312B (en) | 2017-09-27 | 2017-09-27 | Rapid calculation method for contactor dynamic characteristics based on radial basis function neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107679312B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754070A (en) * | 2018-12-28 | 2019-05-14 | 东莞钜威动力技术有限公司 | Insulation resistance value calculation method neural network based and electronic equipment |
| CN110232211A (en) * | 2019-05-07 | 2019-09-13 | 哈尔滨工业大学 | A kind of contactor spring characteristic computing method considering the effect of structure thermal field |
| CN110795862A (en) * | 2019-11-11 | 2020-02-14 | 哈尔滨工业大学 | Method for calculating dynamic characteristics of switching device based on coil current |
| CN111695258A (en) * | 2020-06-11 | 2020-09-22 | 哈尔滨工业大学 | Electromagnetic relay dynamic characteristic simulation kinetic energy injection simulation method |
| CN111898286A (en) * | 2020-04-24 | 2020-11-06 | 中国北方车辆研究所 | Motor modeling analysis and optimization method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101013148A (en) * | 2007-01-31 | 2007-08-08 | 哈尔滨工业大学 | Dynamic property test analysis system for DC hybrid contactor reliable life experiment |
| CN105893701A (en) * | 2016-04-25 | 2016-08-24 | 哈尔滨工业大学 | Method for calculating dynamic characteristics of electromagnetic mechanism based on temperature compensation matrix |
| CN106951664A (en) * | 2017-04-26 | 2017-07-14 | 福州大学 | The automatic modeling and hardware-in-loop simulation method of a kind of contactor |
-
2017
- 2017-09-27 CN CN201710890023.8A patent/CN107679312B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101013148A (en) * | 2007-01-31 | 2007-08-08 | 哈尔滨工业大学 | Dynamic property test analysis system for DC hybrid contactor reliable life experiment |
| CN105893701A (en) * | 2016-04-25 | 2016-08-24 | 哈尔滨工业大学 | Method for calculating dynamic characteristics of electromagnetic mechanism based on temperature compensation matrix |
| CN106951664A (en) * | 2017-04-26 | 2017-07-14 | 福州大学 | The automatic modeling and hardware-in-loop simulation method of a kind of contactor |
Non-Patent Citations (3)
| Title |
|---|
| 许志红等: "基于神经网络的智能交流接触器分断过程设计模型的建立", 《电工电能新技术》 * |
| 赵晓东: "大功率直流接触器动态特性建模及其优化方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 赵晓东等: "串联双绕组式大功率直流接触器动态特性仿真研究", 《低压电器》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109754070A (en) * | 2018-12-28 | 2019-05-14 | 东莞钜威动力技术有限公司 | Insulation resistance value calculation method neural network based and electronic equipment |
| CN109754070B (en) * | 2018-12-28 | 2022-10-21 | 东莞钜威软件科技有限公司 | Neural network-based insulation resistance value calculation method and electronic equipment |
| CN110232211A (en) * | 2019-05-07 | 2019-09-13 | 哈尔滨工业大学 | A kind of contactor spring characteristic computing method considering the effect of structure thermal field |
| CN110232211B (en) * | 2019-05-07 | 2021-11-23 | 哈尔滨工业大学 | Contactor bounce characteristic calculation method considering structure thermal field effect |
| CN110795862A (en) * | 2019-11-11 | 2020-02-14 | 哈尔滨工业大学 | Method for calculating dynamic characteristics of switching device based on coil current |
| CN111898286A (en) * | 2020-04-24 | 2020-11-06 | 中国北方车辆研究所 | Motor modeling analysis and optimization method |
| CN111898286B (en) * | 2020-04-24 | 2024-05-14 | 中国北方车辆研究所 | Motor modeling analysis and optimization method |
| CN111695258A (en) * | 2020-06-11 | 2020-09-22 | 哈尔滨工业大学 | Electromagnetic relay dynamic characteristic simulation kinetic energy injection simulation method |
| CN111695258B (en) * | 2020-06-11 | 2023-05-26 | 哈尔滨工业大学 | Electromagnetic relay dynamic characteristic simulation kinetic energy injection simulation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107679312B (en) | 2021-01-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107679312A (en) | A kind of quick calculation method of the contactor dynamic characteristic based on radial basis function neural network | |
| CN105260516B (en) | A kind of electromagnetical transient emulation method of the sub-network containing switching characteristic | |
| Amrhein et al. | 3-D magnetic equivalent circuit framework for modeling electromechanical devices | |
| CN106229964A (en) | A kind of based on the electrical power distribution network fault location method improving binary particle swarm algorithm | |
| CN108923401B (en) | A kind of calculation method and device of superconductive current limiter fault resstance and cut-off current | |
| Kokotovic | Subsystems, time scales and multimodeling | |
| CN107453928B (en) | A kind of power communication network node different degree evaluation method and device | |
| CN104375876B (en) | Electromagnetical transient emulation method is immunized in a kind of 0+ errors under input quantity catastrophe | |
| CN114744625B (en) | Wind turbine generator model order reduction method and system | |
| CN105701568A (en) | Heuristic power distribution network state estimation measurement position rapid optimization method | |
| CN110428684A (en) | The Physical Experiment artificial intelligence automotive engine system and working method of virtual reality | |
| CN108649562A (en) | A kind of power system load modeling parameter identification method based on blue wolf algorithm | |
| CN100470995C (en) | Distributed computing method of the features of the power system | |
| CN105375484A (en) | PMU-based electric power system distributed dynamic-state estimation method | |
| Hachicha et al. | Finite element method for induction machine parameters identification | |
| CN102542086B (en) | Sequential coupling analysis method for adaptive step length | |
| CN108460226B (en) | Method for calculating movement condition of oil flow in transformer oil tank | |
| CN106250658A (en) | On-load switch electromagnetic mechanism quick calculation method based on RBF | |
| CN107016142A (en) | Electromagnetic relay quick calculation method based on Kriging models | |
| CN110224401A (en) | In conjunction with the Power system transient stability prediction method of manual features and residual error network | |
| CN105281326A (en) | Slow coherency partition method considering damping influence of stator | |
| CN109460575A (en) | A kind of hybrid simulation interface system and interpolation algorithm based on Bei Jielong equivalence route | |
| CN105335550B (en) | A kind of power electronic system emulation mode of switching time self-correcting | |
| CN114091623A (en) | Transient voltage partitioning method and device based on multi-binary-table-AP clustering | |
| CN107196565A (en) | A kind of Computation of Nonlinear Characteristics on Switched Reluctance Motor line modeling method |
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 |