Pipeline internal detection defect quantitative feature extraction and feature selection probability determination method

Abstract

The invention provides a pipeline internal detection defect quantitative feature extraction and feature selection probability determination method, which comprises the following steps: acquiring magnetic flux leakage internal detection data of a target pipeline, and constructing a pipeline magnetic flux leakage data matrix; and intercepting the segmented magnetic flux leakage data according to the preset intercepted data volume. And performing defect labeling on axial magnetic flux leakage data, radial magnetic flux leakage data and circumferential magnetic flux leakage data in the segmented magnetic flux leakage data to obtain pipeline defect fragment data. Noise reduction is carried out on the pipeline defect fragment data, multi-layer discrete wavelet decomposition is carried out with the data corresponding to the sensor channel with the maximum peak-valley value difference as reference signal data, and an approximation coefficient and a detail coefficient corresponding to each layer are obtained. And extracting defect quantitative characteristics according to the denoised pipeline defect fragment data and the approximation coefficient, and determining the selection probability of the defect quantitative characteristics according to the detail coefficient. In this way, the importance degree of the defect quantification characteristics can be accurately determined, and the precision and efficiency of a defect quantification algorithm are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline defect detection, and in particular to a method for extracting quantitative features of defects detected in pipelines and determining the probability of feature selection. Background Art

[0002] Pipeline transportation is widely used due to its low cost, high efficiency and good safety. However, due to the long laying time, transportation medium and natural environment, some pipelines are prone to problems such as pipe wall damage, which poses a potential major danger. Scientific and reasonable pipeline fault diagnosis is the key to ensure the safe transportation of energy. Defect size is an important reference indicator for pipeline safety status assessment, so it is of great significance to achieve quantitative assessment of pipeline defect information. At present, advanced methods with deep learning and machine learning as the mainstream are becoming increasingly mature, providing a new direction for the development of pipeline fault diagnosis technology.

[0003] To achieve defect size quantification using machine learning, the first step is to accurately measure the length, width, depth and other geometric information of pipeline defects to obtain defect data for defect quantification algorithm learning. Then, signal data containing multi-directional and multi-scale defect information is obtained through magnetic flux leakage internal detection to extract the features of multiple signals and match them with the measured values ​​of actual defects to achieve size prediction of new defects.

[0004] Preprocessing of the leakage magnetic internal detection signal data can improve the signal quality, and thus improve the accuracy of the defect quantization algorithm. Improving the quality and diversity of features can improve the generalization performance of the quantization model. In the related art, the threshold method can be used to denoise the leakage magnetic internal detection signal data. Generally, the threshold method is divided into a hard threshold method and a soft threshold method. The hard threshold method may introduce artifacts or enhance the non-noise part of some signals when removing noise, which may cause the quality of the final signal to deteriorate or distort. The soft threshold method tends to introduce a smoothing effect near the signal change, which may cause the signal edge to become blurred and some detail information to be lost. In this way, it is impossible to accurately extract defect quantization features from the leakage magnetic internal detection signal data, reducing the accuracy of the defect quantization algorithm. Summary of the invention

[0005] The technical problem to be solved by the present invention is that it is impossible to accurately extract defect quantification features from magnetic flux leakage internal detection signal data, thereby resulting in low accuracy of the defect quantification algorithm.

[0006] In order to solve the above technical problems, the present invention provides a method for extracting quantitative features of defects detected in pipelines and determining the probability of feature selection, which specifically adopts the following technical solutions:

[0007] The present invention provides a method for extracting quantitative features of defects detected in pipelines and determining the probability of feature selection, comprising: first, obtaining magnetic leakage internal detection data of a target pipeline, and constructing a pipeline magnetic leakage data matrix according to the magnetic leakage internal detection data, wherein the magnetic leakage internal detection data includes magnetic leakage internal detection signal data collected by multiple sensor channels. Then, the pipeline magnetic leakage data matrix is ​​intercepted according to a preset interception data amount to obtain segmented magnetic leakage data, wherein the segmented magnetic leakage data includes axial magnetic leakage data, radial magnetic leakage data, and circumferential magnetic leakage data. Defects are marked on the axial magnetic leakage data, radial magnetic leakage data, and circumferential magnetic leakage data, respectively, and pipeline defect segment data is intercepted. Secondly, the pipeline defect segment data is denoised to obtain the pipeline defect segment data after denoising; the peak-to-valley value differences corresponding to multiple sensor channels are determined according to the pipeline defect segment data after denoising, and the pipeline defect segment data corresponding to the sensor channel with the largest peak-to-valley value difference is used as the reference signal data. Next, the reference signal data is subjected to multi-layer discrete wavelet decomposition according to the first decomposition parameter to obtain the first approximation coefficient and the first detail coefficient corresponding to each layer. Finally, the defect quantitative features are extracted according to the denoised pipeline defect fragment data and the first approximate coefficient corresponding to each layer, and the selection probability of the defect quantitative features is determined according to the first detail coefficient corresponding to each layer.

[0008] In an optional implementation, the defect quantification feature includes one or more of the following features: peak value, valley value, peak-to-valley difference, peak-to-peak spacing, length threshold, width threshold, surface energy, and volume energy.

[0009] In an optional implementation, the expression of the above surface energy is:

[0010]

[0011] Among them, S represents the surface energy, X l Indicates the left valley point of the pipeline defect, X r is the right valley value point of the pipeline defect, F D is the magnetic field strength curve of the sensor, F rl It is the straight line formed by connecting the left valley value point and the right valley value point.

[0012] In an optional implementation, the probability of selecting the defect quantization feature is determined according to the first detail coefficient corresponding to each layer, including: first, determining the inflection point of the curve corresponding to the first detail coefficient according to the difference value of the first detail coefficient corresponding to each layer. Then, determining the probability of selecting the defect quantization feature according to the inflection point of the curve.

[0013] In an optional implementation, the defect quantification feature includes: the original signal feature and the decomposition feature corresponding to each layer; wherein the selection probability expression of the decomposition feature is:

[0014] Pfn =a n ×(1-F kn );

[0015] in,

[0016]

[0017] Where P fn represents the probability of selecting the decomposition feature corresponding to the nth layer, n is an integer greater than or equal to 1, a n is the preset reference probability coefficient corresponding to the nth layer, F kn Indicates the oscillation frequency corresponding to the nth layer, F ni Indicates the number of inflection points of the curve corresponding to the nth layer, D n Represents the length of the first detail coefficient corresponding to the nth layer. The probability expression of the selection of the original signal feature is:

[0018] P O =1-(P f1 +P f2 +…+P fn );

[0019] Where P O Represents the probability of selecting the original signal feature, P f1 Represents the selection probability of the decomposition feature corresponding to the first layer, P f2 represents the selection probability of the decomposition feature corresponding to the second layer, P fn Represents the selection probability of the decomposition feature corresponding to the nth layer.

[0020] In an optional implementation, the pipeline defect segment data is denoised to obtain the pipeline defect segment data after denoising, specifically including: first, performing multi-layer discrete wavelet decomposition on the reference signal data according to the second decomposition parameter to obtain the second approximate coefficient and the second detail coefficient corresponding to each layer. Then, the second detail coefficient is denoised according to the preset threshold function by the threshold method to obtain the second detail coefficient after denoising. Finally, the second detail coefficient and the second approximate coefficient after denoising are inversely transformed to obtain the pipeline defect segment data after denoising.

[0021] In an optional implementation, the expression of the second detail coefficient after noise reduction is:

[0022]

[0023] in,

[0024]

[0025]

[0026] In the formula, y represents the second detail coefficient after denoising, and x represents the second detail coefficient of each layer before denoising. represents the preset threshold function, α represents the preset variable parameter, γ represents the average value of the second detail coefficient corresponding to each layer, N represents the length of the original signal data, D n Indicates the length of the second detail coefficient corresponding to the nth layer.

[0027] In an optional implementation, the first decomposition parameter includes: coefficient lengths of the first approximate coefficient and the first detail coefficient, and a first decomposition layer number. Then, performing multi-layer discrete wavelet decomposition on the reference signal data according to the first decomposition parameter includes: performing multi-layer discrete wavelet decomposition on the reference signal data according to the coefficient length and the first decomposition layer number through a wavelet basis function.

[0028] In an alternative implementation, the expression for the coefficient length is:

[0029] L n =floor((a+b*(n-1))*L n-1 );

[0030] Where, L n It represents the coefficient length corresponding to the nth layer, floor means rounding down the logarithm, and a and b are preset constants.

[0031] In an optional implementation, the peak-to-valley value differences corresponding to the multiple sensor channels are determined according to the pipeline defect segment data after noise reduction, specifically including: first, outlier removal and missing value compensation are performed on the pipeline defect segment data after noise reduction to obtain the pipeline defect segment data after data preprocessing. Then, the peak-to-valley value differences corresponding to the multiple sensor channels are determined according to the pipeline defect segment data after data preprocessing.

[0032] In summary, the method for extracting the quantitative features of internal detection defects and determining the probability of feature selection provided by this application uses a discrete wavelet analysis method to implement soft threshold noise reduction on the defect feature area signal, which improves the restoration degree of the leakage magnetic internal detection signal and reduces the influence of signal distortion caused by noise reduction, so that the importance of the defect quantitative features can be accurately determined. At the same time, based on the importance of the defect quantitative features, a method for determining the selection probability of the defect quantitative features when using the defect quantization algorithm is provided, which effectively improves the accuracy and efficiency of the defect quantization algorithm and optimizes the inversion efficiency of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A flow chart of a method for extracting quantitative features of defects detected in a pipeline and determining the probability of feature selection provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application as detailed in the claims.

[0035] Pipeline transportation is widely used due to its low cost, high efficiency and good safety. However, due to the long laying time, transportation medium and natural environment, some pipelines are prone to problems such as pipe wall damage, which poses a potential major danger. Leakage of oil pipelines not only brings huge economic losses to industrial production and people's lives, but also causes serious damage to the environment and affects the health and life safety of residents along the line. High-precision pipeline defect inversion helps to accurately evaluate the defect shape and provide quantitative analysis and data support for the pipeline operation status. This enables staff to identify serious defects in a timely manner and replace them, thereby reducing the significant economic losses and environmental pollution that may be caused by leakage accidents. At the same time, for defects that are not serious and will not cause leakage, unnecessary pipeline replacement is avoided, providing accurate prediction capabilities for pipeline operation and maintenance.

[0036] Scientific and reasonable pipeline fault diagnosis is the key to ensure the safe transportation of energy. Defect size is an important reference indicator for pipeline safety status assessment, so it is of great significance to achieve quantitative assessment of pipeline defect information. At present, advanced methods with deep learning and machine learning as the mainstream are becoming increasingly mature, providing a new direction for the development of pipeline fault diagnosis technology.

[0037] To achieve defect size quantification using machine learning, the first step is to accurately measure the length, width, depth and other geometric information of pipeline defects to obtain defect data for defect quantification algorithm learning. Then, signal data containing multi-directional and multi-scale defect information is obtained through magnetic flux leakage internal detection to extract the features of multiple signals and match them with the measured values ​​of actual defects to achieve size prediction of new defects.

[0038] Preprocessing of magnetic flux leakage internal detection signal data can improve signal quality, thereby improving the accuracy of defect quantization algorithm. Improving the quality and diversity of features can improve the generalization performance of quantization model. Since pipelines are usually in complex environments, there are also many noise disturbances in magnetic flux leakage internal detection signals. Among them, wavelet denoising method is widely used in magnetic flux leakage internal detection signal denoising due to its advantages such as multi-resolution analysis, energy concentration, adaptability, time-frequency localization and computational efficiency. Wavelet threshold denoising reduces the low-amplitude component in the signal by setting threshold processing on the low-amplitude signal, and then restores the magnetic flux leakage internal detection signal by reconstruction. Generally, threshold method is divided into hard threshold method and soft threshold method. The hard threshold method may introduce artifacts or enhance the non-noise part of some signals when removing noise, which may cause the quality of the final signal to deteriorate or distort. The soft threshold method tends to introduce a smoothing effect near the signal change, which may cause the signal edge to become blurred and lose some detail information. In this way, it is impossible to accurately extract defect quantification features from the magnetic flux leakage internal detection signal data, reducing the accuracy of the defect quantification algorithm.

[0039] In order to solve the above problems, the embodiment of the present application provides a method for extracting quantitative features of defects detected in pipelines and determining the probability of feature selection. The method uses a discrete wavelet analysis method to implement soft threshold noise reduction on the defect feature area signal, thereby improving the restoration degree of the leakage magnetic detection signal and reducing the influence of signal distortion caused by noise reduction, so as to accurately determine the importance of the defect quantitative features. At the same time, based on the importance of the defect quantitative features, a method for determining the probability of selecting the defect quantitative features when using the defect quantization algorithm is provided, which effectively improves the accuracy of the defect quantization algorithm and optimizes the inversion efficiency of the algorithm.

[0040] The solution provided by the embodiment of the present application is introduced below in conjunction with the accompanying drawings.

[0041] For details, see Figure 1 , is a flow chart of a method for extracting quantitative features of internal detection defects and determining the probability of feature selection provided in an embodiment of the present application, such as Figure 1 As shown, the method for extracting quantitative features of internal detection defects and determining the probability of feature selection provided by the present invention includes the following steps S101-S107:

[0042] S101, obtaining magnetic flux leakage internal detection data of a target pipeline, and constructing a pipeline magnetic flux leakage data matrix according to the magnetic flux leakage internal detection data.

[0043] In the embodiment of the present application, the magnetic leakage internal detection data of the target pipeline can be collected by a multi-sensor magnetic leakage internal detection device, that is, the magnetic leakage internal detection data includes magnetic leakage internal detection signal data collected by multiple sensor channels.

[0044] Furthermore, the pipeline magnetic leakage data matrix can be constructed based on the magnetic leakage internal detection data. Specifically, the magnetic leakage data of discrete sampling points collected by different channels of the sensor can be constructed into a column vector matrix in the following form:

[0045]

[0046] Where D represents the pipeline magnetic flux leakage data matrix, x i，j Represents the magnetic flux leakage signal data of the i-th discrete sampling point collected by the j-th sensor channel, i=1, 2, ..., m, j=1, 2, ..., n, n is the number of magnetic flux leakage sensor channels, and m is the number of discrete sampling points of each sensor channel.

[0047] S102, intercepting the pipeline magnetic flux leakage data matrix according to a preset interception data amount to obtain segmented magnetic flux leakage data.

[0048] Next, the pipeline magnetic flux leakage data matrix can be intercepted according to the preset interception data volume to obtain segmented magnetic flux leakage data for defect marking. The segmented magnetic flux leakage data includes: axial magnetic flux leakage data, radial magnetic flux leakage data and circumferential magnetic flux leakage data.

[0049] Among them, the preset intercepted data volume can be preset according to the sampling method of the device and the actual application requirements. For example, taking the collection method of using a multi-sensor magnetic leakage internal detection device as an example, 1000 discrete sampling points are taken as the preset intercepted data volume. For example, the sensor device for mileage sampling has a sampling mileage of 0.2mm, and the mileage of each intercepted data segment is 0.2*1000=200mm, that is, 20cm is an interception cycle. The same is true for the sensor device for time sampling.

[0050] S103, respectively marking defects on the axial magnetic flux leakage data, the radial magnetic flux leakage data, and the circumferential magnetic flux leakage data, and intercepting pipeline defect segment data.

[0051] Furthermore, a unified defect marking standard may be used to mark the axial magnetic flux leakage data, the radial magnetic flux leakage data and the circumferential magnetic flux leakage data in the segmented magnetic flux leakage data extracted in S102, and the defect data may be extracted from the magnetic leakage data to obtain pipeline defect segment data.

[0052] In one implementation, the defect marking standard can be preset based on prior knowledge. When marking defects on the magnetic flux leakage data, the axial magnetic flux leakage data defect marking can be mainly used, and radial magnetic flux leakage data and circumferential magnetic flux leakage data defect marking can be performed at the same mileage position and sensor serial number.

[0053] S104: De-noise the pipeline defect segment data to obtain de-noised pipeline defect segment data.

[0054] Furthermore, the pipeline defect segment data obtained in S103 may be subjected to noise reduction processing to further improve the accuracy of determining the importance of the defect quantitative features.

[0055] S105 , determining peak-to-valley value differences corresponding to a plurality of sensor channels respectively according to the pipeline defect segment data after noise reduction, and taking the pipeline defect segment data corresponding to the sensor channel with the largest peak-to-valley value difference as reference signal data.

[0056] Specifically, among all the multiple sensor channels, the pipeline defect segment data corresponding to the sensor channel with the largest peak-to-valley value difference is used as the reference signal data to extract the defect quantitative feature and determine the selection probability of the defect quantitative feature.

[0057] Among them, the peak-to-valley value difference corresponding to the sensor channel can be determined using the following expression:

[0058] F fg =F max -F min ;

[0059] In the formula, F fg Indicates the peak-to-valley difference, F max represents the maximum value of the noise-reduced pipeline defect fragment data corresponding to the sensor channel, F min It is the minimum value of the pipeline defect fragment data after noise reduction corresponding to the sensor channel.

[0060] In some embodiments, before determining the peak-to-valley value differences corresponding to the multiple sensor channels, the pipeline defect segment data after noise reduction can also be preprocessed to remove abnormal values ​​and compensate for missing values, thereby improving the accuracy of determining the importance of the defect quantitative features. Then, in S105, the peak-to-valley value differences corresponding to the multiple sensor channels are determined according to the pipeline defect segment data after noise reduction, specifically including:

[0061] Firstly, outliers are eliminated and missing values ​​are compensated for in the pipeline defect fragment data after noise reduction to obtain the pipeline defect fragment data after data preprocessing.

[0062] Then, the peak-to-valley value differences corresponding to the multiple sensor channels are determined according to the pipeline defect segment data after data preprocessing.

[0063] Specifically, outliers in the pipeline defect segment data after noise reduction can be identified in the following way:

[0064] F cal =abs(F i -F average );

[0065] The meaning of the above expression is to calculate the data F in the noise-reduced pipeline defect fragment data corresponding to the sensor channeli The average value of all values ​​on the sensor channel F average Then take the absolute value F cal , and get the discriminant value F cal If the F cal Greater than 1.5 times F average Then determine the data F i 1.5 is a given reference value, which is for illustrative purposes only. This multiple value can be preset based on the signal quality of the pipeline defect segment data after noise reduction and actual application requirements.

[0066] Furthermore, for the above abnormal value data F i The following processing can be done to remove outliers and replace them, that is, missing value compensation:

[0067] F err =0.5×(F i-1 +F i+1 );

[0068] In the formula, F err Indicates the data value of compensation, F i-1 and F i+1 Indicates that the data F i The data on both sides, that is, the outlier can be replaced by the average of the values ​​on both sides.

[0069] In one implementation, if a sensor channel has continuous abnormal values, an adjacent sensor channel may be selected as reference signal data.

[0070] S106. Perform multi-layer discrete wavelet decomposition on the reference signal data according to the first decomposition parameter to obtain a first approximation coefficient and a first detail coefficient corresponding to each layer.

[0071] In some embodiments, the first decomposition parameter includes: coefficient lengths of the first approximate coefficient and the first detail coefficient, and the first decomposition layer number. Then, performing multi-layer discrete wavelet decomposition on the reference signal data according to the first decomposition parameter includes:

[0072] The reference signal data is subjected to multi-layer discrete wavelet decomposition according to the coefficient length and the number of first decomposition layers through the wavelet basis function.

[0073] In some embodiments, the expression for coefficient length is:

[0074] L n =floor((a+b*(n-1))*L n-1 );

[0075] Where, L nIndicates the coefficient length corresponding to the nth layer. When n=0, it indicates the original reference signal data. Floor indicates that the logarithmic value is rounded down. a and b are preset constants. For example, a can be 0.6 or 0.5.

[0076] In some embodiments, the wavelet basis function may be a db4 wavelet, and the first decomposition layer number may be, for example, 3 layers.

[0077] S107, extracting defect quantitative features according to the pipeline defect segment data after denoising and the first approximate coefficient corresponding to each layer, and determining the selection probability of the defect quantitative features according to the first detail coefficient corresponding to each layer.

[0078] Finally, the defect quantitative features can be extracted based on the pipeline defect segment data after noise reduction and the first approximate coefficient corresponding to each layer. In some embodiments, the defect quantitative features may include one or more of the following features: peak value, valley value, peak-to-valley difference, peak-to-peak spacing, length threshold, width threshold, surface energy, and volume energy.

[0079] In one example, taking axial magnetic leakage data as an example, the peak value is the maximum value in the pipeline defect segment data after the axial magnetic leakage data is denoised, the valley value is the minimum value in the pipeline defect segment data after the axial magnetic leakage data is denoised, and the peak-to-valley difference is the difference between the peak value and the valley value. For the pipeline defect segment data after denoising with two maximum points, the peak-to-peak spacing is calculated as the difference in mileage corresponding to the two maximum points. The length threshold is the number of rows of the pipeline defect segment data after denoising, that is, the number of mileage direction points intercepted. The width threshold is the number of columns of the pipeline defect segment data after denoising, that is, the number of sensor channels contained in the defect area.

[0080] In yet another example, the expression of the extracted surface energy is:

[0081]

[0082] Among them, S represents the surface energy, X l Indicates the left valley point of the pipeline defect, X r is the right valley value point of the pipeline defect, F D is the magnetic field strength curve of the sensor, F rl It is the straight line formed by connecting the left valley value point and the right valley value point.

[0083] In another example, the method for extracting volume energy is: in the pipeline magnetic flux leakage data matrix, the surface energy of the column vector with the largest peak-to-valley value difference and the two column vectors on the left and right sides are calculated and summed.

[0084] In the embodiment of the present application, wavelet analysis is performed on the axial magnetic flux leakage data, radial magnetic flux leakage data and circumferential magnetic flux leakage data to obtain the first approximate coefficients of multiple layers, and the same defect quantitative feature extraction is performed on the pipeline defect fragment data after noise reduction and the first approximate coefficients of each layer. According to the number of layers of wavelet decomposition, the number of original defect quantitative features can be expanded by a corresponding multiple.

[0085] In some embodiments, a random forest algorithm may be used to invert the defect quantitative features to invert the geometric features of the pipeline defect, such as length, width, depth, etc. Specifically, the random forest algorithm randomly selects the defect quantitative features for calculation, and then derives the size of the pipeline defect.

[0086] In one implementation, a random forest model with weighted feature selection may be used, that is, the importance of defect quantification features is calculated by preprocessing defect signal data, and the weight of the probability of defect quantification feature selection is determined based on the importance.

[0087] In some embodiments, the above-mentioned determination of the selection probability of the defect quantization feature according to the first detail coefficient corresponding to each layer specifically includes:

[0088] First, the inflection point of the curve corresponding to the first detail coefficient is determined according to the differential value of the first detail coefficient corresponding to each layer.

[0089] Specifically, the inflection point corresponding to the first detail coefficient corresponding to each layer is calculated by calculating the difference series of the first detail coefficient, that is, calculating the difference between the latter value and the previous value in the first detail coefficient. Then the inflection point can be determined by the following method:

[0090] M i =(C i-1 ×C i ) (i≥2);

[0091] Where M i is the result of inflection point judgment, C i Represents the value in the differential sequence of the first detail coefficient. Specifically, the above formula is to multiply each value of the differential sequence of the first detail coefficient starting from the second value with the previous value. If the result M i If it is negative, the point is judged to be an inflection point of the first detail coefficient.

[0092] Then, the probability of selecting the defect quantitative feature is determined according to the inflection point of the curve.

[0093] In some embodiments, the defect quantification feature may include: the original signal feature and the decomposition feature corresponding to each layer. The selection probability expression of the decomposition feature is:

[0094] P fn =a n ×(1-Fkn );

[0095] in,

[0096]

[0097] Where P fn represents the probability of selecting the decomposition feature corresponding to the nth layer, n is an integer greater than or equal to 1, a n is the preset reference probability coefficient corresponding to the nth layer, F kn Indicates the oscillation frequency corresponding to the nth layer, F ni Indicates the number of inflection points of the curve corresponding to the nth layer, D n Indicates the length of the first detail coefficient corresponding to the nth layer.

[0098] The selection probability expression of the original signal feature is:

[0099] P O =1-(P f1 +P f2 +…+P fn );

[0100] Where P O Represents the probability of selecting the original signal feature, P f1 Represents the selection probability of the decomposition feature corresponding to the first layer, P f2 represents the selection probability of the decomposition feature corresponding to the second layer, P fn Represents the selection probability of the decomposition feature corresponding to the nth layer.

[0101] Exemplarily, taking the first decomposition layer number as 3 as an example, the defect quantization feature may include: original signal feature, one-layer decomposition feature, two-layer decomposition feature and three-layer decomposition feature.

[0102] Among them, taking the probability of one layer of decomposition features as an example, it can be expressed as:

[0103] P f1 =a1×(1-F k1 ).

[0104] Where P f1 It represents the selection probability of the decomposition feature corresponding to the first layer (ie, the first layer of decomposition feature), and a1 is the preset reference probability coefficient corresponding to the first layer, for example, a1 can be 0.2.

[0105] Then the probability of selecting the original signal feature can be expressed as:

[0106] P O =1-(P f1 +P f2 +P f3 ).

[0107] In some embodiments, in the above S104, the threshold method can be used to reduce the noise of the pipeline defect segment data. Specifically, the pipeline defect segment data can be subjected to wavelet decomposition to obtain the detail coefficient and approximate coefficient corresponding to each layer of the pipeline defect segment data decomposition. Then, a threshold function is set for the detail coefficient of each layer, and the detail coefficient after noise reduction is obtained according to the threshold function. The detail coefficient and approximate coefficient after noise reduction are inversely transformed to obtain the pipeline defect segment data after noise reduction. Then S104 can specifically include:

[0108] S1041. Perform multi-layer discrete wavelet decomposition on the reference signal data according to the second decomposition parameter to obtain a second approximate coefficient and a second detail coefficient corresponding to each layer.

[0109] S1042: De-noise the second detail coefficient using a threshold method according to a preset threshold function to obtain a de-noised second detail coefficient.

[0110] S1043, performing an inverse transformation on the second detail coefficient and the second approximation coefficient after denoising to obtain pipeline defect segment data after denoising.

[0111] In some embodiments, when the pipeline defect segment data is denoised using a threshold method, the expression of the second detail coefficient after denoising is:

[0112]

[0113] in,

[0114]

[0115]

[0116] In the formula, y represents the second detail coefficient after denoising, and x represents the second detail coefficient of each layer before denoising. represents the preset threshold function, α represents the preset variable parameter, γ represents the average value of the second detail coefficient corresponding to each layer, N represents the length of the original signal data, D n Indicates the length of the second detail coefficient corresponding to the nth layer.

[0117] In summary, the method for extracting the quantitative features of internal detection defects and determining the probability of feature selection provided by the above-mentioned embodiment of the present application uses a discrete wavelet analysis method to implement soft threshold noise reduction on the defect feature area signal, which improves the restoration degree of the leakage magnetic internal detection signal and reduces the influence of signal distortion caused by noise reduction, so as to accurately determine the importance of the defect quantitative features. At the same time, based on the importance of the defect quantitative features, a method for determining the selection probability of the defect quantitative features when using the defect quantization algorithm is provided, which effectively improves the accuracy and efficiency of the defect quantization algorithm and optimizes the inversion efficiency of the algorithm.

[0118] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0119] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0120] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0121] Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the general concept of this application and do not constitute a limitation on the protection scope of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without creative work belong to the protection scope of this application.

Claims

1. A method for extracting quantitative features of defects detected in pipelines and determining the probability of feature selection, characterized in that: include: Acquire magnetic flux leakage internal detection data of the target pipeline, and construct a pipeline magnetic flux leakage data matrix according to the magnetic flux leakage internal detection data, wherein the magnetic flux leakage internal detection data includes magnetic flux leakage internal detection signal data collected by multiple sensor channels; The pipeline magnetic flux leakage data matrix is ​​intercepted according to a preset intercepted data amount to obtain segmented magnetic flux leakage data, wherein the segmented magnetic flux leakage data includes: axial magnetic flux leakage data, radial magnetic flux leakage data and circumferential magnetic flux leakage data; respectively marking defects on the axial magnetic flux leakage data, the radial magnetic flux leakage data and the circumferential magnetic flux leakage data, and intercepting pipeline defect segment data; Denoising the pipeline defect segment data to obtain the denoised pipeline defect segment data; determining the peak-to-valley value differences respectively corresponding to the multiple sensor channels according to the denoised pipeline defect segment data, and taking the pipeline defect segment data corresponding to the sensor channel with the largest peak-to-valley value difference as the reference signal data; Perform multi-layer discrete wavelet decomposition on the reference signal data according to the first decomposition parameter to obtain a first approximation coefficient and a first detail coefficient corresponding to each layer; Defect quantification features are extracted according to the pipeline defect segment data after denoising and the first approximation coefficient corresponding to each layer, and the selection probability of the defect quantification features is determined according to the first detail coefficient corresponding to each layer.

2. The method according to claim 1, characterized in that The defect quantification features include one or more of the following features: Peak value, valley value, peak-to-valley difference, peak-to-peak spacing, length threshold, width threshold, surface energy, volume energy.

3. The method according to claim 2, characterized in that The expression of the surface energy is: Among them, S represents the surface energy, X l Indicates the left valley point of the pipeline defect, X r is the right valley value point of the pipeline defect, F D is the magnetic field strength curve of the sensor, F rl It is the straight line formed by connecting the left valley value point and the right valley value point.

4. The method according to any one of claims 1 to 3, characterized in that: The determining the selection probability of the defect quantization feature according to the first detail coefficient corresponding to each layer includes: Determining an inflection point of a curve corresponding to the first detail coefficient according to a differential value of the first detail coefficient corresponding to each layer; The probability of selecting the defect quantitative feature is determined according to the inflection point of the curve.

5. The method according to claim 4, characterized in that The defect quantification feature includes: the original signal feature and the decomposition feature corresponding to each layer; wherein the selection probability expression of the decomposition feature is: P fn =a n ×(1-F kn ); in, Where P fn Indicates the probability of selecting the decomposition feature corresponding to the nth layer, n is an integer greater than or equal to 1, a n is the preset reference probability coefficient corresponding to the nth layer, F kn Indicates the oscillation frequency corresponding to the nth layer, F ni Indicates the number of inflection points of the curve corresponding to the nth layer, D n Indicates the length of the first detail coefficient corresponding to the nth layer; The selection probability expression of the original signal feature is: P O =1-(P f1 +P f2 +…+P fn ); Where P O Represents the probability of selecting the original signal feature, P f1 Represents the selection probability of the decomposition feature corresponding to the first layer, P f2 represents the selection probability of the decomposition feature corresponding to the second layer, P fn Represents the selection probability of the decomposition feature corresponding to the nth layer.

6. The method according to claim 1, characterized in that The denoising of the pipeline defect segment data to obtain the denoised pipeline defect segment data specifically includes: Perform multi-layer discrete wavelet decomposition on the reference signal data according to the second decomposition parameter to obtain a second approximate coefficient and a second detail coefficient corresponding to each layer; Denoising the second detail coefficient according to a preset threshold function by using a threshold method to obtain a denoised second detail coefficient; The denoised second detail coefficient and the second approximation coefficient are inversely transformed to obtain the denoised pipeline defect segment data.

7. The method according to claim 6, characterized in that The expression of the second detail coefficient after noise reduction is: in, In the formula, y represents the second detail coefficient after denoising, and x represents the second detail coefficient of each layer before denoising. represents the preset threshold function, α represents the preset variable parameter, γ represents the average value of the second detail coefficient corresponding to each layer, N represents the length of the original signal data, D n Indicates the length of the second detail coefficient corresponding to the nth layer.

8. The method according to claim 1, characterized in that The first decomposition parameters include: coefficient lengths of the first approximation coefficient and the first detail coefficient, and a first decomposition level number; The performing multi-layer discrete wavelet decomposition on the reference signal data according to the first decomposition parameter comprises: The reference signal data is subjected to multi-layer discrete wavelet decomposition according to the coefficient length and the first decomposition layer number by using a wavelet basis function.

9. The method according to claim 8, characterized in that The expression of the coefficient length is: L n =floor((a+b*(n-1))*L n-1 ); Where, L n It represents the coefficient length corresponding to the nth layer, floor means rounding down the logarithm, and a and b are preset constants.

10. The method according to claim 1, characterized in that The step of determining the peak-to-valley value differences respectively corresponding to the plurality of sensor channels according to the pipeline defect segment data after noise reduction specifically includes: Performing outlier elimination and missing value compensation on the pipeline defect segment data after noise reduction to obtain the pipeline defect segment data after data preprocessing; The peak-to-valley value differences respectively corresponding to the plurality of sensor channels are determined according to the pipeline defect segment data after the data preprocessing.

Images