CN108615053A - Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing - Google Patents
Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing Download PDFInfo
- Publication number
- CN108615053A CN108615053A CN201810339547.2A CN201810339547A CN108615053A CN 108615053 A CN108615053 A CN 108615053A CN 201810339547 A CN201810339547 A CN 201810339547A CN 108615053 A CN108615053 A CN 108615053A
- Authority
- CN
- China
- Prior art keywords
- failure
- optimal
- signal
- entropy
- svm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—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
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/28—Testing of electronic circuits, e.g. by signal tracer
- G01R31/316—Testing of analog circuits
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Abstract
The invention discloses the manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing, mainly comprise the following steps:With the failure of software Simulation Diagnosis object;For each failure Monte Carlo analysis in circuit, the characteristic signal of failure is detected, is decomposed fault-signal with wavelet packet, makes signal decomposition that there is maximum regularity based on optimal wavelet entropy principle, extracts the characteristic value that every group of signal optimal energy value is failure;To using particle cluster algorithm before failure modes, between in the supporting vector in view of being spaced sample data class weight parameter and punishment parameter carry out parameter optimization, make the optimal hyperlane of SVM that there is better classifying quality, improve the accuracy of fault diagnosis.
Description
Technical field
The invention belongs to Analogical Circuit Technique fields, are a kind of manifold SVM analog circuit faults based on particle group optimizing
Diagnostic method is mainly used in the fault diagnosis of analog circuit, is mainly based upon wavelet packet optimal energy entropy and manifold structure, and
SVM (Support Vector Machine, support vector machines) mould that parameter optimization is crossed with weight modified particle swarm optiziation
Quasi- circuit fault diagnosis.
Background technology
The arrival in big data epoch makes industrial intelligentization grow rapidly, and analog circuit also appears in every field therewith,
The diagnosis broken down for circuit also just becomes particularly important.But analog circuit complexity is various, fault parameter is continuous, first device
The many factors such as part tolerance limit the development of its diagnostic techniques.The universal machine learning algorithm that promotes of artificial intelligence concept
Development, more scholars are applied to the fault diagnosis of analog circuit.Xin Jian et al. proposes that wavelet transformation is mutually tied with neural network
The method of conjunction, but neural network Shortcomings itself, training sample is more, there is the problems such as study, poor robustness.What is supported the army,
Zeng Wenying et al. proposes comentropy algorithm of support vector machine and carrys out the failure of diagnostic sensor circuit, although improving extensive energy
Power shortens Diagnostic Time, but has ignored the structure feature inside data, can not obtain accurate diagnostic message, affect
The classification performance of SVM.This paper presents the manifold SVM based on particle group optimizing, the prior information of fused data distributed architecture, together
When particle cluster algorithm is combined with support vector machines, parameter therein is optimized, the diagnosis of SVM can be effectively enhanced
Effect.
Invention content
In view of the drawbacks of the prior art, the present invention proposes the manifold SVM analog circuit faults based on particle group optimizing and examines
It is disconnected, it faster can more accurately be diagnosed to be the failure of analog circuit appearance.Include the following steps:
1) failure of software Simulation Diagnosis object is used, and measures the accuracy rate of classification failure;
2) for each failure Monte Carlo analysis in circuit, the characteristic signal of failure is detected, with wavelet packet point
Circuit alarm is analysed, makes signal decomposition that there is maximum regularity based on optimal wavelet entropy principle, extracts every group of optimal energy of signal
Magnitude is the characteristic value of failure;
3) when to failure modes, the SVM in view of manifold structure inside data is selected, reuses the PSO calculations for improving weight
Method carries out optimizing to the parameter in support vector machines, and a part of data are finally carried out sample training, and remainder is differentiated
Analysis, to obtain the accuracy of classification.
Used software is orcad and Pspice in 1).
It is for the step of 2) middle circuit alarm extracted:
Measured signal is carried out 3 layers of WAVELET PACKET DECOMPOSITION by (2-1), can be derived that the WAVELET PACKET DECOMPOSITION coefficient of 8 frequency bands
(2-2) builds optimal wavelet tree on this basis, is based primarily upon optimal energy entropy:
A. the Energy-Entropy of each node in wavelet tree is calculated;
B. compare from the subspace of minimum one layer of WAVELET PACKET DECOMPOSITION, if the entropy of two sub-spaces is less than generating space entropy, protect
It stays the two subspaces, the sum of entropy of the two subspaces of the entropy of generating space to replace, otherwise retains generating space, give up sub- sky
Between;
C. the reconstruct of the coefficient of remaining optimal energy entropy finally can be obtained into optimal wavelet tree, then calculates the energy value of the signal,
Characteristic value as fault-signal;
When in 3) to failure modes, when constructing SVM, mainly the manifold structure between sample data is taken into account, it should
Method and step is as follows:
A weighting function 3-1) is constructed between homogeneous datas portrays the manifold local geometry of sample data,
Weighting function is:
3-2) set the interior class similarity matrix of sample as:S=ω (W'-W) ω, whereinIt is similar
Weighting function.
Within-cluster variance M=μ S+ (1- μ) S based on manifold structure can 3-3) be obtained according to thisw, for original optimal
Change problem can be described as;
S.t yi(WTXi+b)≥1-ξiI=1,2...l
In 3-3) in when converting primal problem to the optimization problem under inequality constraints, there are weight parameter β and punish
Penalty parameter C carries out optimizing using improved PSO algorithms to parameter therein, and its step are as follows:
(1) data are obtained from sample set, generate primary and establish particle populations;
(2) particle populations of acquisition are initialized, sets some parameters in formula;
(3) Selection of Function:For fitness function, the fitness value of each particle, wherein y are calculatedi
WithRespectively represent actual value and test value in data set;
(4) by fitness function calculate particle each fitness value, and be iterated according to this more new individual it is optimal with
Global optimum position.
(5) if iterations are more than the maximum times T of settingmaxOr final result be less than accuracy value when, iteration knot
Beam exports optimal solution, no to then follow the steps (3).
It is remaining for dividing by a measured characteristic part for training in the failure modes in carrying out step 4
Class differentiates, measures the accuracy of classification.
Advantageous effect
The energy value of the optimal wavelet tree reconstructed based on optimal energy entropy can more be showed and be out of order as feature vector
The feature of signal.In SVM discriminant analyses, the manifold structure of data is considered, construct the weight letter between sample data and class
Number, at the same using improve after PSO algorithms, in SVM weight parameter and punishment parameter carry out optimizing so that classifying quality
It is faster and better.
Description of the drawings
Fig. 1 is selected diagnostic circuit;
The flow chart of diagnostic circuits of the Fig. 2 based on this method.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.Fig. 2 is the flow chart based on this method diagnostic circuit.
Bandpass filter is chosen as diagnosis object, circuit diagram such as attached drawing 1 draws electricity in OrCAD/PSpice softwares
Road, element have tolerance, set up 5% that resistance tolerance is nominal value, inductance 10%.Set up fault mode be respectively C1 ↑, C2
↑, C3 ↑, C4 ↑, R1 ↑, R2 ↓, R3 ↑, R4 ↓, R5 ↑, R6 ↓, R7 ↑, R8 ↓, wherein ↑ indicate that the value is the failure more than nominal value 50%
Value, ↓ for less than nominal value 50% fault value.Thus have 12 groups of fault values, in addition normal condition, one share 13 in state.
200 analyses are carried out to each malfunction with Monte Carlo analysis, then wavelet packet analysis is carried out to signal with matlab,
Optimal wavelet tree is reconstructed, using their optimal energy entropy as characteristic value, composition characteristic vector.Algorithm steps are as follows:
A. the Energy-Entropy of each node in wavelet tree is calculatedWherein eijFor the opposite of each frequency band
Energy
B. compare from the subspace of minimum one layer of WAVELET PACKET DECOMPOSITION, if the entropy of two sub-spaces is less than generating space entropy, i.e.,
Ent(N) > Ent(N1)+Ent(N2), the two subspaces are stayed, the sum of entropy of the two subspaces of the entropy of generating space replaces, on the contrary
Retain generating space, gives up subspace
C. the reconstruct of the coefficient of remaining optimal energy entropy finally can be obtained into optimal wavelet tree, then calculates the energy value of the signal,
Characteristic value T=[E as fault-signal0,E1…En]。
In failure modes, consider that the manifold structure of data makes the optimal hyperlane of SVM, improved PSO algorithms pair are used in combination
The weight parameter and punishment parameter of SVM carries out optimizing, and its step are as follows:
D. a weighting function is constructed between homogeneous data to portray the manifold local geometry of sample data, weight
Function is:
E. set the interior class similarity matrix of sample as:S=ω (W'-W) ω, wherein, WijFor similar power
Weight function.
F. within-cluster variance M=μ S+ (1- μ) S based on manifold structure can be obtained according to thisw, for original optimization
Problem can be described as;
G. when carrying out optimizing to the parameter of SVM using improved PSO algorithms the step of is as follows:
(1) data are obtained from sample set, generate primary and establish particle populations;
(2) particle populations of acquisition are initialized, sets some parameters in formula;
(3) Selection of Function:For fitness function, the fitness value of each particle, wherein y are calculatedi
WithRespectively represent actual value and test value in data set;
(4) by fitness function calculate particle each fitness value, and be iterated according to this more new individual it is optimal with
Global optimum position.
(5) if iterations are more than the maximum times T of settingmaxOr final result be less than accuracy value when, iteration knot
Beam exports optimal solution, no to then follow the steps (3).
Discriminant classification finally is carried out using remaining data, measures the accuracy of classification.
Claims (6)
1. the manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing, which is characterized in that include the following steps:
1) failure of software Simulation Diagnosis object is used, and measures the accuracy rate of classification failure;
2) for each failure Monte Carlo analysis in circuit, the characteristic signal of failure is detected, with wavelet packet analysis electricity
Road fault-signal makes signal decomposition have maximum regularity, extracts every group of signal optimal energy value based on optimal wavelet entropy principle
For the characteristic value of failure;
3) when to failure modes, the SVM in view of manifold structure inside data is selected, reuses the PSO algorithms pair for improving weight
Parameter in support vector machines carries out optimizing, and a part of data are finally carried out sample training, and remainder carries out discriminant analysis,
To obtain the accuracy of classification.
2. the method as described in claim 1, which is characterized in that in step 1) used software be orcad and
Pspice。
3. the method as described in claim 1, which is characterized in that the step of circuit alarm of extraction is in step 2):
Measured signal is carried out 3 layers of WAVELET PACKET DECOMPOSITION by (2-1), can be derived that the WAVELET PACKET DECOMPOSITION coefficient of 8 frequency bands;
(2-2) builds optimal wavelet tree on this basis, is based primarily upon optimal energy entropy:
A. the Energy-Entropy of each node in wavelet tree is calculated;
B. compare from the subspace of minimum one layer of WAVELET PACKET DECOMPOSITION, if the entropy of two sub-spaces is less than generating space entropy, retain this
Two sub-spaces, the sum of entropy of the two subspaces of the entropy of generating space replace, otherwise retain generating space, give up subspace;
C. the reconstruct of the coefficient of remaining optimal energy entropy finally can be obtained into optimal wavelet tree, then calculates the energy value of the signal, as
The characteristic value of fault-signal.
4. the method as described in claim 1, which is characterized in that when in step 3) to failure modes, when constructing SVM, by sample
Manifold structure between notebook data is taken into account, this is as follows:
A weighting function 3-1) is constructed between homogeneous datas to portray the manifold local geometry of sample data, weight
Function is:
3-2) set the interior class similarity matrix of sample as:S=ω (W'-W) ω, whereinWijFor similar weight
Function;
Within-cluster variance M=μ S+ (1- μ) S based on manifold structure can 3-3) be obtained according to thisw, for original optimization problem
It can be described as;
S.t yi(WTXi+b)≥1-ξiI=1,2...l.
5. method as claimed in claim 4, which is characterized in that in 3-3) in convert under inequality constraints primal problem to
When optimization problem, there are weight parameter β and punishment parameter C, and optimizing is carried out to parameter therein using improved PSO algorithms,
Its step are as follows:
(1) data are obtained from sample set, generate primary and establish particle populations;
(2) particle populations of acquisition are initialized, sets some parameters in formula;
(3) Selection of Function:For fitness function, the fitness value of each particle, wherein y are calculatediWithPoint
Actual value and test value in data set are not represented;
(4) by fitness function calculate particle each fitness value, and be iterated according to this more new individual it is optimal with it is global
Optimal location;
(5) if iterations are more than the maximum times T of settingmaxOr final result be less than accuracy value when, iteration terminates, will
Optimal solution exports, no to then follow the steps (3).
6. the method as described in claim 1, which is characterized in that in the failure modes in carrying out step 3), by measured spy
A sign data part is remaining to be used for discriminant classification for training, and measures the accuracy of classification.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810339547.2A CN108615053A (en) | 2018-04-16 | 2018-04-16 | Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810339547.2A CN108615053A (en) | 2018-04-16 | 2018-04-16 | Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108615053A true CN108615053A (en) | 2018-10-02 |
Family
ID=63660255
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810339547.2A Pending CN108615053A (en) | 2018-04-16 | 2018-04-16 | Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108615053A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110175682A (en) * | 2019-04-22 | 2019-08-27 | 广东技术师范大学 | A kind of optimization core pivot element analysis fault monitoring method based on Chaos-Particle Swarm Optimization |
CN110210580A (en) * | 2019-07-01 | 2019-09-06 | 桂林电子科技大学 | Analog-circuit fault diagnosis method based on cloud evolution algorithm optimization SVM |
CN114325352A (en) * | 2022-01-04 | 2022-04-12 | 电子科技大学 | Analog filter circuit fault diagnosis method based on empirical wavelet transform |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017128455A1 (en) * | 2016-01-25 | 2017-08-03 | 合肥工业大学 | Analogue circuit fault diagnosis method based on generalized multiple kernel learning-support vector machine |
CN107202950A (en) * | 2017-05-12 | 2017-09-26 | 南京邮电大学 | Analog-circuit fault diagnosis method based on LMD approximate entropies and SVM |
CN107798343A (en) * | 2017-10-16 | 2018-03-13 | 南京邮电大学 | One kind is based on the improved SVM analog-circuit fault diagnosis methods of manifold structure |
-
2018
- 2018-04-16 CN CN201810339547.2A patent/CN108615053A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017128455A1 (en) * | 2016-01-25 | 2017-08-03 | 合肥工业大学 | Analogue circuit fault diagnosis method based on generalized multiple kernel learning-support vector machine |
CN107202950A (en) * | 2017-05-12 | 2017-09-26 | 南京邮电大学 | Analog-circuit fault diagnosis method based on LMD approximate entropies and SVM |
CN107798343A (en) * | 2017-10-16 | 2018-03-13 | 南京邮电大学 | One kind is based on the improved SVM analog-circuit fault diagnosis methods of manifold structure |
Non-Patent Citations (1)
Title |
---|
陈志凡等: "基于PSO-LSSVM 的模拟电路故障诊断", 《计算机技术与发展》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110175682A (en) * | 2019-04-22 | 2019-08-27 | 广东技术师范大学 | A kind of optimization core pivot element analysis fault monitoring method based on Chaos-Particle Swarm Optimization |
CN110210580A (en) * | 2019-07-01 | 2019-09-06 | 桂林电子科技大学 | Analog-circuit fault diagnosis method based on cloud evolution algorithm optimization SVM |
CN114325352A (en) * | 2022-01-04 | 2022-04-12 | 电子科技大学 | Analog filter circuit fault diagnosis method based on empirical wavelet transform |
CN114325352B (en) * | 2022-01-04 | 2023-04-18 | 电子科技大学 | Analog filter circuit fault diagnosis method based on empirical wavelet transform |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104523268B (en) | Electroencephalogram signal recognition fuzzy system and method with transfer learning ability | |
CN106886023B (en) | A kind of Radar Echo Extrapolation method based on dynamic convolutional neural networks | |
Huo et al. | Artificial neural network models for reference evapotranspiration in an arid area of northwest China | |
CN112395945A (en) | Graph volume behavior identification method and device based on skeletal joint points | |
CN108615053A (en) | Manifold SVM analog-circuit fault diagnosis methods based on particle group optimizing | |
CN107527352A (en) | Remote sensing Ship Target contours segmentation and detection method based on deep learning FCN networks | |
CN107423814A (en) | A kind of method that dynamic network model is established using depth convolutional neural networks | |
CN102938073B (en) | Method for classifying remote sensing images | |
CN110161343A (en) | A kind of non-intrusion type real-time dynamic monitoring method of intelligence train exterior power receiving device | |
CN104133143B (en) | A kind of Guangdong power system diagnostic system and method calculating platform based on Hadoop cloud | |
CN104849650B (en) | One kind is based on improved analog-circuit fault diagnosis method | |
CN110866448A (en) | Flutter signal analysis method based on convolutional neural network and short-time Fourier transform | |
CN104799852A (en) | Method for extracting movement imagination electroencephalogram characteristics based on ultralimit learning machine self encoding | |
CN106997373A (en) | A kind of link prediction method based on depth confidence network | |
CN107609667A (en) | Heating load forecasting method and system based on Box_cox conversion and UFCNN | |
Gao et al. | IMA health state evaluation using deep feature learning with quantum neural network | |
CN110446173A (en) | A kind of energy-efficient satellite-carried wireless Sensor Network data compression method | |
CN105228185A (en) | A kind of method for Fuzzy Redundancy node identities in identification communication network | |
Tiancheng et al. | Application of improved naive bayesian-CNN classification algorithm in sandstorm prediction in inner mongolia | |
CN107798343A (en) | One kind is based on the improved SVM analog-circuit fault diagnosis methods of manifold structure | |
Chandra et al. | Application of cooperative neuro-evolution of Elman recurrent networks for a two-dimensional cyclone track prediction for the South Pacific region | |
CN109861220B (en) | Method for constructing tensor input of deep convolutional neural network for power system analysis | |
Wang et al. | Remote sensing image classification based on artificial neural network: A case study of Honghe Wetlands National Nature Reserve | |
Sun et al. | Technology of structure damage monitoring based on multi-agent | |
Zhao et al. | Knowledge representation model for evolution of crowd evacuation stability |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181002 |
|
RJ01 | Rejection of invention patent application after publication |