CN109194227A - Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication - Google Patents
Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication Download PDFInfo
- Publication number
- CN109194227A CN109194227A CN201811187655.9A CN201811187655A CN109194227A CN 109194227 A CN109194227 A CN 109194227A CN 201811187655 A CN201811187655 A CN 201811187655A CN 109194227 A CN109194227 A CN 109194227A
- Authority
- CN
- China
- Prior art keywords
- support vector
- rotor
- classification
- svm
- sample
- 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.)
- Withdrawn
Links
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02P—CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS
- H02P21/00—Arrangements or methods for the control of electric machines by vector control, e.g. by control of field orientation
- H02P21/14—Estimation or adaptation of machine parameters, e.g. flux, current or voltage
- H02P21/18—Estimation of position or speed
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Control Of Motors That Do Not Use Commutators (AREA)
Abstract
The present invention is directed to brushless DC motor rotor position detection problem, proposes a kind of Brushless DC Motor Position method for sensing for being based on support vector machines (SVM) Decision Classfication.Proposed by the present invention controlled based on the polytypic position sensing of SVM is using brushless DC motor stator voltage and current as the input of decision SVM, rotor position information is as output, DC motor rotor location is divided into 6 regions, rotor-position is indicated with region serial number 1-6,15 bis- classifiers of SVM are set, hierarchical classification decision is carried out；SVM optimized parameter is determined to SVM network training by grid optimization method, trained network model is applied in motor operation again, acquire the input of motor stator voltage and electric current as SVM, final rotor position information is determined by hierarchical decision making, logic commutation signal is calculated by rotor-position, determine that each region corresponds to the on-off of respective switch pipe, i.e. phase change logic signal.
Description
Technical field
It is specifically exactly that one kind is based on the present invention relates to a kind of rotor position sensing method in brshless DC motor field
The Brushless DC Motor Position method for sensing of support vector machines (SVM) Decision Classfication.
Background technique
Brshless DC motor, which controls electronics commutation circuit by rotor-position signal, keeps each winding of stator armature continuous
Commutation is powered, so that stator field and rotor permanent magnet magnetic field be made to remain 90 or so Space Angle, generates torque drive rotor
Operating.
The rotor position information of conventional brushless DC motor is measured by position sensor, and installation site is needed to detect
Device, but there is the brshless DC motor of position detecting device to have the disadvantage in that increase motor volume, it is small-sized to be unfavorable for motor
Change；Position sensor is mounted in space very limited inside motor, it is difficult to installation and maintenance difficult；It is difficult to adapt to badly
Environment；Sensor wire is complicated, is readily incorporated interference.Therefore the heat that brushless DC motor without position sensor is studied at people
Point.
The position measurement technique of rotor position-sensor-free device mainly has Based on Back-EMF Method, electric current testing, intelligence at present
Algorithm.Back-emf is directly proportional to speed, therefore cannot be obtained zero by detection back-emf in revolving speed very low even zero
Signal；And the realization of current method depends on the precision of current sensor, under normal circumstances, sensor can be obtained preferably
The signal of size of current, and the minor variations of current phase and waveform are then difficult to detect.Therefore Based on Back-EMF Method and current method list
Solely using has its limitation.
Summary of the invention
Technical problem: the gyrator channel detection method of position sensor and a variety of position-sensor-free devices has its office
It is sex-limited, therefore more difficult apply to requires motor operation relatively high occasion.
Technical solution: to solve the above-mentioned problems, support vector machines (SVM) Decision Classfication is used in rotor-position detection
It goes, motor steady running as requested is enabled by rate of current control.For brshless DC motor, rotor-position is believed
Mapping model number between electric moter voltage, electric current is set up, and realizes this mapping using support vector machines Decision Classfication.It will
The input of electric moter voltage and electric current as support vector machines, rotor position information is as output, to realize motor rotor position
Judgement.
The present invention be based on SVM on the basis of rotor-position detection system, by winding known to brushless motor working principle
A, the induction magnetic linkage of B, C phase winding is the function of electrical angle θ, and the induction magnetic linkage and three phase terminals voltage of winding A, B, C phase winding
There are certain relationships with electric current, therefore electrical angle θ, i.e. rotor-position signal can be predicted by three phase terminals voltage and current.
SVM is a kind of two disaggregated models, its purpose is to find a hyperplane to be split to sample, the original of segmentation
It is then margin maximization, is eventually converted into a convex quadratic programming problem to solve.It is general to describe nonlinear multivariable disaggregated model
Form are as follows: yi=sgn (g (xi)) whereinI=1,2 ..., N indicates the input of support vector cassification prediction model
Amount, yiIndicate simulated target output quantity.Mapping function Φ (xi):Using kernel function by the sample of original input space
It is mapped in high-dimensional feature space Ω, linear classification is carried out to sample data using mapping function in feature space.
Svm classifier may be expressed as: g (x)=(w Φ (x))+b (1)
Wherein vector w ∈ Rn, bias b ∈ R.SVM minimizes the structure wind of model while minimizing sample error
Danger, it may be assumed that
Constraint condition: yi[(wxi)+b]≤1-ξi, ξi>=0i=1 ..., N
Here ξiFor slack variable, equation (2) passes through the Lagrange multinomial of dual form, final classification is calculated
Function are as follows:
Classification results are determined according to the symbol of g (x), in formula: optimal Lagrange multiplier αi；B is bias term.G(xi, x) be
Kernel function, common part kernel function have radial basis function, global kernel function Polynomial kernel function, sig function etc..
Brushless DC Motor Position method for sensing proposed by the present invention based on support vector machines (SVM) Decision Classfication, A, B
Phase voltage ua(k),ub(k) electric current ia(k),ib(k),ia(k-1),ib(k-1) as the input of support vector machines, S (K) is rotor
Position signal, as the output of support vector machines, each SVM, which is exported, corresponds to an area sample (non-), continuous decision, directly
To separating each classification.Gj(xi, x) and it is kernel function, j=1,2 ..., L are respectively L classifier, i=1,2 ..., m.M is to support
Vector number, is automatically generated by algorithm of support vector machine, and k is time series, and K is rotor-position.
A kind of detection algorithm based on SVM Decision Classfication proposed by the present invention mainly includes support vector cassification
Modeling and model running part.
1. support vector cassification modeled segments mainly realize that steps are as follows:
Step1: signal: A, B phase is detected with model brshless DC motor acquisition system input and output to position sensor
Voltage
ua(k),ub(k) electric current ia(k),ib(k),ia(k-1),ib(k-1) as the input of support vector machines, S (K) is to turn
Sub- position
Confidence number, as the output of support vector machines, the 0-360 degree electrical angle that the rotor of direct current generator is rotated point
It is every
Totally 6 regions, rotor-position are indicated with region serial number 1-6 in 60 degree of regions.
Step2: 15 2 classifiers of support vector machines are set altogether, and 1 classifier of first layer is just classification non-1 (2-6 classification)
Sample be set to positive sample, the sample of classification non-6 (1-5 classification) is set to negative sample, 2 classifiers of the second layer, along non-1 region
The sample of non-2 classification is set to positive sample by (2-6) classification, and the sample of non-6 classification is set to negative sample, obtains second classification
Device so goes down, our available 15 such binary classifiers are as shown in table 1.
1 hierarchical classification device of table
Training data and test data are measured by position sensor brshless DC motor, 5000 will measured
Group training data and 2500 groups of test datas are normalized；
Determine kernel function Gj, kernel function can choose functions, the sig functions such as multinomial, radial base etc..
Step3: choosing the training sample of respective class according to classifier output respectively, and positive sample is+1, and negative sample is -1, adopts
15 support vector machines are trained with Vapnik algorithm, is tested using grid optimization method cross validation and obtains optimal models ginseng
Number penalty coefficient C, slack variable ξ and kernel functional parameter obtain 15 optimal two classifiers of support vector machines.
2. support vector cassification operation part is main to realize that steps are as follows:
Step1: relevant voltage and current input signal is acquired in real time and is normalized；
Step2: relevant voltage and current input signal is inputted into established support vector machine classifier and obtains rotor institute
In region Ki classification results；
When belong to 1,2 ..., 6 classes there are two and when result above keep last time classification results constant；When belong to 1,2 ..., 6
Keep last time classification results constant when class is completely without result.
Step3 carries out real-time control according to the rotor-position that classification obtains.Logic commutation signal is calculated by rotor-position,
Determine that each region corresponds to the on-off of respective switch pipe, i.e. phase change logic signal.
The transformation relation of rotor position information and phase change logic signal is as shown in table 2, wherein 1 indicates open-minded, 0 indicates to close
It is disconnected.
The transformation relation table of 2 rotor position information of table and phase change logic signal
The utility model has the advantages that position sensing method of the invention has many advantages, such as that dynamic property is good, robustness is high.The operation of algorithm
Rate is fast, improves controller response speed.
Detailed description of the invention
Fig. 1 is the Brushless DC Motor Position sensing arrangement figure based on support vector machines Decision Classfication.
Specific embodiment:
It is proposed by the present invention based on a pair of polytypic Brushless DC Motor Position method for sensing of support vector machines, in conjunction with being
Details are as follows for its specific embodiment of structure chart of uniting:
Step1: signal: A, B phase is detected with model brshless DC motor acquisition system input and output to position sensor
Voltage
ua(k),ub(k) electric current ia(k),ib(k),ia(k-1),ib(k-1) as the input of support vector machines, S (K) is to turn
Sub- position
Confidence number, as the output of support vector machines, the 0-360 degree electrical angle that the rotor of direct current generator is rotated point
It is every
Totally 6 regions, rotor-position are indicated with region serial number 1-6 in 60 degree of regions.
Step2: 15 2 classifiers of support vector machines are set altogether, and 1 classifier of first layer is just classification non-1 (2-6 classification)
Sample be set to positive sample, the sample of classification non-6 (1-5 classification) is set to negative sample, 2 classifiers of the second layer, along non-1 region
The sample of non-2 classification is set to positive sample by (2-6) classification, and non-6 sample is set to negative sample, obtains second classifier, such as
This goes down, our available 15 such binary classifiers are as shown in table 3.
3 hierarchical classification device of table
Training data and test data are measured by position sensor brshless DC motor, 5000 will measured
Group training data and 2500 groups of test datas are normalized；Determine that excitation function be G is radial basis function
Wherein x is input data, xiFor radial base center, σiFor radial base radius, i=1,2 ... m, m are center vector
Number.J be corresponding classifier serial number, j=1,2 ..., 15.
Step3: choosing the training sample of respective class according to classifier output respectively, and positive sample is+1, and negative sample is -1, adopts
15 support vector machines are trained with Vapnik algorithm, stay cross validation test to obtain optimal mould using grid optimization method
Shape parameter penalty coefficient C, slack variable ξ and radial base radius sigmaI,Obtain 15 optimal two classifiers of support vector machines.Parameter net
Lattice are chosen as follows by index variation:
Penalty coefficient C:106-10-1；Slack variable ξ: 10-5-10-1, radial base radius sigmai: 10-5-10-1
Second part: mainly realize that steps are as follows in support vector cassification model running part:
Step1: relevant voltage and current input signal is acquired in real time and is normalized；
Step2: relevant voltage and current input signal is inputted into established support vector machine classifier and obtains rotor institute
In region KiClassification results；
When belong to 1,2 ..., 6 classes there are two and when result above keep last time classification results constant；When belong to 1,2 ..., 6
Keep last time classification results constant when class is completely without result.
Step3 judges rotor-position S (K), i.e. rotor region K by support vector cassification networki, according to turn
Sub- position carries out real-time control.Logic commutation signal is calculated by rotor-position, determines that each region corresponds to respective switch pipe
On-off, i.e. phase change logic signal.
Above-mentioned specific implementation is preferable realization of the invention, and certainly, the invention may also have other embodiments,
Without deviating from the spirit and substance of the present invention, those skilled in the art make various in accordance with the present invention
Corresponding changes and modifications, but these corresponding changes and modifications all should belong to scope of protection of the claims of the invention.
Claims (1)
1. the Brushless DC Motor Position method for sensing that one kind is based on support vector machines (SVM) Decision Classfication, it is characterised in that logical
Cross the position signal that SVM Decision Classfication device obtains brushless DC motor rotor, reduce motor because position sensor there are due to band
The error and reduction motor volume come.A kind of detection algorithm based on SVM Decision Classfication proposed by the present invention, using branch
Hold two taxonomic structure of vector machine, by hierarchical decision making determine final classification export, mainly include support vector cassification modeling and
Model running part:
1) support vector cassification modeled segments mainly realize that steps are as follows:
Step1: signal: A, B phase voltage u is detected with model brshless DC motor acquisition system input and output to position sensora
(k),ub(k) electric current ia(k),ib(k),ia(k-1),ib(k-1) as the input of support vector machines, S (K) is rotor-position letter
Number, as the output of support vector machines, each SVM exports a corresponding area sample (non-), continuous decision, until separating
Each classification.The 0-360 degree electrical angle that the rotor of direct current generator rotates is divided into every 60 degree of regions totally 6 regions, rotor
Position is indicated with region serial number 1-6.The training data measured and test data are normalized；
Step2: 15 2 classifiers of support vector machines are set altogether, and 1 classifier of first layer is just the sample of classification non-1 (2-6 classification)
Originally it is set to positive sample, the sample of classification non-6 (1-5 classification) is set to negative sample, 2 classifiers of the second layer, along non-1 region (2-6)
The sample of non-2 classification is set to positive sample by classification, and the sample of non-6 classification is set to negative sample, obtains second classifier, so
Go down, our available 15 such binary classifiers are as shown in table 1, determine kernel function Gj, kernel function can be chosen multinomial
Functions, the sig functions such as formula, radial base etc..
1 hierarchical decision classifier of table
Step3: choosing the training sample of respective class according to classifier output respectively, and positive sample is+1, and negative sample is -1, uses
Vapnik algorithm is trained 15 support vector machines, is tested using grid optimization method cross validation and obtains optimal model parameters
Penalty coefficient C, slack variable ξ and kernel functional parameter obtain 15 optimal two classifiers of support vector machines.
2) support vector cassification operation part is main realizes that steps are as follows:
Step1: relevant voltage and current input signal is acquired in real time and is normalized；
Step2: relevant voltage and current input signal is inputted into established support vector machine classifier and obtains rotor location
Domain Ki classification results；
When belong to 1,2 ..., 6 classes there are two and when result above keep last time classification results constant；When belong to 1,2 ..., 6 classes it is complete
Keep last time classification results constant when no result.
Step3 carries out real-time control according to the rotor-position that classification obtains.Logic commutation signal is calculated by rotor-position, is determined
Each region corresponds to the on-off of respective switch pipe, i.e. phase change logic signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811187655.9A CN109194227A (en) | 2018-10-09 | 2018-10-09 | Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811187655.9A CN109194227A (en) | 2018-10-09 | 2018-10-09 | Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109194227A true CN109194227A (en) | 2019-01-11 |
Family
ID=64948155
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811187655.9A Withdrawn CN109194227A (en) | 2018-10-09 | 2018-10-09 | Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109194227A (en) |
-
2018
- 2018-10-09 CN CN201811187655.9A patent/CN109194227A/en not_active Withdrawn
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ebrahimi et al. | Diagnosis and performance analysis of three-phase permanent magnet synchronous motors with static, dynamic and mixed eccentricity | |
| CN104378038B (en) | Permanent magnet synchronous motor parameter identification method based on artificial neural network | |
| CN106788075A (en) | Rapid vector screening prediction method for controlling torque based on improved Euler method | |
| CN109167545A (en) | Magneto alternator magnetic linkage on-line identification method and system | |
| CN109194225B (en) | Online identification method for parameters of doubly-fed motor | |
| CN102662323B (en) | Adoptive sliding mode control method and adoptive sliding mode control system of wind power generation variable-pitch actuator | |
| Aydin et al. | Artificial immune based support vector machine algorithm for fault diagnosis of induction motors | |
| Sediki et al. | ADALINE approach for induction motor mechanical parameters identification | |
| CN102510253B (en) | Sensorless controller of bearingless synchronous reluctance motor and method for controlling sensorless controller | |
| CN109768749A (en) | Threephase asynchronous machine rotor time constant real-time correction method | |
| CN109946978A (en) | A kind of servo-system fractional model discrimination method considering time delay process | |
| Bohari et al. | Speed tracking of indirect field oriented control induction motor using neural network | |
| CN109194227A (en) | Brushless DC Motor Position method for sensing based on support vector machines Decision Classfication | |
| Attestog et al. | Robust active learning multiple fault diagnosis of PMSM drives with sensorless control under dynamic operations and imbalanced datasets | |
| Kumar et al. | Accurate fault diagnosis and classification scheme based on non-parametric, statistical-frequency features and neural networks | |
| CN109217744A (en) | Brushless DC Motor Position method for sensing based on support vector machines hierarchical classification | |
| Rai et al. | A comparative performance analysis for loss minimization of induction motor drive based on soft computing techniques | |
| CN109245631A (en) | Based on the polytypic Brushless DC Motor Position method for sensing of support vector machines | |
| CN109150027A (en) | Based on a pair of polytypic Brushless DC Motor Position method for sensing of support vector machines | |
| CN107947650A (en) | A kind of brushless direct current motor sensorless control method based on extreme learning machine classification | |
| Wang et al. | Broken rotor bar fault detection of induction motors using a joint algorithm of trust region and modified bare-bones particle swarm optimization | |
| CN109150054A (en) | Brushless DC Motor Position method for sensing based on LSSVM Decision Classfication | |
| CN109120191A (en) | Brushless DC Motor Position method for sensing based on LSSVM hierarchical classification | |
| CN109120192A (en) | Based on a pair of polytypic Brushless DC Motor Position method for sensing of LSSVM | |
| Banerjee et al. | Hybrid intelligent predictive control system for high speed BLDC motor in aerospace application |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190111 |