CN109239585A - A kind of method for diagnosing faults based on the preferred wavelet packet of improvement - Google Patents
A kind of method for diagnosing faults based on the preferred wavelet packet of improvement Download PDFInfo
- Publication number
- CN109239585A CN109239585A CN201811038641.0A CN201811038641A CN109239585A CN 109239585 A CN109239585 A CN 109239585A CN 201811038641 A CN201811038641 A CN 201811038641A CN 109239585 A CN109239585 A CN 109239585A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- wavelet packet
- fault
- sample
- follows
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a kind of based on the decision tree SVM analog-circuit fault diagnosis method for improving preferred wavelet packet, includes the following steps: to extract the circuit response signal under each fault mode using circuit simulating software；The Optimum wavelet basic function for extracting fault signature is selected using improved preferred method of wavelet packet；Wavelet package transforms are carried out to original signal using the optimal wavelet basic function selected, form fault sample set of eigenvectors；The link structure between all fault feature vector collection is sought using minimum spanning tree；Optimal decision tree topology is found out using the distance that connects between group；Decision-tree model is trained according to topological tree Structure, finally trains the corresponding SVM classifier of each node using fault sample vector set.The present invention has good robustness, and Generalization Capability is stronger.
Description
Technical field
It is especially a kind of based on the fault diagnosis for improving preferred wavelet packet the present invention relates to Analog Circuit Fault Diagnosis Technology
Method.
Background technique
Analog circuit test and fault diagnosis play key effect in circuit design, equipment production and instrument maintenance,
It is current academic research person and the engineer challenging important topic in Electronic Testing field.It is fast with electronic technology
Speed development, the complexity and closeness of analog circuit constantly increase, propose to analog circuit operational reliability more stringent
It is required that.In analog circuit test and fault diagnosis, traditional Troubleshooting Theory and method not can be well solved because of element
Tolerance, the continuity of output response and the circuits build-in attribute such as non-linear and caused by failure diversity and complexity problem,
It is especially urgent that research suits practical, high-efficient high performance modern intelligent trouble diagnosis theory and method.
Analog circuit fault diagnosing includes three fault detection, fault identification and failure predication aspects.By nearly 20 years
Development becomes the hot spot of fault diagnosis field using artificial intelligence as the analog-circuit fault diagnosis method of theoretical basis, especially
Using neural network as the method for representative, due to it is stronger classification, identification and inferential capability, and processing Nonlinear Mapping
The advantages of with fault-tolerant aspect, becomes the model of analog circuit intelligent failure diagnosis method.Since this century, by wavelet analysis
The research that theoretical, biological evolution algorithm and information fusion technology etc. are applied to analog circuit fault diagnosing also starts to start to walk, and occurs
The new way that a variety of methods blend.The quality that fault signature is improved using wavelet analysis, in conjunction with fuzzy theory and calculation of evolving
The structure of method Optimum Classification device, it has also become the Main way of intelligent failure diagnosis method.
Method based on preferred wavelet packet is suggested in recent years, is widely used in the fault signature of analog circuit fault diagnosing
It extracts, the existing diagnostic method based on preferred wavelet packet, cannot preferable characteristic feature there are selection standard generalization is not strong
The inter-class separability of vector set estimates the problem of being difficult to pick out actually optimal optimal wavelet basic function.
Summary of the invention
The purpose of the present invention is to provide a kind of based on the decision tree SVM analog circuit fault diagnosing for improving preferred wavelet packet
Method improves circuit fault diagnosis in the validity for extracting fault signature sample link.
Realize the technical solution of the object of the invention are as follows: it is a kind of based on the method for diagnosing faults for improving preferred wavelet packet, including
Following steps:
Step 1, the circuit response signal under each fault mode is extracted using circuit simulating software；
Step 2, the preferred method of wavelet packet using improved based on Energy-Entropy, which is picked out, extracts the best small of fault signature
Wave basic function；
Step 3, wavelet package transforms are carried out to original signal using the optimal wavelet basic function selected, it is special forms fault sample
Levy vector set；
Step 4, the topological link structure between all fault feature vector collection is sought using minimum spanning tree；
Step 5, optimal decision tree topology is found out using the distance that connects between group, is then instructed according to topological tree Structure
Practise decision-tree model；
Step 6, the corresponding SVM classifier of each node is trained using fault sample vector set.
Further, preferred wavelet method of the modified used described in step 2 based on Energy-Entropy can be extracted preferably
Be conducive to increase the wavelet packet basis functions that inter-class separability is estimated out, specific as follows:
A given wavelet basis function is selected in wavelet packet basis functions library to be selected, and wavelet packet change is carried out to original sample
It changes, formula is as follows:
Wavelet packet coefficient recurrence formula is as follows:
The reconstruction formula of wavelet packet is as follows:
Wherein, h (k) and g (k) is respectively low-pass filtering coefficient and high-pass filtering coefficient in multiscale analysis,It indicates
K-th of coefficient corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, n), node (j, n) indicate n-th of frequency band of jth layer；
WAVELET PACKET DECOMPOSITION, definition test i-th of frequency range pair of signal jth layer are carried out to collected analog circuit fault signal
The energy answered are as follows:
In formula, N is the length of i-th of frequency band, di,j(k) it indicates corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, i) k-th
Wavelet packet coefficient.Using the energy value of each frequency band as fault signature, it is normalized, the energy after being normalized
Fault signature are as follows:
In formula,For each frequency band gross energy of jth layer after WAVELET PACKET DECOMPOSITION.
The main improvement of the present invention in step 2 is to measure using the viewpoint of Energy-Entropy separable between sample
Degree, i.e. inter-class separability are estimated.
If M sample is obtained after WAVELET PACKET DECOMPOSITION:
For any component, its Energy-Entropy is calculated:
Wherein, ETiFor the column and T of the i-th column of sample vector matrixkiFor the element of matrix k row i column, k-th of sample is represented
This ith feature component, WiFor the entropy of i-th of component, WEEFor average energy entropy, finally with WEEThe corresponding wavelet basis of minimum value
As best wavelet packet basis.
Further, step 5 use connection distance between group replace it is original only by center of gravity distance or divergence definition
Between class distance, it is specific as follows:
In formula, m and n represent two different failure classes, and a and b respectively represent the sample total number of m and n class, x representative sample
This.
Compared with prior art, the present invention its remarkable advantage are as follows: (1) present invention provides the principle of preferred wavelet packet basis functions
Possess more complete Fundamentals of Mathematics；(2) inter-class separability considered between sample is estimated；(3) more than single center of gravity distance
The good actual separation degree measured between sample class, it is of overall importance stronger.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow diagrams for improving preferred wavelet packet method for diagnosing faults.
Fig. 2 is WAVELET PACKET DECOMPOSITION schematic diagram.
Fig. 3 is minimal spanning tree algorithm flow diagram.
Fig. 4 is the schematic diagram of the experimental circuit sallen-key circuit in embodiment.
Specific embodiment
Present invention is further described in detail with specific embodiment with reference to the accompanying drawing.
In conjunction with Fig. 1, the present invention is based on the method for diagnosing faults for improving preferred wavelet packet, include the following steps:
Step 1, the circuit response signal under each fault mode is extracted using circuit simulating software；
Step 2, the preferred method of wavelet packet using improved based on Energy-Entropy, which is picked out, extracts the best small of fault signature
Wave basic function；
A given wavelet basis function is selected in wavelet packet basis functions library to be selected, and wavelet packet change is carried out to original sample
It changes, formula is as follows:
Wavelet packet coefficient recurrence formula is as follows:
The reconstruction formula of wavelet packet is as follows:
Wherein, h (k) and g (k) is respectively low-pass filtering coefficient and high-pass filtering coefficient in multiscale analysis,It indicates
K-th of coefficient corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, n), node (j, n) indicate n-th of frequency band of jth layer.
WAVELET PACKET DECOMPOSITION, definition test i-th of frequency range pair of signal jth layer are carried out to collected analog circuit fault signal
The energy answered are as follows:
In formula, N is the length of i-th of frequency band, di,j(k) it indicates corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, i) k-th
Wavelet packet coefficient.Using the energy value of each frequency band as fault signature, it is normalized, the energy after being normalized
Fault signature are as follows:
In formula,For each frequency band gross energy of jth layer after WAVELET PACKET DECOMPOSITION.
The main improvement of the present invention in step 2 is to measure using the viewpoint of Energy-Entropy separable between sample
Degree, i.e. inter-class separability are estimated.
If M sample is obtained after WAVELET PACKET DECOMPOSITION:
For any component, its Energy-Entropy is calculated:
Wherein, ETiFor the column and T of the i-th column of sample vector matrixkiFor the element of matrix k row i column, k-th of sample is represented
This ith feature component, WiFor the entropy of i-th of component, WEEFor average energy entropy, finally with WEEThe corresponding wavelet basis of minimum value
As best wavelet packet basis.
Step 3, wavelet package transforms are carried out to original signal using the optimal wavelet basic function selected, it is special forms fault sample
Levy vector set；
Step 4, the topological link structure between all fault feature vector collection is sought using minimum spanning tree；
Step 5, optimal decision tree topology is found out using the distance that connects between group, is then instructed according to topological tree Structure
Practise decision-tree model；
Distance is connected between calculating group, instead of original between class distance only defined by center of gravity distance or divergence, specifically
It is as follows:
In formula, m and n represent two different failure classes, and a and b respectively represent the sample total number of m and n class, x representative sample
This.
Step 6, decision-tree model is trained according to topological tree Structure, finally trains each using fault sample vector set
The corresponding SVM classifier of node.
Embodiment
Method for diagnosing faults proposed by the present invention based on the preferred wavelet packet of improvement, with low-pass filter as shown in Figure 4
It for sallen-key circuit, is verified, steps are as follows:
Step 1, sallen-key low-pass filter electricity as shown in Figure 4 is extracted using circuit simulating software Multisim12.0
Circuit response signal under the fault mode of 15 kinds of road under each fault mode；
Step 2, the preferred method of wavelet packet using improved based on Energy-Entropy, which is picked out, extracts the best small of fault signature
Wave basic function；
A given base is selected in wavelet packet basis functions library to be selected, and as shown in Figure 2 three layers small is carried out to original sample
Wave packet transform, formula are as follows:
Wavelet packet coefficient recurrence formula is as follows:
The reconstruction formula of wavelet packet is as follows:
Wherein, h (k) and g (k) is respectively low-pass filtering coefficient and high-pass filtering coefficient in multiscale analysis,It indicates
K-th of coefficient corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, n), node (j, n) indicate n-th of frequency band of jth layer.
WAVELET PACKET DECOMPOSITION, definition test i-th of frequency range pair of signal jth layer are carried out to collected analog circuit fault signal
The energy answered are as follows:
In formula, N is the length of i-th of frequency band, di,j(k) it indicates corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, i) k-th
Wavelet packet coefficient.Using the energy value of each frequency band as fault signature, it is normalized, the energy after being normalized
Fault signature are as follows:
In formula,For each frequency band gross energy of jth layer after WAVELET PACKET DECOMPOSITION.
The main improvement of the present invention in step 2 is to measure using the viewpoint of Energy-Entropy separable between sample
Degree, i.e. inter-class separability are estimated.
If M sample is obtained after WAVELET PACKET DECOMPOSITION:
For any component, its Energy-Entropy is calculated:
Wherein, ETiFor the column and T of the i-th column of sample vector matrixkiFor the element of matrix k row i column, k-th of sample is represented
This ith feature component, WiFor the entropy of i-th of component, WEEFor average energy entropy, finally with WEEThe corresponding wavelet basis of minimum value
As best wavelet packet basis.
Step 3, three layers of wavelet packet as shown in Figure 2 are carried out to original signal using the optimal wavelet basic function selected to become
It changes, forms fault sample set of eigenvectors；
Step 4, it is sought between all fault feature vector collection using minimal spanning tree algorithm as described in the flow chart of figure 3
Topological link structure；
Step 5, optimal decision tree topology is found out using the distance that connects between group, is then instructed according to topological tree Structure
Practise decision-tree model；
Distance is connected between calculating group, instead of original between class distance only defined by center of gravity distance or divergence, specifically
It is as follows:
In formula, m and n represent two different failure classes, and a and b respectively represent the sample total number of m and n class, x representative sample
This.
Step 6, decision-tree model is trained according to topological tree Structure, finally trains each using fault sample vector set
The corresponding SVM classifier of node.
The present invention is based on matlab language, test by taking sallen-key circuit as an example, the total diagnosis 95.1% of failure.
Claims (3)
1. a kind of based on the method for diagnosing faults for improving preferred wavelet packet, which comprises the steps of:
Step 1, the circuit response signal under each fault mode is extracted using circuit simulating software；
Step 2, the preferred method of wavelet packet using improved based on Energy-Entropy picks out the best wavelet for extracting fault signature
Function；
Step 3, using the optimal wavelet basic function selected to original signal carry out wavelet package transforms, formed fault sample feature to
Quantity set；
Step 4, the topological link structure between all fault feature vector collection is sought using minimum spanning tree；
Step 5, optimal decision tree topology is found out using the distance that connects between group, is then trained according to topological tree Structure
Decision-tree model；
Step 6, the corresponding SVM classifier of each node is trained using fault sample vector set.
2. according to claim 1 based on the method for diagnosing faults for improving preferred wavelet packet, which is characterized in that step 2 tool
Body are as follows:
A given wavelet basis function is selected in wavelet packet basis functions library to be selected, and wavelet package transforms are carried out to original sample, it is public
Formula is as follows:
Wavelet packet coefficient recurrence formula is as follows:
The reconstruction formula of wavelet packet is as follows:
Wherein, h (k) and g (k) is respectively low-pass filtering coefficient and high-pass filtering coefficient in multiscale analysis,Indicate small echo
Packet decomposes k-th of coefficient corresponding to posterior nodal point (j, n), and node (j, n) indicates n-th of frequency band of jth layer；
WAVELET PACKET DECOMPOSITION is carried out to collected analog circuit fault signal, definition test i-th of frequency range of signal jth layer is corresponding
Energy are as follows:
In formula, N is the length of i-th of frequency band, di,j(k) k-th of small echo corresponding to WAVELET PACKET DECOMPOSITION posterior nodal point (j, i) is indicated
Packet coefficient；Using the energy value of each frequency band as fault signature, it is normalized, the energy failure after being normalized
Feature are as follows:
In formula,For each frequency band gross energy of jth layer after WAVELET PACKET DECOMPOSITION；
If M sample is obtained after WAVELET PACKET DECOMPOSITION:
For any component, its Energy-Entropy is calculated:
Wherein, ETiFor the column and T of the i-th column of sample vector matrixkiFor the element of matrix k row i column, k-th of sample is represented
Ith feature component, WiFor the entropy of i-th of component, WEEFor average energy entropy, finally with WEEThe corresponding wavelet basis conduct of minimum value
Best wavelet packet basis.
3. according to claim 1 based on the method for diagnosing faults for improving preferred wavelet packet, which is characterized in that step 5 is adopted
Original between class distance only defined by center of gravity distance or divergence is replaced with distance is connected between group, specific as follows:
In formula, m and n represent two different failure classes, and a and b respectively represent the sample total number of m and n class, x representative sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811038641.0A CN109239585A (en) | 2018-09-06 | 2018-09-06 | A kind of method for diagnosing faults based on the preferred wavelet packet of improvement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811038641.0A CN109239585A (en) | 2018-09-06 | 2018-09-06 | A kind of method for diagnosing faults based on the preferred wavelet packet of improvement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109239585A true CN109239585A (en) | 2019-01-18 |
Family
ID=65060144
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811038641.0A Pending CN109239585A (en) | 2018-09-06 | 2018-09-06 | A kind of method for diagnosing faults based on the preferred wavelet packet of improvement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109239585A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058110A (en) * | 2019-04-16 | 2019-07-26 | 重庆大学 | A kind of active inverter intermittent fault diagnostic method |
| CN110414612A (en) * | 2019-07-31 | 2019-11-05 | 精英数智科技股份有限公司 | A kind of processing method, device, system and the medium of firedamp sensor fault data |
| WO2022033685A1 (en) | 2020-08-13 | 2022-02-17 | Siemens Aktiengesellschaft | Simulation-augmented decision tree analysis method, computer program product and system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130002262A1 (en) * | 2011-06-30 | 2013-01-03 | Denso Corporation | Vehicle-mounted system and insulation failure diagnosis unit |
| CN104008398A (en) * | 2014-06-11 | 2014-08-27 | 北京大学 | Gait classification method based on multi-sensor information fusion |
| CN105701470A (en) * | 2016-01-13 | 2016-06-22 | 合肥工业大学 | Analog circuit fault characteristic extraction method based on optimal wavelet packet decomposition |
| CN106156789A (en) * | 2016-05-09 | 2016-11-23 | 浙江师范大学 | Towards the validity feature sample identification techniques strengthening grader popularization performance |
| CN106437687A (en) * | 2016-08-31 | 2017-02-22 | 中国船舶重工集团公司第七八研究所 | While-drilling mud impulse signal noise reduction method based on optimal wavelet packet decomposition tree |
| CN107220670A (en) * | 2017-05-27 | 2017-09-29 | 重庆大学 | Supervised Artifical Taste system features extracting method is had based on wavelet transform |
| CN107894564A (en) * | 2017-11-09 | 2018-04-10 | 合肥工业大学 | A kind of analog-circuit fault diagnosis method based on intersection wavelet character |
-
2018
- 2018-09-06 CN CN201811038641.0A patent/CN109239585A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130002262A1 (en) * | 2011-06-30 | 2013-01-03 | Denso Corporation | Vehicle-mounted system and insulation failure diagnosis unit |
| CN104008398A (en) * | 2014-06-11 | 2014-08-27 | 北京大学 | Gait classification method based on multi-sensor information fusion |
| CN105701470A (en) * | 2016-01-13 | 2016-06-22 | 合肥工业大学 | Analog circuit fault characteristic extraction method based on optimal wavelet packet decomposition |
| CN106156789A (en) * | 2016-05-09 | 2016-11-23 | 浙江师范大学 | Towards the validity feature sample identification techniques strengthening grader popularization performance |
| CN106437687A (en) * | 2016-08-31 | 2017-02-22 | 中国船舶重工集团公司第七八研究所 | While-drilling mud impulse signal noise reduction method based on optimal wavelet packet decomposition tree |
| CN107220670A (en) * | 2017-05-27 | 2017-09-29 | 重庆大学 | Supervised Artifical Taste system features extracting method is had based on wavelet transform |
| CN107894564A (en) * | 2017-11-09 | 2018-04-10 | 合肥工业大学 | A kind of analog-circuit fault diagnosis method based on intersection wavelet character |
Non-Patent Citations (2)
| Title |
|---|
| 宋国明等: "最小生成树SVM的模拟电路故障诊断方法", 《电子科技大学学报》 * |
| 袁莉芬等: "基于小波包优选的模拟电路故障特征提取方法", 《电工技术学报》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058110A (en) * | 2019-04-16 | 2019-07-26 | 重庆大学 | A kind of active inverter intermittent fault diagnostic method |
| CN110414612A (en) * | 2019-07-31 | 2019-11-05 | 精英数智科技股份有限公司 | A kind of processing method, device, system and the medium of firedamp sensor fault data |
| WO2022033685A1 (en) | 2020-08-13 | 2022-02-17 | Siemens Aktiengesellschaft | Simulation-augmented decision tree analysis method, computer program product and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2015158198A1 (en) | Fault recognition method and system based on neural network self-learning | |
| CN104665849B (en) | A kind of high ferro dispatcher based on the interaction of many physiological signals multi-model stress detection method | |
| CN109933881A (en) | A kind of Fault Diagnosis of Power Electronic Circuits method based on optimization deepness belief network | |
| CN103901882B (en) | A kind of system and method for train dynamics system on-line monitoring fault diagnosis | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN109239585A (en) | A kind of method for diagnosing faults based on the preferred wavelet packet of improvement | |
| CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
| CN106443297A (en) | Decision tree SVM fault diagnosis method for three-level inverter of photovoltaic diode clamp type | |
| CN103064008B (en) | A kind of Nolinear analog circuit soft fault diagnostic method based on Hilbert-Huang transform | |
| CN106971152A (en) | A kind of method of Bird's Nest in detection transmission line of electricity based on Aerial Images | |
| CN105975504A (en) | Recurrent neural network-based social network message burst detection method and system | |
| CN107563069A (en) | A kind of wind power generating set intelligent fault diagnosis method | |
| CN103728551A (en) | Analog circuit fault diagnosis method based on cascade connection integrated classifier | |
| CN107909118A (en) | A kind of power distribution network operating mode recording sorting technique based on deep neural network | |
| CN109660206A (en) | A kind of diagnosing failure of photovoltaic array method based on Wasserstein GAN | |
| CN103439647A (en) | Analog circuit fault diagnosis method for dynamic supply current monitoring | |
| CN109858503A (en) | The traction converter failure diagnostic method of decision tree is promoted based on gradient | |
| CN105334436A (en) | Cross-linked cable partial discharge mode identification method based on SOM-BP combined neural network | |
| CN103018660A (en) | Multi-fault intelligent diagnosing method for artificial circuit utilizing quantum Hopfield neural network | |
| CN103345637A (en) | On-line detecting system and method of hydrophobicity of composite insulator of electric transmission line | |
| CN104318485A (en) | Power transmission line fault identification method based on nerve network and fuzzy logic | |
| CN109389325A (en) | Transformer substation electronic transducer state evaluating method based on wavelet neural network | |
| CN108334902A (en) | A kind of track train equipment room smog fireproof monitoring method based on deep learning | |
| Moloi et al. | High impedance fault classification and localization method for power distribution network | |
| CN104280253B (en) | A kind of fault diagnosis method and system based on immune detectors |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190118 |