CN107092934A - A kind of large scale structure damnification recognition method based on three-level data fusion - Google Patents
A kind of large scale structure damnification recognition method based on three-level data fusion Download PDFInfo
- Publication number
- CN107092934A CN107092934A CN201710278069.4A CN201710278069A CN107092934A CN 107092934 A CN107092934 A CN 107092934A CN 201710278069 A CN201710278069 A CN 201710278069A CN 107092934 A CN107092934 A CN 107092934A
- Authority
- CN
- China
- Prior art keywords
- data
- fusion
- large scale
- scale structure
- level data
- 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
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of large scale structure damnification recognition method based on three-level data fusion, comprise the steps of:A, the displacement of collection large scale structure to be detected and acceleration information simultaneously carry out data processing；B, using three-level data anastomosing algorithm the information gathered in step A is merged；C, the step B results drawn are identified, testing result is drawn by numerical simulation, reduce the transmission of hash, the transmission of redundancy）.A kind of three-level convergence strategy of present invention design completes the non-destructive tests to structure jointly, carries out pixel-based fusion first by the information that consensus algorithm is gathered to single sensor, improves the precision of data acquisition, reduces the volume of transmitted data of node;Reuse ACGA BP neural networks and carry out preliminary non-destructive tests using the intrinsic frequency of the displacement of static measurement data and dynamic measuring data as the input parameter of network respectively;Finally the recognition result based on static measurement data and the recognition result based on dynamic measuring data are merged again using D S evidence theories so that final recognition result is more accurate.
Description
Technical field
The present invention relates to a kind of sensor technology, specifically a kind of large scale structure non-destructive tests based on three-level data fusion
Method.
Background technology
In the bridge of Transport Hub, the high-rise building of city symbol, the big sport of civil plantation amusement and recreation
The isostructural health and safety of the heart, arts center and the people are closely related.Yet with these structural volumes are huge, structure is multiple
Miscellaneous, service life length, floor space are wide, if effectively monitoring and protecting and Gernral Check-up can not be implemented to it, it will generation is permitted
More uncertain factor.In the last few years, these civil structures were influenceed by environment or human factor during service, and destruction is collapsed
The accident of collapsing occurs repeatedly, and causes serious social influence.
Although countries in the world begin to focus on the degree of impairment of structure very early, because traditional damage identification technique falls
Afterwards, it is difficult to which comprehensively the damage status of structure is identified.At the beginning of 21 century, the in-service engineering structure in global range is entered
The reparation phase.Need to spend very huge financial fund due to reconstructing the large civil structures such as bridge, dam, therefore, entirely
Old civil structure is all considered as the wealth of preciousness by countries in the world, it is desirable to gone forward side by side the hand-manipulating of needle pair by the diagnostic assessment comprehensive to its
The repairing and reinforcement of property is to extend its service life, and this will save substantial amounts of manpower and materials.Wireless sensor network technology has ten
Divide good prospect and important Research Significance, large scale structure is laid between substantial amounts of wireless sensor node, node and passed through
ZigBee communications protocol carries out data transmission, expense that is so not only neat and artistic but also having saved cable, but simultaneously we also this see
Arrive, the laying of big quantity sensor equally also brings another problem, how accurately to be realized from mass of redundancy data to structure
The identification of damage, this is accomplished by the Data fusion technique of research emphasis one that another technology is also this paper.
The content of the invention
It is an object of the invention to provide a kind of large scale structure damnification recognition method based on three-level data fusion, to solve
The problem of being proposed in above-mentioned background technology.
To achieve the above object, the present invention provides following technical scheme：
A kind of large scale structure damnification recognition method based on three-level data fusion, is comprised the steps of:
A, the displacement of collection large scale structure to be detected and acceleration information simultaneously carry out data processing；
B, using three-level data anastomosing algorithm the information gathered in step A is merged；
C, the step B results drawn are identified, testing result is drawn by numerical simulation.
It is used as the further technical scheme of the present invention：The three-level data anastomosing algorithm includes pixel-based fusion, feature level
Fusion and decision level fusion.
It is used as the further technical scheme of the present invention：The pixel-based fusion uses consensus algorithm, first will be single
Some groups of data deleting unrelidble datas that individual sensor is collected, then fusion treatment is carried out with consensus algorithm, obtain
More accurate data.
It is used as the further technical scheme of the present invention：The feature-based fusion is used as pattern using ACGA-BP neutral nets
Identifier, respectively using frequency and displacement as input parameter, realizes the preliminary identification of structure.
It is used as the further technical scheme of the present invention：The decision level fusion uses D-S evidence theory, and analysis, which is discussed, adopts
With D-S evidence theory two kinds of preliminary recognition results is carried out with decision level fusion it is compared to mix frequency and displacement as neural
Network inputs parameter carries out the superiority of non-destructive tests.
It is used as the further technical scheme of the present invention：The step A comprises the following steps：A, acquisition of information, according to research
The actual conditions of object use a variety of sensors, and the signal that sensor is obtained incoming calculating after A/D is converted
Machine system, b, data prediction carry out the pretreatment of data using filtering or wild point elimination method, c, feature extraction will be sensed
The signal of device collection carries out feature extraction, and extraction is characterized in have the physical quantity of clear and definite physical significance or without any physical significance
Statistic and its deformation.
Compared with prior art, the beneficial effects of the invention are as follows：A kind of three-level convergence strategy of present invention design is completed jointly
To the non-destructive tests of structure, pixel-based fusion is carried out first by the information that consensus algorithm is gathered to single sensor,
The precision of data acquisition is improved, the volume of transmitted data of node is reduced;ACGA-BP neutral nets are reused respectively with static measurement number
According to displacement and the intrinsic frequency of dynamic measuring data as the input parameter of network carry out preliminary non-destructive tests;Finally utilize D-
S evidence theories are melted again to the recognition result based on static measurement data and the recognition result based on dynamic measuring data
Close so that final recognition result is more accurate.
Brief description of the drawings
Fig. 1 is convergence strategy structure chart.
Fig. 2 is Data Fusion Structure figure.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not made
Embodiment, belongs to the scope of protection of the invention.
Fig. 1-2 is referred to, a kind of large scale structure damnification recognition method based on three-level data fusion is comprised the steps of:
A, the displacement of collection large scale structure to be detected and acceleration information simultaneously carry out data processing；
B, using three-level data anastomosing algorithm the information gathered in step A is merged；
C, the step B results drawn are identified, testing result is drawn by numerical simulation.
Three-level data anastomosing algorithm includes pixel-based fusion, feature-based fusion and decision level fusion.
Pixel-based fusion uses consensus algorithm, and some groups of data for first collecting single sensor are rejected
Suspicious data, then fusion treatment is carried out with consensus algorithm, obtain more accurate data.
Feature-based fusion, as mode discriminator, is used as input using frequency and displacement respectively using ACGA-BP neutral nets
Parameter, realizes the preliminary identification of structure.
Decision level fusion uses D-S evidence theory, and analysis is discussed using D-S evidence theory to two kinds of preliminary recognition results
Progress decision level fusion, which is compared to mix frequency and displacement, is used as nerve network input parameter to carry out the superior of non-destructive tests
Property.
Step A comprises the following steps：A, acquisition of information, a variety of sensings are used according to the actual conditions of research object
Device, and the signal that sensor is obtained incoming computer system after A/D is converted, b, data prediction use filtering or wild
Point elimination method carries out the pretreatment of data, and c, feature extraction, the signal that sensor is gathered carry out feature extraction, the spy of extraction
It is physical quantity or the statistic without any physical significance and its deformation for having clear and definite physical significance to levy.
The present invention operation principle be：Current large scale structure non-destructive tests are faced with following difficulty:(1) for structure
Certain damage status, single non-destructive tests means are difficult to accurately identify, and frequently resulted in using a variety of different recognition methods
As a result not consistent, therefore, it is difficult to accurate identification information is obtained from a variety of results.(2) for the monitoring system of large scale structure
A variety of different characteristic informations are often obtained, the result obtained using different parameter progress non-destructive tests is also inconsistent.In order to
Problem above is solved, a kind of three-level convergence strategy is devised according to Data-Fusion theory herein.Melted in data level using uniformity
The method of conjunction, the error that reduction accidentalia is caused;Merged in feature level using ACGA-BP neutral nets;In decision level
Finally merged using D-S evidence theory.In order to improve the accuracy of result, surveyed herein using static measurement data with dynamic
The method that is combined of amount data is collectively as the basis for estimation of structural damage, and convergence strategy structure chart is as shown in Figure 1.
Every data message is gathered by sensor first, melted collecting one group of data progress data to single sensor
, it is necessary to first deleting unrelidble data before closing, and how to determine whether data meet requirement, it is necessary to a threshold value be set, when data are big
It is suspicious data that this data is just can consider when this threshold value, selects Greenland sea-ice determination methods can to distinguish herein
Doubt data.
Easily influenceed because sensor collects data by ambient noise and accidentalia, cause measurement data not
Accurately, therefore by single sensor the advanced row suspicious data of same category information collected is rejected uses consensus algorithm again
Pixel-based fusion is carried out, this pixel-based fusion can effectively improve sensor and collect the accuracy of data, and subtract
Lack the volume of transmitted data of terminal node, reduce node power consumption.
Feature-based fusion selects ACGA-BP neutral nets to complete, and BP neural network has good accuracy of identification, but its
Convergence rate is slow, is easily trapped into local optimum.Therefore here using improved self-adapting synergizing evolution genetic algorithm to nerve
The weights and threshold value of network carry out global search.So as to improve the precision of network and the time of training.Respectively with based on dynamic
The intrinsic frequency of measurement data and displacement based on static measurement data are obtained to research object as the input parameter of network
Preliminary non-destructive tests result.
The preliminary non-destructive tests result obtained using ACGA-BP networks still can not accomplish that height is accurate on accuracy of identification
Really, the recognition methods either based on static measurement data, which is also based on the recognition methods of dynamic measuring data, the office of oneself
It is sex-limited, when identification parameter has noise, different degrees of erroneous judgement just occurs, and D-S evidence theory employed herein is made
The possibility that erroneous judgement occurs can be effectively reduced for decision level fusion scheme, so as to considerably increase the standard of final recognition result
True property.
The training of BP neural network is the method amendment weights and threshold value reduced with gradient, is so easily trapped into part most
It is excellent, influence the overall performance of algorithm;Along with widely using for BP neural network, traditional utilization BP networks solve the disadvantage of problem
End is also showing, for the shortcoming of BP networks, and many scholars propose various improvement strategies, such as adaptive learning rate method, altogether
Gradient method etc. is gripped, these methods improve the performance of BP neural network to a certain extent, but still can not fundamentally solve
BP network weights and threshold value training problem.And genetic algorithm has stronger ability of searching optimum, but its local search ability is not
Foot.So being solved using genetic algorithm to the weights and threshold value of BP networks, two kinds of algorithms are combined, so that whole net
Network performance is more superior.
BP neural network is optimized using genetic algorithm and is broadly divided into three parts:The knot of neutral net is determined first
Structure, and then determine genetic algorithm individual information.Secondly using self-adapting synergizing evolution genetic algorithm Optimizing BP Network weights and
Threshold value, individual fitness value is calculated by fitness function to be obtained, and genetic algorithm finds optimum individual by genetic manipulation.Most
Afterwards, using the weights in optimum individual and threshold value information as network initial value.
Claims (6)
1. a kind of large scale structure damnification recognition method based on three-level data fusion, it is characterised in that comprise the steps of:
A, the displacement of collection large scale structure to be detected and acceleration information simultaneously carry out data processing；
B, using three-level data anastomosing algorithm the information gathered in step A is merged；
C, the step B results drawn are identified, testing result is drawn by numerical simulation.
2. a kind of large scale structure damnification recognition method based on three-level data fusion according to claim 1, its feature exists
In the three-level data anastomosing algorithm includes pixel-based fusion, feature-based fusion and decision level fusion.
3. a kind of large scale structure damnification recognition method based on three-level data fusion according to claim 2, its feature exists
In the pixel-based fusion uses consensus algorithm, and some groups of data for first collecting single sensor are rejected
Suspicious data, then fusion treatment is carried out with consensus algorithm, obtain more accurate data.
4. a kind of large scale structure damnification recognition method based on three-level data fusion according to claim 2, its feature exists
In the feature-based fusion, as mode discriminator, is joined using frequency and displacement as input respectively using ACGA-BP neutral nets
Number, realizes the preliminary identification of structure.
5. a kind of large scale structure damnification recognition method based on three-level data fusion according to claim 2, its feature exists
In the decision level fusion uses D-S evidence theory, and analysis is discussed using D-S evidence theory to two kinds of preliminary recognition results
Progress decision level fusion, which is compared to mix frequency and displacement, is used as nerve network input parameter to carry out the superior of non-destructive tests
Property.
6. a kind of large scale structure damnification recognition method based on three-level data fusion according to claim 1, its feature exists
In the step A comprises the following steps：A, acquisition of information, a variety of sensings are used according to the actual conditions of research object
Device, and the signal that sensor is obtained incoming computer system after A/D is converted, b, data prediction use filtering or wild
Point elimination method carries out the pretreatment of data, and c, feature extraction, the signal that sensor is gathered carry out feature extraction, the spy of extraction
It is physical quantity or the statistic without any physical significance and its deformation for having clear and definite physical significance to levy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710278069.4A CN107092934A (en) | 2017-04-25 | 2017-04-25 | A kind of large scale structure damnification recognition method based on three-level data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710278069.4A CN107092934A (en) | 2017-04-25 | 2017-04-25 | A kind of large scale structure damnification recognition method based on three-level data fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107092934A true CN107092934A (en) | 2017-08-25 |
Family
ID=59637991
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710278069.4A Pending CN107092934A (en) | 2017-04-25 | 2017-04-25 | A kind of large scale structure damnification recognition method based on three-level data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107092934A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298143A (en) * | 2019-07-30 | 2019-10-01 | 中冶建筑研究总院有限公司 | A kind of prestressing force truss string structure damnification recognition method based on two stages data fusion |
| CN110414602A (en) * | 2019-07-30 | 2019-11-05 | 中冶建筑研究总院有限公司 | A kind of prestressing force string chord member of truss damnification recognition method based on multi-modal data fusion |
| CN114764088A (en) * | 2021-12-17 | 2022-07-19 | 中国飞机强度研究所 | Fusion recognition method for layered damage state of composite plate on airplane |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB846722A (en) * | 1956-04-02 | 1960-08-31 | Ibm | Improvements in character sensing apparatus |
| KR20110035171A (en) * | 2009-09-30 | 2011-04-06 | 성균관대학교산학협력단 | Method and apparatus for context estimating |
| CN103147972A (en) * | 2013-03-19 | 2013-06-12 | 北京化工大学 | Reciprocating-type compressor fault diagnosis method based on multi-sensor information fusion |
| CN104766129A (en) * | 2014-12-31 | 2015-07-08 | 华中科技大学 | Subway shield construction surface deformation warning method based on temporal and spatial information fusion |
-
2017
- 2017-04-25 CN CN201710278069.4A patent/CN107092934A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB846722A (en) * | 1956-04-02 | 1960-08-31 | Ibm | Improvements in character sensing apparatus |
| KR20110035171A (en) * | 2009-09-30 | 2011-04-06 | 성균관대학교산학협력단 | Method and apparatus for context estimating |
| CN103147972A (en) * | 2013-03-19 | 2013-06-12 | 北京化工大学 | Reciprocating-type compressor fault diagnosis method based on multi-sensor information fusion |
| CN104766129A (en) * | 2014-12-31 | 2015-07-08 | 华中科技大学 | Subway shield construction surface deformation warning method based on temporal and spatial information fusion |
Non-Patent Citations (2)
| Title |
|---|
| 卢伟: "《基于现场总线的大跨空间结构健康监测系统研究》", 《中国博士学位论文全文数据库工程科技Ⅱ辑》 * |
| 奥特内斯等: "《数字时间序列分析》", 30 June 1982, 国防工业出版社 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298143A (en) * | 2019-07-30 | 2019-10-01 | 中冶建筑研究总院有限公司 | A kind of prestressing force truss string structure damnification recognition method based on two stages data fusion |
| CN110414602A (en) * | 2019-07-30 | 2019-11-05 | 中冶建筑研究总院有限公司 | A kind of prestressing force string chord member of truss damnification recognition method based on multi-modal data fusion |
| CN114764088A (en) * | 2021-12-17 | 2022-07-19 | 中国飞机强度研究所 | Fusion recognition method for layered damage state of composite plate on airplane |
| CN114764088B (en) * | 2021-12-17 | 2024-09-20 | 中国飞机强度研究所 | Method for identifying layered damage state fusion of composite material plate on airplane |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106548343B (en) | Illegal transaction detection method and device | |
| CN108777873A (en) | The wireless sensor network abnormal deviation data examination method of forest is isolated based on weighted blend | |
| CN106804046A (en) | Mobile location method and device based on measurement report | |
| CN107103754A (en) | A kind of road traffic condition Forecasting Methodology and system | |
| CN102495919B (en) | Extraction method for influence factors of carbon exchange of ecosystem and system | |
| CN107092934A (en) | A kind of large scale structure damnification recognition method based on three-level data fusion | |
| CN103886030B (en) | Cost-sensitive decision-making tree based physical information fusion system data classification method | |
| CN107958031A (en) | Resident trip OD distribution extracting methods based on fused data | |
| CN103400299B (en) | Method for detecting network overlapped communities based on overlapped point identification | |
| CN111382472A (en) | Method and device for predicting shield-induced proximity structure deformation by random forest fusion SVM (support vector machine) | |
| CN108632832B (en) | Network coverage analysis method and system | |
| CN102902981A (en) | Violent video detection method based on slow characteristic analysis | |
| CN106535204A (en) | Service coverage quality evaluation method and device | |
| CN109615616A (en) | A kind of crack identification method and system based on ABC-PCNN | |
| CN106127161A (en) | Fast target detection method based on cascade multilayer detector | |
| CN106846380A (en) | A kind of medical image registration method and equipment | |
| CN105930900A (en) | Method and system for predicting hybrid wind power generation | |
| CN116578931A (en) | Urban waterlogging and ponding prediction method based on Bagging-XGB model | |
| CN116843196A (en) | Intelligent training method and system applied to military training | |
| CN105228185A (en) | A kind of method for Fuzzy Redundancy node identities in identification communication network | |
| CN104280253B (en) | A kind of fault diagnosis method and system based on immune detectors | |
| CN104680118A (en) | Method and system for generating face character detection model | |
| CN103605670B (en) | A kind of method and apparatus for determining the crawl frequency of network resource point | |
| CN108833461A (en) | Social networks key node based on network decomposition finds method and system | |
| CN115239733B (en) | Crack detection method and apparatus, terminal device and storage medium |
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: 20170825 |
|
| RJ01 | Rejection of invention patent application after publication |