CN115096987B - Pipeline defect quantification method based on magnetic leakage signal characteristics - Google Patents
Pipeline defect quantification method based on magnetic leakage signal characteristics Download PDFInfo
- Publication number
- CN115096987B CN115096987B CN202210729456.6A CN202210729456A CN115096987B CN 115096987 B CN115096987 B CN 115096987B CN 202210729456 A CN202210729456 A CN 202210729456A CN 115096987 B CN115096987 B CN 115096987B
- Authority
- CN
- China
- Prior art keywords
- defect
- peak
- depth
- width
- length
- 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.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 366
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000011002 quantification Methods 0.000 title claims abstract description 13
- 238000013139 quantization Methods 0.000 claims abstract description 51
- 238000013528 artificial neural network Methods 0.000 claims abstract description 43
- 238000012549 training Methods 0.000 claims description 28
- 230000004907 flux Effects 0.000 claims description 19
- 238000012360 testing method Methods 0.000 claims description 19
- 239000008186 active pharmaceutical agent Substances 0.000 claims description 16
- 238000012937 correction Methods 0.000 claims description 9
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 abstract description 3
- 239000011295 pitch Substances 0.000 description 22
- 238000004088 simulation Methods 0.000 description 11
- 230000008859 change Effects 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 8
- 238000001514 detection method Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 5
- 230000005284 excitation Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000006698 induction Effects 0.000 description 3
- 238000004445 quantitative analysis Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000005415 magnetization Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000002950 deficient Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N27/00—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means
- G01N27/72—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating magnetic variables
- G01N27/82—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating magnetic variables for investigating the presence of flaws
- G01N27/83—Investigating or analysing materials by the use of electric, electrochemical, or magnetic means by investigating magnetic variables for investigating the presence of flaws by investigating stray magnetic fields
Landscapes
- Chemical & Material Sciences (AREA)
- Chemical Kinetics & Catalysis (AREA)
- Electrochemistry (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
Abstract
The invention discloses a pipeline defect quantification method based on magnetic leakage signal characteristics, which comprises the steps of obtaining triaxial magnetic leakage signals of defects under different defect sizes and constructing a sample set; selecting the peak-to-valley spacing of the axial component differential signal as a quantization parameter of the defect length to obtain a defect length quantization formula: selecting the circumferential component peak Gu Zhongzhi interval as a quantization parameter of the defect width to obtain defect width quantization; and obtaining a mapping relation between the length and width of the defect and a defect depth fitting parameter by establishing a neural network, and calculating the defect depth by taking the defect depth fitting parameter and a radial component peak-valley value as quantization parameters of the defect depth. The method can more accurately quantify the defect size under the conditions of unknown geometric parameters of the residual defects or fluctuation within a certain range of the lift-off distance, and the like, and has stronger anti-interference capability and simpler quantification mode.
Description
Technical Field
The invention relates to the technical field of pipeline defect detection, in particular to a pipeline defect quantification method based on magnetic leakage signal characteristics.
Background
In pipeline security engineering, pipeline inspection is a basic method of ensuring pipeline security. Among the various pipeline detection technologies, the magnetic flux leakage internal detection technology is the most widely applied and mature magnetic pipeline defect detection technology. The magnetic flux leakage internal detection technology is characterized in that the size of the pipeline defect cannot be directly measured in the operation process, the core problem of the magnetic flux leakage internal detection technology is the inversion problem of the size of the pipeline defect, and the defect size is reduced through the collected magnetic flux leakage signals. The complex relationship between the magnetic leakage signal and the defect size causes the analysis of the magnetic leakage signal to realize defect quantification, which is a technical problem. In the traditional defect quantification method, axial component characteristics are used as evaluation, signal characteristic sources are single, the identification degree is not high, and the defect quantification accuracy is reduced.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides the pipeline defect quantification method based on the magnetic leakage signal characteristics, which can identify the type of the pipeline defect according to the magnetic leakage signal of the pipeline, has stronger anti-interference capability and simpler quantification mode, and has great engineering significance and good application prospect.
In order to achieve the above purpose, the present invention adopts the following technical scheme, including:
A pipeline defect quantification method based on magnetic leakage signal characteristics comprises the following steps:
s1, acquiring triaxial magnetic leakage signals of defects under different defect sizes, and constructing a sample set;
the defect sizes include: defect length, defect width, defect depth;
the triaxial magnetic leakage signal of the defect is as follows: axial magnetic leakage signals are axial components, radial magnetic leakage signals are radial components, and circumferential magnetic leakage signals are circumferential components;
The axial direction is the direction along the length of the pipeline, the radial direction is the direction vertical to the inner wall of the pipeline, and the circumferential direction is the circumferential direction of the pipeline;
The circumferential component is selected in the following way: determining the peak position of a radial component on an axial path, and selecting the value of the radial component on a circumferential path where the peak position is located as the value of the circumferential component;
S2, differentiating the axial components in the sample set to obtain an axial component differential signal, extracting the peak-to-valley spacing DS xp-p of the axial component differential signal as a quantization parameter of the defect length L, and performing linear fitting on the peak-to-valley spacing DS xp-p of the axial component differential signal in the sample set and the defect length L to obtain a defect length quantization formula:
L＝a1*DSxp-p+b1
Wherein a 1 is a length scale factor, and b 1 is a length correction factor;
S3, extracting a circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set as a quantization parameter of the defect width W, and performing linear fitting on the circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set and the defect width W to obtain a defect width quantization formula:
W＝a2*Sy-50％+b2
wherein a 2 is a width scaling factor, and b 2 is a width correction factor;
The circumferential component peak Gu Zhongzhi pitch S y-50％ is extracted by: extracting peaks and troughs of the circumferential component, calculating a median value of the peaks and the troughs, namely 50% of a difference value between the peaks and the troughs, as a peak Gu Zhongzhi, and obtaining a pitch of a median value of the peaks and the troughs on the circumferential component, namely a pitch S y-50％ of peaks Gu Zhongzhi of the circumferential component;
S4, extracting a radial component peak-valley value B zp-p in the sample set, namely, a difference value between a peak and a trough of a radial component, taking the radial component peak-valley value B zp-p as a quantization parameter of the defect depth D, and fitting the radial component peak-valley value B zp-p in the sample set and the defect depth D to obtain a defect depth quantization formula:
D＝a*Bzp-p+b
Wherein a and b are defect depth fitting parameters and are variables, and when the defect length or defect width is changed, the corresponding defect depth fitting parameters a and b are also changed;
s5, constructing a neural network, wherein the input of the neural network is defect length L and defect width W, the output of the neural network is defect depth fitting parameters a and b, and training is performed by using a sample set to generate the neural network;
S6, quantifying the unknown defects as follows:
s61, calculating the defect length L of the unknown defect according to the axial component differential signal peak-valley spacing DS xp-p of the unknown defect and a defect length quantization formula;
S62, calculating the defect width W of the unknown defect according to the circumferential component peak Gu Zhongzhi interval S y-50％ of the unknown defect and a defect width quantization formula;
s63, inputting the calculated length L and width W of the unknown defect into a neural network model, and predicting to obtain defect depth fitting parameters a and b;
S64, substituting the radial component peak-valley value B zp-p of the unknown defect and predicted defect depth fitting parameters a and B into a defect depth quantization formula, and calculating to obtain the defect depth D of the unknown defect.
Preferably, in step S1, the triaxial magnetic leakage signals of defects under different defect sizes are simulated by finite element software modeling.
Preferably, in step S5, the sample set is divided into a training set and a test set, and the neural network is trained by using the training set:
And fitting the radial component peak-valley value B zp-p with the same defect length and defect width in the training set with the defect depth D to obtain defect depth fitting parameters a and B under the defect length and the defect width, so as to obtain training data of the neural network, and training the neural network by using the training data of the neural network.
Preferably, in step S5, the sample set is divided into a training set and a test set, and the neural network is tested by using the test set:
inputting the defect length L and the defect width W of known defects in the test set into the trained neural network, and predicting to obtain corresponding defect depth fitting parameters a and b; substituting the corresponding defect depth fitting parameters a and B and the radial component peak-to-valley value B zp-p of the defect into a defect depth quantization formula, and calculating to obtain a predicted value of the defect depth D; and comparing the predicted value of the defect depth D with the true value of the defect depth D to test the prediction accuracy of the neural network.
The invention has the advantages that:
(1) According to analysis of sample data, the defect length is most stable by utilizing the peak-valley spacing of the axial component differential signal, the peak-valley spacing of the axial component differential signal is not influenced by factors such as defect width, defect depth, sensor lift-off value and the like, namely, the peak-valley spacing of the axial component differential signal is not changed along with the change of the defect length when the disturbance variable is changed, and the peak-valley spacing of the axial component differential signal is only changed along with the change of the defect length, so that the peak-valley spacing of the axial component differential signal is selected as a quantization parameter of the defect length according to the invention, and a defect length quantization formula is constructed according to a sample set.
(2) According to analysis of sample data, only the signal characteristic quantity of the circumferential component peak Gu Zhongzhi interval can be called as a stable quantity and does not change along with the change of the defect depth and the lifting distance, so that the circumferential component peak Gu Zhongzhi interval is selected as a quantization parameter of the defect width according to the invention, and a defect width quantization formula is constructed according to a sample set.
(3) According to analysis of sample data, although the two signal characteristic quantities, namely the radial component peak-valley value and the axial component peak-valley value, are suitable to be used as signal characteristic quantities for quantifying defect depth, the defect length is determined by the axial component differential signal peak-valley distance, the defect width size is determined by the circumferential component peak Gu Zhongzhi distance, the mapping relation between the length and width of the defect and defect depth fitting parameters a and b is obtained by establishing a BP neural network model, and then the defect depth is calculated by taking the defect depth fitting parameters and the radial component peak-valley value as quantization parameters of the defect depth, so that the defect depth is more accurate than a mode of directly quantifying the defect depth by utilizing the radial component peak-valley value.
(4) According to the method, the magnetic flux leakage signals of the pipeline under different defect sizes are simulated through finite element modeling, the mutual influence relation among the characteristic quantities of the signals, the defect length, the defect width and the defect depth is researched and analyzed, the characteristic effectiveness of the defect length, the defect width and the defect depth and the characteristic quantities of the signals can be obtained, the signal characteristic quantities can be used for representing the superiority and inferiority of geometric parameters of the defects and obtaining an optimal representation result.
(5) The method of the invention can acquire different sample data under different working scenes, different equipment, different materials, different magnetization conditions and the like, but the whole quantitative prediction method is applicable to all occasions.
Drawings
FIG. 1 is a flow chart of a method for quantifying pipeline defects based on magnetic flux leakage signal characteristics.
Fig. 2 is a graph of leakage flux of an axial component differential signal.
Fig. 3 is a magnetic flux leakage graph of the circumferential component.
Fig. 4 is a graph of leakage flux for radial components.
FIG. 5 is a graph of leakage flux of differential signals of axial components of different defect lengths.
FIG. 6 is a graph of axial component differential signal peak-to-valley spacing versus defect length.
FIG. 7 is a graph of axial component differential signal peak-to-valley spacing versus defect width for different defect lengths.
FIG. 8 is a graph of axial component differential signal peak-to-valley spacing versus defect depth for different defect lengths.
FIG. 9 is a graph of peak-to-valley spacing of axial component differential signals for different defect lengths versus lift-off.
Fig. 10 is a graph of leakage flux of circumferential components of different defect widths.
Fig. 11 is a graph of circumferential component peak Gu Zhongzhi pitch versus defect width.
FIG. 12 is a graph of circumferential component peak Gu Zhongzhi spacing versus defect depth for different defect widths.
Fig. 13 is a graph of circumferential component peak Gu Zhongzhi spacing versus lift-off values for different defect widths.
FIG. 14 is a graph of leakage flux for radial components of different defect depths.
FIG. 15 is a graph of radial component peak to valley versus defect depth for the same defect length and width.
FIG. 16 is a graph of radial component peak to valley versus defect depth for different defect widths.
Fig. 17 is a schematic diagram of training results of a BP neural network.
FIG. 18 is a graph comparing predicted values of defect depth with actual values of defect depth for different defect lengths and widths.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, a pipeline defect quantification method based on magnetic leakage signal characteristics is characterized by comprising the following steps:
s1, acquiring triaxial magnetic leakage signals of defects under different defect sizes, and constructing a sample set.
The defect sizes include: defect length, defect width, defect depth.
The triaxial magnetic leakage signal of the defect is as follows: axial magnetic leakage signals are axial components, radial magnetic leakage signals are radial components, and circumferential magnetic leakage signals are circumferential components;
the axial direction is the direction along the length of the pipeline, the radial direction is the direction perpendicular to the inner wall of the pipeline, namely the normal direction of the inner wall of the pipeline, and the circumferential direction is the circumferential direction along the pipeline, namely the circumferential direction around the pipeline;
The circumferential component is constructed by selecting the value of the radial component on the circumferential path of the pipe in the following specific manner: determining the peak position of a radial component on an axial path, and selecting the value of the radial component on a circumferential path where the peak position is located as the value of the circumferential component;
S2, differentiating the axial components in the sample set to obtain an axial component differential signal, extracting the peak-to-valley spacing DS xp-p of the axial component differential signal as a quantization parameter of the defect length L, and performing linear fitting on the peak-to-valley spacing DS xp-p of the axial component differential signal in the sample set and the defect length L to obtain a defect length quantization formula:
L＝a1*DSxp-p+b1
Wherein a 1 is a length scale factor, and b 1 is a length correction factor;
as shown in fig. 2, fig. 2 is a graph of leakage flux of the axial component differential signal, in which the vertical axis represents the magnetic induction intensity differential value and the horizontal axis represents the axial distance.
S3, extracting a circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set as a quantization parameter of the defect width W, and performing linear fitting on the circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set and the defect width W to obtain a defect width quantization formula:
W＝a2*Sy-50％+b2
wherein a 2 is a width scaling factor, and b 2 is a width correction factor;
as shown in fig. 3, the circumferential component peak Gu Zhongzhi pitch S y-50％ is extracted by: and extracting peaks and troughs of the circumferential component, calculating 50% of a median value of the peaks and the troughs, namely a difference value S max between the peaks and the troughs, as peaks Gu Zhongzhi, and obtaining a pitch of peaks and troughs on the circumferential component, namely a pitch S y-50％ of peaks Gu Zhongzhi of the circumferential component. Fig. 3 is a graph of leakage magnetic flux of circumferential components, in which the vertical axis represents magnetic induction intensity and the horizontal axis represents circumferential distance.
S4, extracting a radial component peak-valley value B zp-p in the sample set, namely, a difference value between a peak and a trough of a radial component, taking the radial component peak-valley value B zp-p as a quantization parameter of the defect depth D, and fitting the radial component peak-valley value B zp-p in the sample set and the defect depth D to obtain a defect depth quantization formula:
D＝a*Bzp-p+b
wherein a and b are defect depth fitting parameters, and are variables, and when the defect length or defect width is changed, the corresponding defect depth fitting parameters a and b are also changed.
The radial component peak-to-valley value B zp-p is shown in fig. 4, and S zp-p in fig. 4 represents the radial component peak-to-valley pitch. Fig. 4 is a graph of leakage magnetic flux of radial component, in which the vertical axis represents magnetic induction intensity and the horizontal axis represents axial distance.
S5, constructing a neural network, wherein the input of the neural network is defect length L and defect width W, the output of the neural network is defect depth fitting parameters a and b, and the neural network is trained and generated by using a sample set.
Dividing a sample set into a training set and a testing set, and training the neural network by using the training set:
And fitting the radial component peak-valley value B zp-p with the same defect length and defect width in the training set with the defect depth D to obtain defect depth fitting parameters a and B under the defect length and the defect width, so as to obtain training data of the neural network, and training the neural network by using the training data of the neural network.
Testing the neural network by using the test set: inputting the defect length L and the defect width W of known defects in the test set into the trained neural network, and predicting to obtain corresponding defect depth fitting parameters a and b; substituting the corresponding defect depth fitting parameters a and B and the radial component peak-to-valley value B zp-p of the defect into a defect depth quantization formula, and calculating to obtain a predicted value of the defect depth D; and comparing the predicted value of the defect depth D with the true value of the defect depth D to test the prediction accuracy of the neural network.
S6, quantifying the unknown defects as follows:
s61, calculating the defect length L of the unknown defect according to the axial component differential signal peak-valley spacing DS xp-p of the unknown defect and a defect length quantization formula;
S62, calculating the defect width W of the unknown defect according to the circumferential component peak Gu Zhongzhi interval S y-50％ of the unknown defect and a defect width quantization formula;
s63, inputting the calculated length L and width W of the unknown defect into a neural network model, and predicting to obtain defect depth fitting parameters a and b;
S64, substituting the radial component peak-valley value B zp-p of the unknown defect and predicted defect depth fitting parameters a and B into a defect depth quantization formula, and calculating to obtain the defect depth D of the unknown defect.
In the embodiment, the defect magnetic flux leakage signals under different defect sizes are simulated through finite element software modeling to construct a sample set. The sample data are: the size of the defect and the corresponding triaxial magnetic leakage signal. The size of the defect includes the length, width, depth of the defect. The defective triaxial magnetic leakage signal means: axial magnetic leakage signals are axial components, radial magnetic leakage signals are radial components, and circumferential magnetic leakage signals are circumferential components; in this embodiment, the circumferential component is constructed by selecting the value of the radial component on the circumferential path of the pipe in the following manner: the peak position of the radial component on the axial path is determined, and the value of the radial component on the circumferential path where the peak position is located is selected as the value of the circumferential component.
In this embodiment, after the length, width and depth of the defect are quantitatively analyzed according to the sample set, a specific quantization mode of the length, width and depth of the defect is obtained, and the quantization analysis is specifically as follows:
(1) Quantitative analysis of defect length:
According to the sample set, as the defect length increases, the axial component peak-to-valley distance, the axial component differential signal peak-to-valley distance and the radial component peak-to-valley distance are monotonically increased, namely a linear relation is formed; the radial component peak-valley value is only approximate single subtraction; the axial component peaks and valleys are non-linear and decrease progressively as the defect length increases.
Taking the defect width of 10mm and the defect depth of 7mm as an example, as shown in fig. 5, axial component differential signals with different defect lengths are obtained by finite element simulation, wherein the axial component differential signals refer to axial component derivation along the axial direction; in this embodiment, the axial component differential signals with defect lengths of 3, 6, 9, 12, 15, 18mm are obtained by finite element simulation. From the simulation results, it can be seen that the peak-to-valley distances DS xp-p of the differential signal of the defect axial components with different lengths are in direct proportion to the defect lengths, as shown in fig. 6.
In the three signal characteristic quantities of axial component peak-valley spacing, axial component differential signal peak-valley spacing and radial component peak-valley spacing, the defect length is most stable by utilizing the axial component differential signal peak-valley spacing, the axial component differential signal peak-valley spacing is not influenced by factors such as defect width, defect depth, sensor lift-off value and the like, namely, when the disturbance quantity is changed, the axial component differential signal peak-valley spacing is not changed along with the change of the defect length, the axial component differential signal peak-valley spacing is only changed along with the change of the defect length, and the axial component differential signal peak-valley spacing is verified in finite element simulation, and the fluctuation precision errors of the axial component differential signal peak-valley spacing are all within 10%. In order to verify the conclusion, the extraction distances are set to be 1, 2, 3, 4 and 5mm respectively, the defect widths are set to be 6, 8, 10, 12 and 14mm respectively, when the defect depths are set to be 4, 6, 8, 10 and 12mm respectively, the leakage detection is carried out to obtain the peak-valley pitches of the axial component differential signals when the defect sizes and the extraction distances are respectively changed, wherein one defect length is a group, 5 groups are divided, the defect lengths are respectively 6, 9, 12 and 15 and 18mm, the peak-valley pitches of the axial component differential signals and the corresponding defect widths, defect depths and extraction distances are obtained, the obtained data are generated into a line graph, and as shown in fig. 7, 8 and 9, the fluctuation accuracy errors of the peak-valley pitches of the axial component differential signals are verified to be within 10% by calculating the fluctuation percentages of each line graph. Secondly, the axial component peak-to-valley spacing and the radial component peak-to-valley spacing are relatively stable, and the two signal feature quantities are not influenced by the defect width and the defect depth, but are influenced by the fluctuation of the sensor lift-off value. As the lift-off increases, the axial component peak-to-valley spacing decreases and then becomes a single peak, while the radial component peak-to-valley spacing increases.
In summary, the axial component differential signal peak-to-valley spacing DS xp-p is selected as a quantization parameter of the defect length L, and linear fitting is performed on the axial component differential signal peak-to-valley spacing DS xp-p and the defect length L in the sample set to obtain a defect length quantization formula:
L＝a1*DSxp-p+b1
Wherein a 1 is a length proportionality coefficient, b 1 is a length correction coefficient, a 1、b1 is a constant value in general, and can be adjusted according to specific conditions of different working environments, pipes, excitation strength, detector properties and the like.
(2) Quantitative analysis of defect width:
When the magnetic leakage detector is axially excited, the sensors for detecting the magnetic leakage signals are circumferentially arranged, the circumferential distance of the magnetic leakage signals detected by the sensors is approximately the width dimension of the defect, the number of the sensors subjected to the defect excitation signals is close to the proportional relation between the defect width, and therefore the number of the sensors subjected to the defect excitation signals can be used as an index for quantifying the defect width, and because of the special property, the signal characteristic quantity of the circumferential component peak Gu Zhongzhi interval is generated.
And extracting the circumferential component, namely determining the maximum value position of the radial component on the axial path, namely the peak position, and then extracting the radial component on the circumferential path where the peak position is located as the circumferential component.
Circumferential components of different defect widths are obtained through finite element simulation, as shown in fig. 10, and according to simulation results, it is known that the circumferential component peak Gu Zhongzhi pitch S y-50％ of different widths is in direct proportion to the defect width, as shown in fig. 11.
As shown by finite element simulation analysis, as the defect width increases, the axial component peak-valley value and the radial component peak-valley value are monotonically increased, but the two signal characteristic quantities are influenced by other factors such as defect depth change, lift-off value fluctuation and the like. Only the signal feature quantity of the circumferential component peak Gu Zhongzhi interval can be called a stable quantity, does not change along with the change of the defect depth and the lifting distance, and is verified in finite element simulation, and the fluctuation precision error is within 10%. It should be emphasized that the characteristic parameters such as the axial component peak-to-valley pitch, the radial component peak-to-valley pitch, the axial component differential signal peak-to-valley pitch and the like remain unchanged when the defect width is changed, and correspond to the judgment of the defect length quantization parameter.
Verifying stability of a circumferential component peak Gu Zhongzhi pitch S y-50％ when quantifying defect width, setting lifting distances to be 1,2,3 and 4mm respectively, setting defect depth to be a value range of 4, 6, 8, 10 and 12mm respectively, performing magnetic leakage detection to obtain circumferential components when the defect depth and the lifting distances are respectively changed, wherein one defect width is a group, 5 groups are altogether, the defect widths are respectively 6, 8, 10, 12 and 14mm, obtaining a circumferential component peak Gu Zhongzhi pitch S y-50％ and a corresponding defect depth and lifting distance, generating a line graph according to the obtained data, and verifying that fluctuation accuracy errors of the circumferential component peak Gu Zhongzhi pitch S y-50％ are all within 10% by calculating fluctuation percentages of each line graph as shown in fig. 12 and 13.
In summary, the circumferential component peak Gu Zhongzhi pitch S y-50％ is selected as a quantization parameter of the defect width W, and linear fitting is performed on the circumferential component peak Gu Zhongzhi pitch S y-50％ and the defect width W in the sample set to obtain a defect width quantization formula:
W＝a2*Sy-50％+b2
Wherein a 2 is a width scaling factor, and b 2 is a width correction factor; a 2、b2 is a constant value in general, and can be adjusted according to specific conditions such as different working environments, pipes, excitation strength, detector properties and the like.
(3) Quantitative analysis of defect depth
According to simulation result analysis, the radial component peak-valley value and the axial component peak-valley value are in good proportional linear relation with the defect depth, increase along with the increase of the defect depth, and are suitable for being used as signal characteristic quantity for quantifying the defect depth. For example, radial components of different defect depths are obtained through finite element simulation, as shown in fig. 14, and it is known from the simulation result that radial component peak-to-valley values B zp-p of different depths are in direct proportion to defect widths, as shown in fig. 15.
Although the two signal feature values, i.e., the radial component peak-valley value and the axial component peak-valley value, are suitable as the signal feature values for quantifying the defect depth, they are also affected by the defect size variation and the lift-off fluctuation, and still belong to unstable signal feature values. However, according to the above-mentioned quantization analysis of defect length and width, the defect length can be determined from the axial component differential signal peak-to-valley pitch, the defect width dimension can be determined from the circumferential component peak Gu Zhongzhi pitch, and the defect depth can be quantized by using the radial component peak-to-valley value as the quantization parameter.
The linear fitting is carried out on the radial component peak-valley value and the corresponding defect depth, the linear fitting effect of the radial component peak-valley value and the defect depth under the defect samples with different sizes is shown in fig. 16, and the fitting straight line can represent the defect depth in the corresponding range and satisfies the following relation:
D＝a*Bzp-p+b
wherein B zp-p is the peak-valley value of the radial component, namely the difference value between the peak and the trough of the radial component, a and B are the fitting parameters of the depth of the defect, and are variables, when the length and the width of the defect are changed, the corresponding a and B are also changed, and a and B respectively represent the proportionality coefficient and the correction coefficient of the depth of the defect.
Therefore, the mapping relation between the length and width of the defect and the fitting parameters a and b of the depth of the defect can be obtained by establishing a BP neural network model.
In this embodiment, 200 groups of defects with different lengths and widths are designed, the defect depths are respectively 4mm, 6mm, 8mm, 10mm and 12mm, 1000 defects are combined together, the radial component peak-valley values B zp-p of the 200 groups of defects with different lengths and widths are fitted with the defect depth D, 200 groups of corresponding defect depth fitting parameters a and B are obtained, and 200 groups of neural network sample data are obtained, so that a neural network sample set is constructed.
The length and the width of known defects are taken as inputs of the neural network, the defect depth fitting parameters a and b are taken as outputs, and the sample data is trained by adopting a GA depth optimization BP neural network method. Substituting the predicted defect depth fitting parameters a and B and the radial component peak-to-valley value B zp-p into the above formula to obtain the predicted value of the defect depth D. And comparing the predicted defect depth value with an initial actual defect depth value, so as to judge the accuracy of defect depth prediction in the method.
In this embodiment, the neural network sample set is divided into a training set, a test set and a verification set, the number of data in the training set is 140, the number of data in the test set is 30, and the number of data in the verification set is 30. The number of nodes of the input layer of the BP neural network is 2, the number of nodes of the hidden layer is automatically optimized, and the number of nodes of the output layer is 2. The comparison result and the correlation coefficient of the predicted value of the defect depth and the actual value of the defect depth in the training set, the test set, the verification set and the neural network total sample set are shown in fig. 17, wherein R is the correlation coefficient of the predicted value of the defect depth and the actual value of the defect depth in the set, and is used for representing the prediction accuracy of the defect depth in the set, and the closer R is to 1, the closer is the prediction accuracy of the defect depth is to 100%, wherein the correlation coefficient r= 0.99801 of the training set, the correlation coefficient r= 0.99448 of the test set, the correlation coefficient r= 0.97216 of the verification set and the correlation coefficient r= 0.99312 of the neural network total sample set. The defect depth predicted value and defect depth actual value pair of part of the sample data are as shown in fig. 18. As can be seen from fig. 18, the effect of indirectly quantifying the depth of the defect is very remarkable by predicting the fitting parameters a and b of the depth of the defect, and the prediction accuracy is higher than 97%, which is higher than that of a method for quantifying the depth of the defect by directly using the peak-valley values of the radial components.
In summary, radial component peak-to-valley value B zp-p is selected as the quantization parameter of defect depth D to obtain a defect depth quantization formula:
D＝a*Bzp-p+b
The defect depth fitting parameters a and b are obtained through predicting the quantized defect length and defect width through a neural network model.
According to the method, the magnetic flux leakage signals of the pipeline under different defect sizes are simulated through finite element modeling, the mutual influence relation among the characteristic quantities of the signals, the defect length, the defect width and the defect depth is researched and analyzed, the characteristic effectiveness of the defect length, the defect width and the defect depth and the characteristic quantities of the signals can be obtained, the signal characteristic quantities can be used for representing the superiority and inferiority of geometric parameters of the defects and obtaining an optimal representation result.
The method of the invention can acquire different sample data under different working scenes, different equipment, different materials, different magnetization conditions and the like, but the whole quantitative prediction method is applicable to all occasions.
The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.
Claims (4)
1. The pipeline defect quantification method based on the magnetic leakage signal characteristics is characterized by comprising the following steps of:
s1, acquiring triaxial magnetic leakage signals of defects under different defect sizes, and constructing a sample set;
the defect sizes include: defect length, defect width, defect depth;
the triaxial magnetic leakage signal of the defect is as follows: axial magnetic leakage signals are axial components, radial magnetic leakage signals are radial components, and circumferential magnetic leakage signals are circumferential components;
The axial direction is the direction along the length of the pipeline, the radial direction is the direction vertical to the inner wall of the pipeline, and the circumferential direction is the circumferential direction of the pipeline;
The circumferential component is selected in the following way: determining the peak position of a radial component on an axial path, and selecting the value of the radial component on a circumferential path where the peak position is located as the value of the circumferential component;
S2, differentiating the axial components in the sample set to obtain an axial component differential signal, extracting the peak-to-valley spacing DS xp-p of the axial component differential signal as a quantization parameter of the defect length L, and performing linear fitting on the peak-to-valley spacing DS xp-p of the axial component differential signal in the sample set and the defect length L to obtain a defect length quantization formula:
L＝a1*DSxp-p+b1
Wherein a 1 is a length scale factor, and b 1 is a length correction factor;
S3, extracting a circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set as a quantization parameter of the defect width W, and performing linear fitting on the circumferential component peak Gu Zhongzhi interval S y-50％ in the sample set and the defect width W to obtain a defect width quantization formula:
W＝a2*Sy-50％+b2
wherein a 2 is a width scaling factor, and b 2 is a width correction factor;
The circumferential component peak Gu Zhongzhi pitch S y-50％ is extracted by: extracting peaks and troughs of the circumferential component, calculating a median value of the peaks and the troughs, namely 50% of a difference value between the peaks and the troughs, as a peak Gu Zhongzhi, and obtaining a pitch of a median value of the peaks and the troughs on the circumferential component, namely a pitch S y-50％ of peaks Gu Zhongzhi of the circumferential component;
S4, extracting a radial component peak-valley value B zp-p in the sample set, namely, a difference value between a peak and a trough of a radial component, taking the radial component peak-valley value B zp-p as a quantization parameter of the defect depth D, and fitting the radial component peak-valley value B zp-p in the sample set and the defect depth D to obtain a defect depth quantization formula:
D＝a*Bzp-p+b
Wherein a and b are defect depth fitting parameters and are variables, and when the defect length or defect width is changed, the corresponding defect depth fitting parameters a and b are also changed;
s5, constructing a neural network, wherein the input of the neural network is defect length L and defect width W, the output of the neural network is defect depth fitting parameters a and b, and training is performed by using a sample set to generate the neural network;
S6, quantifying the unknown defects as follows:
s61, calculating the defect length L of the unknown defect according to the axial component differential signal peak-valley spacing DS xp-p of the unknown defect and a defect length quantization formula;
S62, calculating the defect width W of the unknown defect according to the circumferential component peak Gu Zhongzhi interval S y-50％ of the unknown defect and a defect width quantization formula;
s63, inputting the calculated length L and width W of the unknown defect into a neural network model, and predicting to obtain defect depth fitting parameters a and b;
S64, substituting the radial component peak-valley value B zp-p of the unknown defect and predicted defect depth fitting parameters a and B into a defect depth quantization formula, and calculating to obtain the defect depth D of the unknown defect.
2. The method for quantifying pipeline defects based on magnetic flux leakage signal features according to claim 1, wherein in step S1, triaxial magnetic flux leakage signals of defects under different defect sizes are simulated by finite element software modeling.
3. The method for quantifying pipeline defects based on magnetic leakage signal features according to claim 1, wherein in step S5, the sample set is divided into a training set and a test set, and the training set is used to train the neural network:
And fitting the radial component peak-valley value B zp-p with the same defect length and defect width in the training set with the defect depth D to obtain defect depth fitting parameters a and B under the defect length and the defect width, so as to obtain training data of the neural network, and training the neural network by using the training data of the neural network.
4. A method for quantifying pipeline defects based on magnetic leakage signal features according to claim 1 or 3, wherein in step S5, the sample set is divided into a training set and a test set, and the neural network is tested by using the test set:
inputting the defect length L and the defect width W of known defects in the test set into the trained neural network, and predicting to obtain corresponding defect depth fitting parameters a and b; substituting the corresponding defect depth fitting parameters a and B and the radial component peak-to-valley value B zp-p of the defect into a defect depth quantization formula, and calculating to obtain a predicted value of the defect depth D; and comparing the predicted value of the defect depth D with the true value of the defect depth D to test the prediction accuracy of the neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210729456.6A CN115096987B (en) | 2022-06-24 | 2022-06-24 | Pipeline defect quantification method based on magnetic leakage signal characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210729456.6A CN115096987B (en) | 2022-06-24 | 2022-06-24 | Pipeline defect quantification method based on magnetic leakage signal characteristics |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115096987A CN115096987A (en) | 2022-09-23 |
| CN115096987B true CN115096987B (en) | 2024-04-26 |
Family
ID=83292192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210729456.6A Active CN115096987B (en) | 2022-06-24 | 2022-06-24 | Pipeline defect quantification method based on magnetic leakage signal characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115096987B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104514987A (en) * | 2014-12-19 | 2015-04-15 | 清华大学 | Three-dimensional pipeline flux leakage imaging defect quantizing method |
| JP2017009549A (en) * | 2015-06-26 | 2017-01-12 | コニカミノルタ株式会社 | Non destructive testing device |
| CN106770627A (en) * | 2016-12-16 | 2017-05-31 | 北京华航无线电测量研究所 | A kind of axial magnetic leakage signal length quantization method |
| CN106870957A (en) * | 2017-03-21 | 2017-06-20 | 东北大学 | A kind of feature extracting method of pipeline defect and magnetic leakage signal |
| WO2020133639A1 (en) * | 2018-12-29 | 2020-07-02 | 东北大学 | Intelligent analysis system for magnetic flux leakage detection data in pipeline |
| CN113049676A (en) * | 2021-03-09 | 2021-06-29 | 中国矿业大学 | Quantitative analysis method for pipeline defects |
-
2022
- 2022-06-24 CN CN202210729456.6A patent/CN115096987B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104514987A (en) * | 2014-12-19 | 2015-04-15 | 清华大学 | Three-dimensional pipeline flux leakage imaging defect quantizing method |
| JP2017009549A (en) * | 2015-06-26 | 2017-01-12 | コニカミノルタ株式会社 | Non destructive testing device |
| CN106770627A (en) * | 2016-12-16 | 2017-05-31 | 北京华航无线电测量研究所 | A kind of axial magnetic leakage signal length quantization method |
| CN106870957A (en) * | 2017-03-21 | 2017-06-20 | 东北大学 | A kind of feature extracting method of pipeline defect and magnetic leakage signal |
| WO2020133639A1 (en) * | 2018-12-29 | 2020-07-02 | 东北大学 | Intelligent analysis system for magnetic flux leakage detection data in pipeline |
| CN113049676A (en) * | 2021-03-09 | 2021-06-29 | 中国矿业大学 | Quantitative analysis method for pipeline defects |
Non-Patent Citations (3)
| Title |
|---|
| Defect quantitative recognition technology of circumferential magnetic flux leakage inspection in pipeline;Zhang Guoguang;《2010 International Conference on Computer Application and System Modeling (ICCASM 2010)》;20101104;全文 * |
| 油气管道缺陷漏磁检测试验;杨涛, 王太勇, 李清, 冷永刚;天津大学学报;20040825(08);全文 * |
| 长输管道漏磁内检测缺陷识别量化技术研究;田野;高涛;许光达;丁融;;油气田地面工程;20181020(10);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115096987A (en) | 2022-09-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103336049A (en) | Pulsed eddy current detection method and device for eliminating lift-off effect | |
| Kandroodi et al. | Defect detection and width estimation in natural gas pipelines using MFL signals | |
| CN102095793A (en) | Quantitative magnetic flux leakage testing method for defect at root part of butt weld of pipeline | |
| CN105891321A (en) | Calibration method for micromagnetic detection of ferromagnetic material structural mechanics performance | |
| CN113049676A (en) | Quantitative analysis method for pipeline defects | |
| Chen et al. | Three-dimensional defect reconstruction from magnetic flux leakage signals in pipeline inspection based on a dynamic taboo search procedure | |
| US12018939B2 (en) | System and method for analyzing an oilfield casing using an apparent thickness approach | |
| CN112149296A (en) | Method for judging stability type of hydrological time sequence | |
| Zhang et al. | Analysis of MFL model for sucker rod defects and its MFL signal processing | |
| CN113158558B (en) | High-speed railway roadbed continuous compaction analysis method, device and analyzer | |
| CN115096987B (en) | Pipeline defect quantification method based on magnetic leakage signal characteristics | |
| CN110220967A (en) | A kind of ACFM technology tufted crack surfaces profile automatic identifying method | |
| CN111141809B (en) | Soil nutrient ion content detection method based on non-contact type conductivity signal | |
| Peng et al. | A lift-off revision method for magnetic flux leakage measurement signal | |
| CN107178710B (en) | It is a kind of based on it is inside and outside detection signal characteristic abstraction defect of pipeline inside and outside discrimination method | |
| CN108267502B (en) | Eddy current detection system and method for depth of hardened layer | |
| CN115371536A (en) | Metal pipeline crack measuring method and system | |
| CN111197471B (en) | Transient electromagnetic detection calculation model and detection method for underground screen pipe | |
| CN111504402B (en) | Sensor parameter optimization method and device, controller and medium | |
| Zhou et al. | Structural optimisation of a standard alternating current field measurement excitation probe for concave and convex surfaces | |
| CN114740081B (en) | Micromagnetic detection method for stress distribution along depth | |
| Sukamta et al. | Prototype of Temperature, Humidity and Soil PH Measurement as a Analysis Tool Soil Resistance in Grounding System | |
| CN111272060B (en) | Method for quantifying size of opening defect in width direction of non-conductive material based on single-pair electrode capacitance imaging detection technology | |
| CN114184154B (en) | Oil and gas well casing inner diameter detection method based on random forest and direct-current magnetic field | |
| Shamgholi et al. | Proposing a method for eddy current thickness measurements of coatings on ferrous metal base material |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |