CN100483126C - Defect distinguish based on three-dimensional finite element NN and quantified appraisal method - Google Patents
Defect distinguish based on three-dimensional finite element NN and quantified appraisal method Download PDFInfo
- Publication number
- CN100483126C CN100483126C CNB2006101649236A CN200610164923A CN100483126C CN 100483126 C CN100483126 C CN 100483126C CN B2006101649236 A CNB2006101649236 A CN B2006101649236A CN 200610164923 A CN200610164923 A CN 200610164923A CN 100483126 C CN100483126 C CN 100483126C
- Authority
- CN
- China
- Prior art keywords
- defect
- magnetic flux
- finite element
- flux leakage
- leakage field
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
Abstract
This invention relates to deficiency identification and quantification based on three dimensional limit element neutral network, which comprises the following steps: a, according to deficiency leakage magnetic field three dimensional element computation module forming three dimensional element neutral network; b, measuring and extracting deficiency magnetic field characteristics value and setting measurement values and computing error valve conditions; c, given deficiency characteristics parameters initial values by use of three dimensional limit element neutral network for overlap computation to realize deficiency identification and evaluation through comparing deficiency computation values and error size.
Description
Technical field
The present invention relates to a kind of defect recognition and method for quantitatively evaluating, be used for the feature identification and the quantitatively evaluating of defect and magnetic leakage detection signal, belong to technical field of nondestructive testing based on three-dimensional finite element NN.
Background technology
Magnetic Flux Leakage Inspecting is a lossless detection method commonly used, be widely used aspect ferrimagnet quality testing and the safety monitoring, but it is the technical barrier of Non-Destructive Testing research field with quantitatively evaluating with the feature identification that realizes defective that the Magnetic Flux Leakage Inspecting signal is carried out reasonable dismissal always.That Chinese patent literature discloses is a kind of " pipeline defect and magnetic leakage detects the analytical approachs of data " (publication number: 1458442, open day: 2003.11.26), this technology relates to a kind of defect and magnetic leakage that can improve various steel pipe defect detection signal analysis efficiencies and accuracy of quantitative analysis and detects data analysis processing method, but this technology is by the video data cloud atlas, the defective that can only qualitatively judge pipeline has or not, and is difficult to realize quantifying defects; And patent " pipeline corrosion default kind identification method " (application number: 200410068973.5, publication number: 1587785A), this technology relates to a kind of pipeline corrosion default kind identification method, realize a large amount of Magnetic Flux Leakage Inspecting data Fast Classification are handled by simple algorithm and international standard, this technology only just can obtain comparatively ideal accuracy of identification to the regularization defective; Patent " a kind of quantization method of detecting corrosion defect by magnetic leakage " (application number: 200510011116.6, publication number: 1641347A), this technology relates to a kind of corrosion default Magnetic Flux Leakage Inspecting data analysis processing method, utilize detected stray field distribution characteristics, according to measuring the quantification of the match quantitative model realization of foundation by the Artificial discontinuity in advance to three direction sizes of corrosion default, its quantized result is big to the dependence of aforementioned statistical fit model data, and the difference of any testing conditions all may produce bigger influence to quantized result.
Summary of the invention
The purpose of this invention is to provide a kind of defect recognition and method for quantitatively evaluating based on three-dimensional finite element NN.Its essence is the dimensional Finite Element model is embedded in the neural network structure that the advantage of comprehensive finite element and neural network solves in the electromagnetism Nondestructive Evaluation nonlinear problem and finds the solution big, the slow-footed problem of calculated amount.Simultaneously, this invention can be avoided the dependence of conventional neural net method to training data, helps improving recognition capability and quantified precision to various random defects.
Technical scheme of the present invention is as follows: a kind of defect recognition and method for quantitatively evaluating based on three-dimensional finite element NN is characterized in that this method comprises the steps:
1) according to the dimensional Finite Element model construction three-dimensional finite element NN of defect magnetic flux leakage field: finite element NN comprises three layers: input layer, output layer and hidden layer.If defect magnetic flux leakage field dimensional Finite Element model comprises M grid cell, a N node, and there is i characteristic parameter each unit, and the input layer of neural network structure just comprises i * M neuron so, and these neurons multiply by the weight matrix element respectively
Afterwards, as the input of hidden layer; Hidden layer comprises N
2Individual neuron, N is one group, the weights φ of respective nodes is multiply by in the output of hidden layer again
j, just can obtain the output valve corresponding with this node; Output layer is a N neuron, represents the output valve at N node place;
2) measure defect magnetic flux leakage field, extract the defect magnetic flux leakage field eigenwert, and set the threshold value δ of error between defect magnetic flux leakage field measured value A and neural network calculated value B: according to the distribution characteristics of defect magnetic flux leakage field, employing waits the spatial sampling method to measure the stray field intensity at a certain regional some spots place around the defective, and extract its axially, circumferentially and radial component as the eigenwert of defect magnetic flux leakage field;
3) according to the size of measured defect magnetic flux leakage field, set the initial estimate of defect characteristic parameter, the three-dimensional finite element NN that utilizes step 1) to set up carries out iterative computation to it: if the error A-B between the measured value A of defect magnetic flux leakage field and neural network calculated value B is greater than step 2) set error threshold δ, then need upgrade the estimated value of defect characteristic parameter and calculate again; As error A-B during, think that promptly being used for the defect characteristic estimates of parameters that finite element NN calculates is desired defect characteristic parameter less than threshold value δ.
The defect magnetic flux leakage field eigenwert of extraction of the present invention is axial, the circumferential and radial component of defect magnetic flux leakage field.
Feature of the present invention also is: the defect magnetic flux leakage field eigenwert of being extracted above-mentioned steps 2) will be carried out normalized.
Defect recognition and method for quantitatively evaluating that the present invention proposes based on three-dimensional finite element NN, come down to limited element calculation model with the parallel mode realization, both had the high characteristics of FEM (finite element) calculation precision, had neural network concurrent calculating, fireballing advantage again.Simultaneously, because the weights of finite element NN depend on the differential equation and the boundary condition of waiting the problem of finding the solution, do not need to train in advance, when therefore utilizing it to carry out the quantifying defects evaluation, can avoid the dependence of conventional neural net method to training sample, help improving recognition capability and quantified precision, have more wide application prospect various random defects.
Description of drawings
Fig. 1 is the ultimate principle that makes up three-dimensional finite element NN according to the stray field limited element calculation model.
Fig. 2 is the defect parameters iteration evaluation method based on three-dimensional finite element NN.
Fig. 3 is linear tetrahedron element figure.
Fig. 4 is the structure of finite element NN.
Embodiment
Carry out defect recognition and method for quantitatively evaluating based on three-dimensional finite element NN, mainly comprise following three basic steps: 1), make up three-dimensional finite element NN according to the dimensional Finite Element model of defect magnetic flux leakage field; 2) eigenwert of measurement and extraction defect magnetic flux leakage field, the threshold condition of error between setting defect magnetic flux leakage field measured value and calculated value; 3) initial estimate of given defect characteristic parameter utilizes three-dimensional finite element NN to carry out iterative computation, realizes defect geometry feature identification and quantitatively evaluating by the stray field calculated value that compares and measures and the error size between eigenwert.
Below in conjunction with accompanying drawing above steps is further described:
1) according to the dimensional Finite Element model construction three-dimensional finite element NN of defect magnetic flux leakage field: finite element NN comprises three layers: input layer, output layer and hidden layer.If the three-dimensional finite element mesh structure of defect magnetic flux leakage field computation model correspondence comprises M linear tetrahedron element (as Fig. 3), a N node, the characteristic parameter of setting each unit is α, β and γ, the input layer of so constructed finite element NN just should comprise 3M neuron, i.e. M α, a M β and M γ.They multiply by weights respectively
With
(i=1,2 ..., M; J=1,2 ..., N) afterwards, as the input of hidden layer; Hidden layer comprises N
2Individual neuron, N is one group; The weights φ of respective nodes is multiply by in the output of hidden layer again
j, just can obtain the b corresponding with this node
jValue; Output layer is a N neuron, represents the b at N node place
jValue.In view of the above, finite element NN structural representation such as Fig. 4 of structure; The weights of three-dimensional finite element NN
With
And weights φ
jDepend on the differential equation and the boundary condition of waiting the problem of finding the solution, do not need to train in advance, therefore do not have dependence training sample.
2) at the quantitative evaluation problem of Magnetic Flux Leakage Inspecting pipeline corrosion default, at first employing waits the spatial sampling method to measure the defective stray field intensity at a certain regional some spots place on every side, and it is carried out normalized respectively, select respectively then its axially, circumferentially and radial component as the eigenwert of defect magnetic flux leakage field.Simultaneously, the error threshold between setting measurement value and network calculations value is δ;
3) according to the size of measured defect magnetic flux leakage field, with the α, the β that set and the γ initial estimate as the defect characteristic parameter, the three-dimensional finite element NN that the utilization step 1) is set up carries out iterative computation to it.
(1) suppose iterative computation to the t time circulation time, the characteristic parameter estimated value of defective is α (t), β (t), γ (t), if with α (t), β (t), γ (t) during respectively as the input of neural network, the network that then calculates is output as:
(2) calculate the error that neural network is exported:
If error E<δ, then α (t), β (t), γ (t) promptly are desired defect characteristic estimates of parameters; Otherwise carry out following calculating.
(3) utilize the gradient descent method to upgrade α (t), β (t) and γ (t):
(4) with α (t+1), β (t+1), γ (t+1) respectively as the input of neural network, repeating step (1)~(4) are till error E<δ.At this moment, think that promptly the neural network input value after upgrading is the quantized result of desired defect characteristic parameter.
Claims (3)
1. defect recognition and method for quantitatively evaluating based on a three-dimensional finite element NN is characterized in that this method comprises the steps:
1) according to the dimensional Finite Element model construction three-dimensional finite element NN of defect magnetic flux leakage field: finite element NN comprises three layers: input layer, output layer and hidden layer, if defect magnetic flux leakage field dimensional Finite Element model comprises M grid cell, a N node, and there is i characteristic parameter each unit, the input layer of neural network structure just comprises i * M neuron so, and these neurons multiply by the weight matrix element respectively
Afterwards, as the input of hidden layer; Hidden layer comprises N
2Individual neuron, N is one group, the weights φ of respective nodes is multiply by in the output of hidden layer again, just can obtain the output valve corresponding with this node; Output layer is a N neuron, represents the output valve at N node place;
2) measure defect magnetic flux leakage field, extract the defect magnetic flux leakage field eigenwert, and set the threshold value δ of error between defect magnetic flux leakage field measured value A and neural network calculated value B: according to the distribution characteristics of defect magnetic flux leakage field, employings waits the spatial sampling method measurement defective stray field intensity at a certain regional some spots place on every side;
3) according to the size of measured defect magnetic flux leakage field, set the initial estimate of defect characteristic parameter, the three-dimensional finite element NN that utilizes step 1) to set up carries out iterative computation to it: if the error A-B between the measured value A of defect magnetic flux leakage field and calculated value B is greater than step 2) set error threshold δ, the estimated value that then need utilize the gradient descent method to upgrade the defect characteristic parameter is calculated again; As error A-B during, think that promptly being used for the defect characteristic estimates of parameters that finite element NN calculates is desired defect characteristic parameter less than threshold value δ.
2, defect recognition and method for quantitatively evaluating based on three-dimensional finite element NN according to claim 1 is characterized in that: the defect magnetic flux leakage field eigenwert of described extraction is axial, the circumferential and radial component of defect magnetic flux leakage field.
3, defect recognition and method for quantitatively evaluating based on three-dimensional finite element NN according to claim 1 and 2, it is characterized in that: defect magnetic flux leakage field eigenwert measured above-mentioned steps 2) is carried out normalized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006101649236A CN100483126C (en) | 2006-12-08 | 2006-12-08 | Defect distinguish based on three-dimensional finite element NN and quantified appraisal method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2006101649236A CN100483126C (en) | 2006-12-08 | 2006-12-08 | Defect distinguish based on three-dimensional finite element NN and quantified appraisal method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1963491A CN1963491A (en) | 2007-05-16 |
CN100483126C true CN100483126C (en) | 2009-04-29 |
Family
ID=38082632
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB2006101649236A Expired - Fee Related CN100483126C (en) | 2006-12-08 | 2006-12-08 | Defect distinguish based on three-dimensional finite element NN and quantified appraisal method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100483126C (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101995436A (en) * | 2010-10-26 | 2011-03-30 | 江苏大学 | Neural network based drawing part crack monitoring method |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102735747A (en) * | 2012-04-10 | 2012-10-17 | 南京航空航天大学 | Defect quantitative identification method of high-speed magnetic flux leakage inspection of high-speed railway rails |
CN103196356B (en) * | 2013-04-07 | 2015-10-21 | 克拉玛依市金牛工程建设有限责任公司 | Based on the oil tube defect quantitative recognition method of support vector machine |
ES2738319T3 (en) | 2014-09-12 | 2020-01-21 | Microsoft Technology Licensing Llc | Computer system to train neural networks |
CN104458895A (en) * | 2014-12-08 | 2015-03-25 | 清华大学 | Three-dimensional pipeline leakage flux imaging detection method and system |
CN104897771B (en) * | 2015-05-12 | 2018-01-23 | 清华大学 | Three-dimensional Magnetic Flux Leakage Inspecting defect profile reconstructing method and device |
CN104899868B (en) * | 2015-05-12 | 2017-08-25 | 清华大学 | Three-dimensional Magnetic Flux Leakage Inspecting defect inversion imaging method |
CN104990977B (en) * | 2015-06-29 | 2018-02-09 | 清华大学 | Three-dimensional Magnetic Flux Leakage Inspecting defect neutral net iterative inversion imaging method |
CN105021694B (en) * | 2015-07-02 | 2017-12-19 | 清华大学 | Magnetic Flux Leakage Inspecting quantifying defects and display methods under endless entire signal |
CN107055732A (en) * | 2017-05-17 | 2017-08-18 | 北京易沃特科技有限公司 | A kind of heavy metals minimizing technology and device |
CN107255671B (en) * | 2017-06-29 | 2019-10-29 | 清华大学 | Steel plate defect magnetic rotation battle array imaging detection method and detection device |
EP3467489A1 (en) * | 2017-10-06 | 2019-04-10 | Rosen Swiss AG | Method for determining the geometry of a defect and for determining a load capacity limit |
CN108828059B (en) * | 2018-06-29 | 2020-04-07 | 清华大学 | Electromagnetic multi-field coupling defect comprehensive detection and evaluation method and device |
CN111562305B (en) * | 2019-02-14 | 2024-01-26 | 爱科维申科技(天津)有限公司 | Automobile shock absorber piston defect detection method based on electromagnetic tomography technology |
CN110108783B (en) * | 2019-05-14 | 2023-04-28 | 上海市特种设备监督检验技术研究院 | Pipeline defect detection method and device based on convolutional neural network |
CN110470729B (en) * | 2019-07-31 | 2022-12-02 | 浙江树人学院(浙江树人大学) | Eddy current-based nondestructive testing method for oil field casing pipe defects |
CN113390956B (en) * | 2021-06-18 | 2024-02-20 | 西安建筑科技大学 | Double-magnetic-sensor probe and magnetic leakage detection defect quantitative evaluation method based on same |
CN117540283B (en) * | 2024-01-10 | 2024-04-26 | 河北省药品医疗器械检验研究院(河北省化妆品检验研究中心) | Intelligent electromagnetic appliance performance evaluation method |
-
2006
- 2006-12-08 CN CNB2006101649236A patent/CN100483126C/en not_active Expired - Fee Related
Non-Patent Citations (2)
Title |
---|
基于神经网络的冷轧带钢表面缺陷检测. 刘红冰等.中国图象图形学报,第10卷第10期. 2005 |
基于神经网络的冷轧带钢表面缺陷检测. 刘红冰等.中国图象图形学报,第10卷第10期. 2005 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101995436A (en) * | 2010-10-26 | 2011-03-30 | 江苏大学 | Neural network based drawing part crack monitoring method |
CN101995436B (en) * | 2010-10-26 | 2012-08-15 | 江苏大学 | Neural network based drawing part crack monitoring method |
Also Published As
Publication number | Publication date |
---|---|
CN1963491A (en) | 2007-05-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN100483126C (en) | Defect distinguish based on three-dimensional finite element NN and quantified appraisal method | |
CN103293014B (en) | Bridge fatigue damage state and residual life evaluating method | |
Sony et al. | Vibration-based multiclass damage detection and localization using long short-term memory networks | |
CN103076394B (en) | Safety evaluation method for ocean platform based on integration of vibration identification frequencies and vibration mode | |
CN103884593B (en) | Construction damage positionings based on a small amount of static measurement data and degree recognition methods | |
Laxman et al. | Automated crack detection and crack depth prediction for reinforced concrete structures using deep learning | |
CN107356677B (en) | Ultrasonic nondestructive testing method based on travel time tomography and reverse time migration imaging | |
CN107024531A (en) | Magnetic Flux Leakage Inspecting cell defect is stretched reconstructing method | |
CN104299033A (en) | Magnetic flux leakage defect reconstruction method based on cuckoo searching and particle filter hybrid algorithm | |
CN112762362A (en) | Underwater pipeline leakage acoustic emission detection method based on convolutional neural network | |
Truong et al. | An effective framework for real-time structural damage detection using one-dimensional convolutional gated recurrent unit neural network and high performance computing | |
CN112668526A (en) | Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing | |
CN110555235A (en) | Structure local defect detection method based on vector autoregressive model | |
CN112836758A (en) | Independent metering area pipe network leakage detection method based on double-layer circulating neural network | |
CN106997410B (en) | Damage occurrence judgment method based on modal strain energy | |
CN115047064A (en) | Pipeline defect quantification method, processor and pipeline defect quantification device | |
Amanollah et al. | Damage detection of structures based on wavelet analysis using improved AlexNet | |
CN117233347B (en) | Carbon steel spheroidization grade measuring method, system and equipment | |
CN114660180A (en) | Sound emission and 1D CNNs-based light-weight health monitoring method and system for medium and small bridges | |
Pan et al. | A novel method for defects marking and classifying in MFL inspection of pipeline | |
Li et al. | Quantitative nondestructive testing of broken wires for wire rope based on magnetic and infrared information | |
CN106849873A (en) | Solar cell Radiation hardness non-destructive method and device | |
Schneider et al. | Effect of different inspection strategies on the reliability of Daniels systems subjected to fatigue | |
CN112131781A (en) | Steel structure damage detection method based on full-connection neural network and transfer rate function | |
CN101319890A (en) | Compensation method for engineering scientific information treatment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090429 Termination date: 20171208 |
|
CF01 | Termination of patent right due to non-payment of annual fee |