CN106934157A - Primary cut-out recognition methods based on SVMs and Dynamics Simulation - Google Patents
Primary cut-out recognition methods based on SVMs and Dynamics Simulation Download PDFInfo
- Publication number
- CN106934157A CN106934157A CN201710146696.2A CN201710146696A CN106934157A CN 106934157 A CN106934157 A CN 106934157A CN 201710146696 A CN201710146696 A CN 201710146696A CN 106934157 A CN106934157 A CN 106934157A
- Authority
- CN
- China
- Prior art keywords
- svms
- primary cut
- classification
- threedimensional model
- grader
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Computer Graphics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of primary cut-out recognition methods based on SVMs and Dynamics Simulation, including the threedimensional model of primary cut-out is set up, calibrate and verify threedimensional model；By changing size, material, the characterisitic parameter of threedimensional model, the various failures of analogue simulation primary cut-out obtain corresponding various mechanical property waveforms；Waveform interception, mathematical morphology filter and feature extraction treatment are carried out to the corresponding mechanical property waveform of various failures, characteristic vector sample set is obtained；Based on SVMs and its one-to-many sorting technique, grader is trained, and calculate the confidence level of each classification；The unknown failure characteristic vector of actual measurement is input to the SVMs of training, fault type recognition is carried out, realizes that Mechanical Failure of HV Circuit Breaker is classified, and complete self study.The present invention need not carry out a large amount of physical simulation experiments, and the Fault Identification for primary cut-out can be accurately identified, and with self-learning function, be had broad application prospects.
Description
Technical field
The present invention relates to electrical equipment malfunction identification technology field, and in particular to one kind is dynamic based on SVMs and many bodies
The primary cut-out recognition methods of Mechanics Simulation.
Background technology
Primary cut-out is one of power equipment of quantity maximum in power system, while being also that most important switch sets
It is standby, the dual role of control and protection is responsible for, its functional reliability is the key factor for determining safe operation of power system.
But, because the internal structure of primary cut-out is invisible, it is difficult to intuitively know whether component therein is normal.
However, measure analysis to intraware after operating primary cut-out is disassembled seems unrealistic again, so, in order to
Know the machine performance of primary cut-out, typically processed and analyzed by the stroke-time graph for measuring, judge the machine
Whether tool mechanism is working properly.To further pass through the machine performance that stroke-time graph judges primary cut-out, generally adopt
With to mechanical defects such as the artificial setting divide-shut brake spring fatigue of primary cut-out, not in place, the oily buffer invalidations of energy storage, by right
Defect primary cut-out carries out substantial amounts of physical simulation experiment and obtains enough sample sets, builds on this basis for failure classes
The grader of type identification, and then its working condition is assessed, but, still have the following disadvantages：
（1）Only some failure can be realized by the adjustment to primary cut-out mechanical mechanism, also many failures
Cannot simulate so that the rich excessively dullness of sample set, Classification and Identification cannot be carried out for some unknown states；
（2）Some failure, such as failure of transmission damping exception class, are difficult to accomplish quantitative analysis by the method tested,
So that not careful enough for the classification of primary cut-out working condition；
（3）Stroke-time graph is obtained in the method tested, the mechanical life of primary cut-out can be lost, particularly some events
Experiment under barrier state, can more be such that its mechanical life loses significantly, cause very big economic loss；
（4）The product type of primary cut-out is more, architectural difference is big, and fault simulating test result is only applicable with fault diagnosis algorithm
In experiment breaker same model product, do not have versatility for the breaker of other models, structure.
How to overcome above-mentioned deficiency, be urgent problem during the mechanical Fault Identification of current primary cut-out.
The content of the invention
The purpose of the present invention is the problem for overcoming Shortcomings during the mechanical Fault Identification of current primary cut-out.This
The primary cut-out recognition methods based on SVMs and Dynamics Simulation of invention, using Dynamics Simulation and
Grader based on SVMs, realizes accurately identifying for the mechanical failure of primary cut-out, the method phase with traditional experiment
Than without carrying out a large amount of physical simulation experiments, and with self-learning function, constantly improve grader, for primary cut-out
Fault Identification, more economical, convenient, safety, has broad application prospects.
In order to achieve the above object, the technical solution adopted in the present invention is：
A kind of primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：Including with
Lower step,
The threedimensional model of primary cut-out is set up, the mechanical property waveform of threedimensional model is obtained by Dynamics Simulation, and
Contrasted with experimental test data by by simulation result, calibrated and verified threedimensional model；
By changing size, material, the characterisitic parameter of threedimensional model, based on each of Dynamics Simulation high voltage breaker simulator
Failure is planted, corresponding various mechanical property waveforms are obtained；
Waveform interception, mathematical morphology filter, feature extraction treatment are carried out to the corresponding mechanical property waveform of various failures, is obtained
Characteristic vector sample set；
Based on SVMs and its one-to-many sorting technique, grader is trained, and calculate the confidence level of each classification；
The unknown failure characteristic vector of actual measurement is input to the SVMs of training, fault type recognition is carried out, high pressure is realized
Breaker mechanical failure modes, and complete self study.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
The process of the threedimensional model of primary cut-out is set up, is comprised the following steps,
（1）The major loop and each parts of operating mechanism of primary cut-out are drawn by 3D sculpting software；
（2）Real work situation according to primary cut-out is assembled.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
The mechanical property waveform of threedimensional model is obtained by Dynamics Simulation, and is entered with experimental test data by by simulation result
Row contrast, calibrates and verifies threedimensional model, comprises the following steps,
（1）Simulation calculation is carried out to threedimensional model using multi-body Dynamic Analysis software, each type kinematic pair is added to threedimensional model,
Divide-shut brake spring, oil bumper characterisitic parameter are defined, and to all parts addition density parameters of threedimensional model, obtains normal shape
Mechanical characteristic of high-voltage circuit breaker simulation result under state；
（2）Surveyed with the experiment to actual high-voltage breaker by by the mechanical characteristic of high-voltage circuit breaker simulation result under normal condition
Examination data are contrasted, and calibrate the simulation parameter of threedimensional model.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
Divide-shut brake spring fatigue is simulated by changing the pre compressed magnitude or stiffness coefficient of divide-shut brake spring in primary cut-out；By changing
Become the damped coefficient of oil bumper in primary cut-out to simulate oil bumper failure；By changing primary cut-out handle ratchet wheel position
Put not in place to simulate energy storage；For per above-mentioned each emulation mode, 1% undulating value being given when setting simulation parameter, it is used for
Simulate the different dispersivenesses for organizing data under same emulation mode.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
Waveform interception, mathematical morphology filter, feature extraction treatment are carried out to the corresponding mechanical property waveform of various failures, feature is obtained
Vectorial sample set, comprises the following steps,
（1）Stroke curve to High Voltage Circuit Breaker Contacts is intercepted, and reduces data volume, is with the point that first stroke is not zero
Beginning flag, is maintained between 199.5 ~ 200.5mm with stroke, and reaches 20 data points for end mark carries out interception ripple
Shape；
（2）Treatment is filtered to the stroke curve of contact using Mathematical Morphology Method, i.e., it is special to pending machinery successively
Property waveform be opened and closed computing and make and break computing, and the result of opening and closing operation and make and break calculation process is averaged, be to ensure
The accuracy that follow-up maximum is obtained, the curve to averaged carries out interpolation processing；
（3）Feature extraction is carried out to doing the mechanical property waveform after difference treatment, the transverse and longitudinal coordinate conduct residing for its maximum is taken
Two-dimensional feature vector, so as to form characteristic vector sample set.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
Based on SVMs and its one-to-many sorting technique, training grader, and the confidence level of each classification is calculated, including it is following
Step,
（1）Based on one-to-many sorting technique, will be extended to suitable for the SVMs of two classification can carry out polytypic point
Class device, SVMs Selection of kernel function RBF, penalty factor c and kernel functional parameter g passes through particle swarm optimization algorithm
Obtain, the purpose for finding optimal penalty factor and kernel functional parameter is to make classification error rate minimum, and wherein classification error rate is adopted
Calculated with ten folding cross validation methods, training obtains the grader based on SVMs；
（2）Using sample training（1）The grader based on SVMs for obtaining, generates the decision hyperplane of each classification；
（3）The region comprising all characteristic values is taken, mesh generation is carried out, dot matrix is generated, lattice spacing selection is less than 0.1mm*
0.1mm；
（4）Utilize（2）In have each classification decision hyperplane grader pair（3）Each point in dot matrix carries out Classification and Identification,
Based on each point and the relation of each categorised decision hyperplane, assessment fraction is given respectively for each categorised decision hyperplane, and
The point is defined as assessment fraction highest that class, if respectively assessment fraction is below the threshold value of setting, then it is assumed that the point is not for
Know classification；
（5）According to（4）, every kind of classification is all included corresponding dot matrix, the assessment a little of the included institute of every kind of classification is divided
Number does normalized, and each point assesses the normalization computing formula such as formula of fraction（1）Shown, the assessment fraction after treatment is characterized and divided
The relative reliability of class result；
Value=(Value-min(Value))/(max(Value)-min(Value)) （1）
Wherein, Value is assessment fraction；Max (Value) is place classification assessment fraction maximum；Min (Value) is place
Classification assessment fraction minimum value.
The foregoing primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：
The unknown failure characteristic vector of actual measurement is input to the SVMs of training, fault type recognition is carried out, high pressure open circuit is realized
Device mechanical breakdown is classified, and completes self study, is comprised the following steps,
（1）Testing mechanical characteristic is carried out to primary cut-out, the mechanical property ripple tested under the unknown state for obtaining in experiment is taken
Shape carries out mathematical morphology filter and waveform interception, and carries out characteristic vector pickup；
（2）The characteristic vector extracted is identified by grader, the region according to corresponding to this feature vector is carried out to it
Classification, and provide confidence level target；
（3）If by finding that this characteristic vector is not belonging to known all kinds of states in grader after identification, being regarded as new
Increasing state, and perform（4）；
（4）After characteristic vector is identified into classification, then it is used as sample grader is trained, and to decision hyperplane
Delimitation make amendment, and relative reliability to each classification makes amendment so that grader completes self study.
The beneficial effects of the invention are as follows：Primary cut-out based on SVMs and Dynamics Simulation of the invention
Recognition methods, using Dynamics Simulation and the grader based on SVMs, realizes the mechanical failure of primary cut-out
Accurately identify, compared with the method for traditional experiment, without carrying out a large amount of physical simulation experiments, and with self-learning function, no
It is disconnected to improve grader, for the Fault Identification of primary cut-out, can accurately identify, without transforming device structure, peace
It is complete reliable, greatly reduce physical simulation experiment workload, economy is convenient, and grader possesses self-learning function, more economical, side
Just, safety, has broad application prospects.
Brief description of the drawings
Fig. 1 is the flow of the primary cut-out recognition methods based on SVMs and Dynamics Simulation of the invention
Figure；
Fig. 2 is the flow chart for finding optimal penalty factor and kernel functional parameter of the invention；
Fig. 3 is the flow chart that SVMs of the invention was trained and was carried out to test data Classification and Identification.
Specific embodiment
Below in conjunction with Figure of description, the present invention is further illustrated.
As shown in figure 1, the primary cut-out recognition methods based on SVMs and Dynamics Simulation of the invention,
Comprise the following steps,
Step（A）, the threedimensional model of primary cut-out is set up, the mechanical property of threedimensional model is obtained by Dynamics Simulation
Waveform, and contrasted with experimental test data by by simulation result, threedimensional model is calibrated and verifies, specifically include following step
Suddenly,
（A1）The major loop and each parts of operating mechanism of primary cut-out are drawn by 3D sculpting software, including it is dynamic tactile
Head, static contact, closing magnet master part, separating brake keep sincere son, cam, ratchet, connecting lever and other each drive links and locating part etc.；
（A2）Real work situation according to primary cut-out is assembled, specially by drawn part according to high pressure open circuit
The real work situation of device is assembled；
Step（B）, it is disconnected based on Dynamics Simulation simulated high-pressure by changing size, material, the characterisitic parameter of threedimensional model
The various failures of road device, obtain corresponding various mechanical property waveforms, specifically include following steps,
（B1）Simulation calculation is carried out to threedimensional model using multi-body Dynamic Analysis software, each type games are added to threedimensional model
Pair, defines divide-shut brake spring, oil bumper characterisitic parameter, and to all parts addition density parameters of threedimensional model, obtain just
Mechanical characteristic of high-voltage circuit breaker simulation result under normal state；
（B2）By by the mechanical characteristic of high-voltage circuit breaker simulation result under normal condition and the experiment to actual high-voltage breaker
Test data is contrasted, and calibrates the simulation parameter of threedimensional model；
For example, it is tired to simulate divide-shut brake spring by changing the pre compressed magnitude or stiffness coefficient of divide-shut brake spring in primary cut-out
Labor；Oil bumper failure is simulated by changing the damped coefficient of oil bumper in primary cut-out；By changing high pressure open circuit
Device handle ratchet wheel position is not in place to simulate energy storage；For per above-mentioned each emulation mode, 1% being given when setting simulation parameter
Undulating value, the dispersivenesses for simulating different group data under same emulation mode；
Step（C）, waveform interception, mathematical morphology filter, feature extraction are carried out to the corresponding mechanical property waveform of various failures
Treatment, obtains characteristic vector sample set, specifically includes following steps,
（C1）Stroke curve to High Voltage Circuit Breaker Contacts is intercepted, and reduces data volume, with the point that first stroke is not zero
It is beginning flag, is maintained between 199.5 ~ 200.5mm with stroke, and reaches 20 data points for end mark carries out interception ripple
Shape；
（C2）Treatment is filtered to the stroke curve of contact using Mathematical Morphology Method, i.e., it is special to pending machinery successively
Property waveform be opened and closed computing and make and break computing, and the result of opening and closing operation and make and break calculation process is averaged, be to ensure
The accuracy that follow-up maximum is obtained, the curve to averaged carries out interpolation processing；
（C3）Feature extraction is carried out to doing the mechanical property waveform after difference treatment, the transverse and longitudinal coordinate conduct residing for its maximum is taken
Two-dimensional feature vector, so as to form characteristic vector sample set；
Step（D）, based on SVMs and its one-to-many sorting technique, train grader, and calculate each classification can
Reliability, specifically includes following steps,
（D1）Based on one-to-many sorting technique, will be extended to suitable for the SVMs of two classification can carry out polytypic point
Class device, SVMs Selection of kernel function RBF, penalty factor c and kernel functional parameter g passes through particle swarm optimization algorithm
Obtain, the purpose for finding optimal penalty factor and kernel functional parameter is to make classification error rate minimum, and wherein classification error rate is adopted
Calculated with ten folding cross validation methods, training obtains the grader based on SVMs；
（D2）Using sample training（D1）The grader based on SVMs for obtaining, generates the decision hyperplane of each classification；
（D3）The region comprising all characteristic values is taken, mesh generation is carried out, dot matrix is generated, lattice spacing selection is less than 0.1mm*
0.1mm, dot matrix is more intensive, then classify thinner；
（D4）Utilize（D2）In have each classification decision hyperplane grader pair（D3）Each point in dot matrix carries out classification knowledge
Not, based on each point and the relation of each categorised decision hyperplane, assessment fraction is given respectively for each categorised decision hyperplane,
And think that the point belongs to assessment fraction highest that class, if respectively assessment fraction is below threshold value（Can be -0.4）, then it is assumed that should
Point is unknown classification；
（D5）According to（D4）, every kind of classification is all included corresponding dot matrix, included to every kind of classification assessment a little
Fraction does normalized, and each point assesses the normalization computing formula such as formula of fraction（1）Shown, the assessment fraction after treatment is characterized
The relative reliability of classification results；
Value=(Value-min(Value))/(max(Value)-min(Value)) （1）
Wherein, Value is assessment fraction；Max (Value) is place classification assessment fraction maximum；Min (Value) is place
Classification assessment fraction minimum value.
Step（E）, the unknown failure characteristic vector of actual measurement is input to the SVMs of training, carry out fault type knowledge
Not, realize that Mechanical Failure of HV Circuit Breaker is classified, and complete self study, specifically include following steps,
（E1）Testing mechanical characteristic is carried out to primary cut-out, the mechanical property ripple tested under the unknown state for obtaining in experiment is taken
Shape carries out mathematical morphology filter and waveform interception, and carries out characteristic vector pickup；
（E2）The characteristic vector extracted is identified by grader, the region according to corresponding to this feature vector is carried out to it
Classification, and provide confidence level target；
（E3）If by finding that this characteristic vector is not belonging to known all kinds of states in grader after identification, being regarded as new
Increasing state, and perform（E4）；
（E4）After characteristic vector is identified into classification, then it is used as sample grader is trained, and it is super to decision-making flat
Amendment is made in the delimitation in face, and relative reliability to each classification makes amendment so that grader completes self study.
For step（D1）The process of optimal penalty factor and kernel functional parameter is found, as shown in Fig. 2
（1）The emulation data of n class different conditions are stored under n different file respectively, are easy to distinguish, it is imitative in matlab
Spot file first in true software, then extracts wherein all of csv files, mechanical property Wave data is imported, by preceding
Waveform interception, mathematical morphology filter and feature extraction described in text, characteristic quantity is saved as respectively according to status categories
Data1, data2, data3 ..., data (n), be then merged into a matrix and conveniently call, data=[data1;
data2;data3;……;data(n)]；
（2）Be that n classes state creates class formative, such as theclass1=1, theclass2=2 etc., this class formative and feature to
The classification of amount is corresponded, and quantity and storage order both of which are consistent, with（1）It is similar, class formative is also merged into one
Individual matrix is conveniently called,
theclass=[theclass1;theclass2;theclass3……theclass(n)]；
（3）Iterations maxgen, population scale sizepop, speed undated parameter c1, c2, particle position span are set
[popmin, popmax], particle rapidity span [Vmin, Vmax]；
（4）Initialization particle position pop, i.e., the initial value of optimal penalty factor c and kernel functional parameter g initializes particle rapidity
V, calculates the n error rate of the folding cross validation method of grader ten under original state, takes maximum therein and saves as
Fitness, the meaning that maximum is taken herein is, in order to find the c and g of the equal very little of error rate for causing n grader, therefore will
Maximum is used as Judging index in the n error rate for iterating to calculate every time；
（5）By successive ignition, speed and position to particle are constantly updated, the fitness in the case of calculating respectively, and will
The minimum value fitness occurred in iterative process is stored in result, and so far, can obtain makes the classification of n grader poor
The optimal solution c and g of the equal very little of error rate.
It should be noted that it is in order that specification is more succinct straight that the present invention is explained using two-dimensional feature vector
See, do not represent characteristic vector involved in the present invention and be confined to two-dimensional feature vector.
For step（E）, the unknown failure characteristic vector of actual measurement is input to the SVMs of training, carry out failure classes
Type identification, realizes that Mechanical Failure of HV Circuit Breaker is classified, and completes self study, a specific embodiment, as shown in figure 3,
（1）The emulation data of n class different conditions are stored under n different file respectively, are easy to distinguish, in matlab
Spot file first, then extracts wherein all of csv files, imports Wave data, by previously described waveform interception,
Mathematical morphology filter and feature extraction, characteristic quantity is saved as according to status categories respectively data1, data2,
Data3 ..., data (n), be then merged into a matrix and conveniently call, data=[data1;data2;
data3;……;data(n)]；
（2）Be that n classes state creates class formative, such as theclass1=1, theclass2=2 etc., this class formative and feature to
The classification of amount is corresponded, and quantity and storage order both of which are consistent, with 1）It is similar, class formative is also merged into one
Matrix is conveniently called,
theclass=[theclass1;theclass2;theclass3……theclass(n)]；
（3）C and g parameters are obtained by population optimizing algorithm, based on one-to-many method, n grader is trained；
（4）Plane where characteristic vector is generated into dot matrix by step-length d, Classification and Identification is carried out using n grader to each point,
N assessment fraction is provided based on each point to the distance of decision surface, the maximum classification of fraction is the classification belonging to it, by a point good class
Point carry out induction-arrangement, draw out all kinds of confidential intervals；
（5）By the assessment score normalization of all points for being identified as the i-th class, using formula（1）, Value=(Value-min
(Value))/(max (Value)-min (Value)), obtains the confidence level in the i-th class state confidential interval；
（6）The unknown state data for measuring will be tested carries out Classification and Identification as input, exports its generic and confidence level；
（7）If by finding that this characteristic vector is not belonging to known all kinds of states in grader after identification, by the experiment after identification
Data update sample set and group indication, data=[data as sample;Data (n+1)], theclass=[theclass;
Theclass (n+1)], and with this Training Support Vector Machines, the confidential interval and confidence level of all kinds of states are regained, complete to divide
The self study of class device.
In sum, the primary cut-out recognition methods based on SVMs and Dynamics Simulation of the invention,
Using Dynamics Simulation and the grader based on SVMs, the accurate knowledge of the mechanical failure of primary cut-out is realized
Not, compared with the method for traditional experiment, without carrying out a large amount of physical simulation experiments, and with self-learning function, constantly improve point
Class device, for the Fault Identification of primary cut-out, can accurately identify, safe and reliable without transforming device structure, greatly
Big to reduce physical simulation experiment workload, economy is convenient, and grader possesses self-learning function, and more economical, convenient, safety has
Wide application prospect.
General principle of the invention, principal character and advantage has been shown and described above.The technical staff of the industry should
Understand, the present invention is not limited to the above embodiments, simply original of the invention is illustrated described in above-described embodiment and specification
Reason, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes and improvements
All fall within the protetion scope of the claimed invention.The claimed scope of the invention is by appending claims and its equivalent circle.
It is fixed.
Claims (7)
1. the primary cut-out recognition methods based on SVMs and Dynamics Simulation, it is characterised in that：Including following
Step,
The threedimensional model of primary cut-out is set up, the mechanical property waveform of threedimensional model is obtained by Dynamics Simulation, and
Contrasted with experimental test data by by simulation result, calibrated and verified threedimensional model；
By changing size, material, the characterisitic parameter of threedimensional model, based on each of Dynamics Simulation high voltage breaker simulator
Failure is planted, corresponding various mechanical property waveforms are obtained；
Waveform interception, mathematical morphology filter, feature extraction treatment are carried out to the corresponding mechanical property waveform of various failures, is obtained
Characteristic vector sample set；
Based on SVMs and its one-to-many sorting technique, grader is trained, and calculate the confidence level of each classification；
The unknown failure characteristic vector of actual measurement is input to the SVMs of training, fault type recognition is carried out, high pressure is realized
Breaker mechanical failure modes, and complete self study.
2. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：The process of the threedimensional model of primary cut-out is set up, is comprised the following steps,
（1）The major loop and each parts of operating mechanism of primary cut-out are drawn by 3D sculpting software；
（2）Real work situation according to primary cut-out is assembled.
3. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：Obtain the mechanical property waveform of threedimensional model by Dynamics Simulation, and by by simulation result and examination
Test test data to be contrasted, calibrate and verify threedimensional model, comprise the following steps,
（1）Simulation calculation is carried out to threedimensional model using multi-body Dynamic Analysis software, according to equipment actual conditions to three-dimensional mould
Type adds each type kinematic pair, defines divide-shut brake spring, oil bumper characterisitic parameter, and close to all parts additions of threedimensional model
Degree parameter, obtains the mechanical characteristic of high-voltage circuit breaker simulation result under normal condition；
（2）Surveyed with the experiment to actual high-voltage breaker by by the mechanical characteristic of high-voltage circuit breaker simulation result under normal condition
Examination data are contrasted, and calibrate the simulation parameter of threedimensional model.
4. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：Tripping spring, the pre compressed magnitude of switching-in spring, stiffness coefficient in change primary cut-out is passed sequentially through to distinguish
Simulation point, switching-in spring fatigue；Oil bumper failure is simulated by changing the damped coefficient of oil bumper in primary cut-out；
It is not in place to simulate energy storage by changing primary cut-out handle ratchet wheel position；For each above-mentioned each emulation mode, emulation is set
1% undulating value is given during parameter, the dispersivenesses for simulating different group data under same emulation mode.
5. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：Waveform interception, mathematical morphology filter are carried out to the corresponding mechanical property waveform of various failures, at feature extraction
Reason, obtains characteristic vector sample set, comprises the following steps,
（1）Stroke curve to High Voltage Circuit Breaker Contacts is intercepted, and reduces data volume, is with the point that first stroke is not zero
Beginning flag, is maintained between 199.5 ~ 200.5mm with stroke, and reaches 20 data points for end mark carries out waveform section
Take；
（2）Treatment is filtered to the stroke curve of contact using Mathematical Morphology Method, i.e., it is special to pending machinery successively
Property waveform be opened and closed computing and make and break computing, and the result of opening and closing operation and make and break calculation process is averaged, be to ensure
The accuracy that follow-up maximum is obtained, the curve to averaged carries out interpolation processing；
（3）Feature extraction is carried out to doing the mechanical property waveform after difference treatment, its maximum and its corresponding transverse and longitudinal coordinate is taken
As two-dimensional feature vector, so as to form characteristic vector sample set.
6. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：Based on SVMs and its one-to-many sorting technique, training grader, and the confidence level of each classification is calculated,
Comprise the following steps,
（1）Based on one-to-many sorting technique, will be extended to suitable for the SVMs of two classification can carry out polytypic point
Class device, SVMs Selection of kernel function RBF, penalty factor c and kernel functional parameter g pass through particle swarm optimization algorithm
Obtain, the purpose for finding optimal penalty factor and kernel functional parameter is to make classification error rate minimum, and wherein classification error rate is adopted
Calculated with ten folding cross validation methods, training obtains the grader based on SVMs；
（2）Using sample training（1）The grader based on SVMs for obtaining, generates the decision hyperplane of each classification；
（3）The region comprising all characteristic values is taken, mesh generation is carried out, dot matrix is generated, lattice spacing selection is less than 0.1mm*
0.1mm；
（4）Utilize（2）In have each classification decision hyperplane grader pair（3）Each point in dot matrix carries out Classification and Identification,
Based on each point and the relation of each categorised decision hyperplane, assessment fraction is given respectively for each categorised decision hyperplane, and
The point is defined as assessment fraction highest that class, if respectively assessment fraction is below the threshold value of setting, then it is assumed that the point is not for
Know classification；
（5）According to（4）, every kind of classification is all included corresponding dot matrix, the assessment a little of the included institute of every kind of classification is divided
Number does normalized, and each point assesses the normalization computing formula such as formula of fraction（1）Shown, the assessment fraction after treatment is characterized and divided
The relative reliability of class result；
Value=(Value-min(Value))/(max(Value)-min(Value)) （1）
Wherein, Value is assessment fraction；Max (Value) is place classification assessment fraction maximum；Min (Value) is place
Classification assessment fraction minimum value.
7. the primary cut-out recognition methods based on SVMs and Dynamics Simulation according to claim 1,
It is characterized in that：The unknown failure characteristic vector of actual measurement is input to the SVMs of training, fault type recognition is carried out, it is real
Existing Mechanical Failure of HV Circuit Breaker classification, and self study is completed, comprise the following steps,
（1）Testing mechanical characteristic is carried out to primary cut-out, the mechanical property ripple tested under the unknown state for obtaining in experiment is taken
Shape carries out mathematical morphology filter and waveform interception, and carries out characteristic vector pickup；
（2）The characteristic vector extracted is identified by grader, the region according to corresponding to this feature vector is carried out to it
Classification, and provide confidence level target；
（3）If by finding that this characteristic vector is not belonging to known all kinds of states in grader after identification, being regarded as new
Increasing state, and perform（4）；
（4）After characteristic vector is identified into classification, then it is used as sample grader is trained, and to decision hyperplane
Delimitation make amendment, and relative reliability to each classification makes amendment so that grader completes self study.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710146696.2A CN106934157B (en) | 2017-03-13 | 2017-03-13 | High-voltage circuitbreaker recognition methods based on support vector machines and Dynamics Simulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710146696.2A CN106934157B (en) | 2017-03-13 | 2017-03-13 | High-voltage circuitbreaker recognition methods based on support vector machines and Dynamics Simulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106934157A true CN106934157A (en) | 2017-07-07 |
| CN106934157B CN106934157B (en) | 2018-08-03 |
Family
ID=59432220
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710146696.2A Active CN106934157B (en) | 2017-03-13 | 2017-03-13 | High-voltage circuitbreaker recognition methods based on support vector machines and Dynamics Simulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106934157B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633114A (en) * | 2017-08-21 | 2018-01-26 | 江苏省电力试验研究院有限公司 | A kind of primary cut-out latency Mechanical Fault Vibration Signals emulation mode and device |
| CN108133096A (en) * | 2017-12-20 | 2018-06-08 | 大连交通大学 | Hitch dynamic reliability Forecasting Methodology based on MBD and SVM |
| CN108267687A (en) * | 2018-01-02 | 2018-07-10 | 国网江苏省电力有限公司电力科学研究院 | Based on probability density and the Mechanical Failure of HV Circuit Breaker diagnostic method being locally linear embedding into |
| CN108318810A (en) * | 2017-12-27 | 2018-07-24 | 国网北京市电力公司 | High voltage isolator fault determination method and device |
| CN109524972A (en) * | 2018-10-10 | 2019-03-26 | 华南理工大学 | Low-frequency oscillation method for parameter estimation based on GSO and SVM algorithm |
| CN109726467A (en) * | 2018-12-26 | 2019-05-07 | 哈尔滨工业大学 | Quick calculation method suitable for interaction force between part each in breaker mechanism |
| CN110703075A (en) * | 2019-09-10 | 2020-01-17 | 国网浙江省电力有限公司嘉兴供电公司 | High-voltage circuit breaker quick overhaul method based on big data technology |
| CN111308336A (en) * | 2020-03-24 | 2020-06-19 | 广西电网有限责任公司电力科学研究院 | High-voltage circuit breaker fast overhaul method and device based on big data |
| CN112698194A (en) * | 2020-12-10 | 2021-04-23 | 云南电网有限责任公司保山供电局 | Comprehensive evaluation method and system for state of circuit breaker operating mechanism |
| CN114527733A (en) * | 2022-02-23 | 2022-05-24 | 上海电气集团股份有限公司 | Simulation test method and device for electrical control system |
| CN114881110A (en) * | 2022-04-02 | 2022-08-09 | 西安交通大学 | Real-time detection method for total pressure change mode in on-orbit spacecraft cabin |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105718958A (en) * | 2016-01-27 | 2016-06-29 | 江苏省电力公司检修分公司 | Breaker fault diagnosis method based on linear discriminant analysis and support vector machine |
-
2017
- 2017-03-13 CN CN201710146696.2A patent/CN106934157B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105718958A (en) * | 2016-01-27 | 2016-06-29 | 江苏省电力公司检修分公司 | Breaker fault diagnosis method based on linear discriminant analysis and support vector machine |
Non-Patent Citations (3)
| Title |
|---|
| XIAOHUA WANG, WU YANG, MINGZHE RONG, SHENLI JIA, YONGHONG CHENG: "《Simulation and Analysis of High Voltage Circuit Breaker"s Mechanism Dynamical Characteristic》", 《TRANSMISSION AND DISTRIBUTION CONFERENCE AND EXHIBITION 2002:ASIA PACIFIC.IEEE/PES》 * |
| 杨景刚，刘媛，高山，陈曦: "《高压断路器机械故障诊断技术》", 《江苏电机工程》 * |
| 郝爽，仲林林，王小华，李高扬，荣命哲: "《基于支持向量机的高压断路器机械状态预测算法研究》", 《高压电器》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107633114B (en) * | 2017-08-21 | 2018-08-14 | 江苏省电力试验研究院有限公司 | A kind of high-voltage circuitbreaker latency Mechanical Fault Vibration Signals emulation mode and device |
| CN107633114A (en) * | 2017-08-21 | 2018-01-26 | 江苏省电力试验研究院有限公司 | A kind of primary cut-out latency Mechanical Fault Vibration Signals emulation mode and device |
| CN108133096A (en) * | 2017-12-20 | 2018-06-08 | 大连交通大学 | Hitch dynamic reliability Forecasting Methodology based on MBD and SVM |
| CN108133096B (en) * | 2017-12-20 | 2021-04-09 | 大连交通大学 | Coupler dynamic reliability prediction method based on MBD and SVM |
| CN108318810B (en) * | 2017-12-27 | 2021-07-20 | 国网北京市电力公司 | High-voltage isolating switch fault determination method and device |
| CN108318810A (en) * | 2017-12-27 | 2018-07-24 | 国网北京市电力公司 | High voltage isolator fault determination method and device |
| CN108267687A (en) * | 2018-01-02 | 2018-07-10 | 国网江苏省电力有限公司电力科学研究院 | Based on probability density and the Mechanical Failure of HV Circuit Breaker diagnostic method being locally linear embedding into |
| CN108267687B (en) * | 2018-01-02 | 2019-03-15 | 国网江苏省电力有限公司电力科学研究院 | Based on probability density and the Mechanical Failure of HV Circuit Breaker diagnostic method being locally linear embedding into |
| CN109524972A (en) * | 2018-10-10 | 2019-03-26 | 华南理工大学 | Low-frequency oscillation method for parameter estimation based on GSO and SVM algorithm |
| CN109524972B (en) * | 2018-10-10 | 2022-03-29 | 华南理工大学 | Low-frequency oscillation parameter estimation method based on GSO and SVM algorithms |
| CN109726467A (en) * | 2018-12-26 | 2019-05-07 | 哈尔滨工业大学 | Quick calculation method suitable for interaction force between part each in breaker mechanism |
| CN109726467B (en) * | 2018-12-26 | 2021-12-31 | 哈尔滨工业大学 | Rapid calculation method suitable for interaction force between parts in circuit breaker mechanism |
| CN110703075B (en) * | 2019-09-10 | 2022-03-15 | 国网浙江省电力有限公司嘉兴供电公司 | High-voltage circuit breaker quick overhaul method based on big data technology |
| CN110703075A (en) * | 2019-09-10 | 2020-01-17 | 国网浙江省电力有限公司嘉兴供电公司 | High-voltage circuit breaker quick overhaul method based on big data technology |
| CN111308336A (en) * | 2020-03-24 | 2020-06-19 | 广西电网有限责任公司电力科学研究院 | High-voltage circuit breaker fast overhaul method and device based on big data |
| CN112698194A (en) * | 2020-12-10 | 2021-04-23 | 云南电网有限责任公司保山供电局 | Comprehensive evaluation method and system for state of circuit breaker operating mechanism |
| CN114527733A (en) * | 2022-02-23 | 2022-05-24 | 上海电气集团股份有限公司 | Simulation test method and device for electrical control system |
| CN114881110A (en) * | 2022-04-02 | 2022-08-09 | 西安交通大学 | Real-time detection method for total pressure change mode in on-orbit spacecraft cabin |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106934157B (en) | 2018-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106934157A (en) | Primary cut-out recognition methods based on SVMs and Dynamics Simulation | |
| CN104712542B (en) | A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults | |
| CN104598813B (en) | Computer intrusion detection method based on integrated study and semi-supervised SVM | |
| CN103914064B (en) | Based on the commercial run method for diagnosing faults that multi-categorizer and D-S evidence merge | |
| CN101587155B (en) | Oil soaked transformer fault diagnosis method | |
| CN105046279A (en) | Analog circuit fault mode classification method | |
| CN105129109B (en) | Method for evaluating health of aircraft aileron actuator system based on multi-fractal theory and self-organizing map (SOM) network | |
| CN101221213A (en) | Analogue circuit fault diagnosis neural network method based on particle swarm algorithm | |
| CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
| CN108363896B (en) | Fault diagnosis method for hydraulic cylinder | |
| CN106323635A (en) | Rolling bearing fault on-line detection and state assessment method | |
| CN105653450A (en) | Software defect data feature selection method based on combination of modified genetic algorithm and Adaboost | |
| CN108664690A (en) | Long-life electron device reliability lifetime estimation method under more stress based on depth belief network | |
| CN103336243A (en) | Breaker fault diagnosis method based on separating/closing coil current signals | |
| CN104596780B (en) | Diagnosis method for sensor faults of motor train unit braking system | |
| CN106154163A (en) | Battery life state identification method | |
| CN103150454A (en) | Dynamic machine learning modeling method based on sample recommending and labeling | |
| CN105606914A (en) | IWO-ELM-based Aviation power converter fault diagnosis method | |
| CN107132310A (en) | Transformer equipment health status method of discrimination based on gauss hybrid models | |
| CN102842043B (en) | Particle swarm classifying method based on automatic clustering | |
| CN105678343A (en) | Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set | |
| CN105572572A (en) | WKNN-LSSVM-based analog circuit fault diagnosis method | |
| CN107957946A (en) | Software Defects Predict Methods based on neighborhood insertion protection algorism support vector machines | |
| CN110298374A (en) | A kind of driving locus energy consumption analysis method and apparatus based on deep learning | |
| Zhang et al. | Automatic generation of test cases based on multi-population genetic 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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: PA Weier road Jiangning District of Nanjing City, Jiangsu Province, No. 1 211103 Applicant after: STATE GRID JIANGSU ELECTRIC POWER COMPANY Research Institute Applicant after: State Grid Corporation of China Applicant after: XI'AN JIAOTONG University Address before: PA Weier road Jiangning District of Nanjing City, Jiangsu Province, No. 1 211103 Applicant before: STATE GRID JIANGSU ELECTRIC POWER COMPANY Research Institute Applicant before: State Grid Corporation of China Applicant before: Xi'an Jiaotong University |
|
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20180306 Address after: No. 243, Fenglou District, Nanjing, Jiangsu, Jiangsu Applicant after: JIANGSU ELECTRIC POWER RESEARCH INSTITUTE Co.,Ltd. Applicant after: STATE GRID JIANGSU ELECTRIC POWER COMPANY Research Institute Applicant after: State Grid Corporation of China Applicant after: XI'AN JIAOTONG University Address before: PA Weier road Jiangning District of Nanjing City, Jiangsu Province, No. 1 211103 Applicant before: STATE GRID JIANGSU ELECTRIC POWER COMPANY Research Institute Applicant before: State Grid Corporation of China Applicant before: Xi'an Jiaotong University |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |