A method and apparatus for analyzing pipeline weld defects based on magnetic flux leakage detection

Abstract

This invention relates to the field of pipeline inspection technology, and discloses a method and apparatus for analyzing pipeline weld defects based on magnetic flux leakage detection. The method includes: generating and optimizing a composite excitation signal using a multi-frequency excitation source; calculating the magnetic flux leakage attenuation coefficient and scattering characteristics in conjunction with pipeline parameters; if the attenuation coefficient exceeds a preset attenuation threshold, adjusting the excitation signal for optimization; extracting the high-frequency transient and low-frequency stable characteristics of the optimized magnetic flux leakage signal to generate a preliminary feature vector of the weld defect; mapping the preliminary feature vector to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location; optimizing the preliminary feature vector to clarify ambiguous areas and obtain the depth and direction of the defect; analyzing the spatial distribution characteristics of the preliminary feature vector based on the depth and direction of the defect to determine the defect type; dynamically adjusting the excitation signal parameters for optimization; recalculating the signal attenuation coefficient; extracting the optimized scattering characteristics; and obtaining high-precision weld defect location results.

Description

Technical Field

[0001] This invention relates to the field of pipeline inspection technology, and in particular to a method and apparatus for analyzing pipeline weld defects based on magnetic flux leakage detection. Background Technology

[0002] Pipeline transportation, as a core infrastructure in the energy and chemical industries, is directly related to the national economic lifeline and public safety in terms of safety and reliability. Among these factors, the quality of pipeline welds is particularly critical, as weld defects can lead to leaks, fractures, or even catastrophic accidents. Magnetic flux leakage (MFL) testing technology, due to its non-destructive characteristics and high sensitivity to internal metal defects, has become the mainstream method for pipeline weld inspection.

[0003] In an existing magnetic flux leakage detection method, a strong magnetic field is applied to the surface of the pipe wall by excitation with DC / low-frequency AC current; a magnetic sensor array scans along the weld seam to capture the intensity and gradient of the surface magnetic flux leakage field; finally, the signal is processed to extract features, classify and evaluate them, and match the signal pattern against a defect database.

[0004] Traditional methods rely on a single excitation method, making them ill-suited for complex pipeline structures and variable working environments. This results in insufficient detection accuracy, especially when dealing with deep or minute defects, where signal capture and resolution capabilities significantly decrease, making it difficult to accurately distinguish the type and spatial location of defects. Therefore, existing technologies lead to insufficient accuracy in pipeline weld defect detection, easily resulting in misjudgments or missed detections. Summary of the Invention

[0005] This invention provides a method and apparatus for analyzing pipeline weld defects based on magnetic flux leakage detection, in order to solve the problem of insufficient accuracy in pipeline weld defect detection caused by existing technologies.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for analyzing pipeline weld defects based on magnetic flux leakage detection, comprising:

[0007] The first composite excitation signal is generated by a multi-frequency excitation source and adjusted and optimized to obtain the initial leakage magnetic signal set;

[0008] Based on the initial magnetic flux leakage signal set and the calibrated pipe material parameters, the attenuation coefficient of the magnetic flux leakage signal is calculated, and the scattering characteristics of the signal are determined.

[0009] If the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated.

[0010] High-frequency transient features and low-frequency stable features are extracted from the signal time series and spatial distribution characteristics of the optimized leakage magnetic field signal set to generate a preliminary feature vector of weld defects.

[0011] By using a pre-established signal propagation path model, the preliminary feature vectors are mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location.

[0012] If there are ambiguous areas in the spatial distribution map, the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect;

[0013] Based on the depth and direction of the defect, the spatial distribution characteristics of the preliminary feature vector are analyzed to determine the defect type, generate a detection report containing the defect type and spatial location, and dynamically adjust the excitation signal parameters to generate a second composite excitation signal.

[0014] Based on the second composite excitation signal, the signal attenuation coefficient is recalculated, the optimized scattering features are extracted, and high-precision weld defect location results are obtained.

[0015] Secondly, the present invention provides a pipe weld defect analysis device based on magnetic flux leakage detection, comprising:

[0016] Initial leakage flux signal generation module: Generates the first composite excitation signal through a multi-frequency excitation source, and adjusts and optimizes it to obtain the initial leakage flux signal set;

[0017] Magnetic leakage signal characteristic determination module: Based on the initial magnetic leakage signal set and the calibrated pipe material parameters, calculate the attenuation coefficient of the magnetic leakage signal and determine the scattering characteristics of the signal;

[0018] Leakage magnetic signal attenuation optimization module: If the attenuation coefficient of the signal exceeds the preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated;

[0019] Weld defect feature extraction module: Extracts high-frequency transient features and low-frequency stable features from the signal time series and spatial distribution characteristics of the optimized leakage magnetic signal set to generate a preliminary feature vector of weld defects;

[0020] Defect location spatial distribution module: By using a pre-established signal propagation path model, the preliminary feature vector is mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location;

[0021] Defect detection shape determination module: If there are ambiguous areas in the spatial distribution map, the scattering features and the attenuation coefficient are used to optimize the preliminary feature vector to determine the depth and direction of the defect;

[0022] The second composite excitation signal module: Based on the depth and direction of the defect, analyzes the spatial distribution characteristics of the preliminary feature vector, determines the defect type, generates a detection report containing the defect type and spatial location, and dynamically adjusts the excitation signal parameters to generate the second composite excitation signal;

[0023] Weld defect report generation module: Based on the second composite excitation signal, recalculate the signal attenuation coefficient, extract the optimized scattering features, and obtain high-precision weld defect location results.

[0024] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the pipe weld defect analysis method described in any one of the above.

[0025] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to execute the pipe weld defect analysis method described in any one of the above.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] (1) This invention successfully separates high-frequency transient features and low-frequency stable features by performing multi-layer convolution processing on the time series data and spatial distribution data collected by magnetic flux leakage detection, effectively improving the resolution and analysis depth of the defect signal. The design of multi-kernel convolution enables the detection system to simultaneously capture the dynamic changes and overall morphological features of weld defects, significantly enhancing the accuracy and robustness of detection and meeting the application requirements under complex working conditions;

[0028] (2) This invention organically combines high-frequency and low-frequency features to generate a unified feature vector, and effectively reduces the dimensionality through data reconstruction algorithms such as principal component analysis or autoencoder, which not only greatly reduces the consumption of computing resources, but also ensures the complete preservation of key defect information.

[0029] (3) The multi-layer data processing and mapping mechanism constructed in this invention not only realizes the structured management and efficient retrieval of weld defect features, but also provides a solid foundation for subsequent defect classification, location and evaluation. This method effectively distinguishes different types of defects by integrating time and space feature information, enhances the system's ability to identify defects such as cracks and corrosion, and at the same time, it intuitively displays the defect situation through a three-dimensional distribution map, which greatly improves the accuracy and reliability of pipeline weld safety inspection. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of a pipeline weld defect analysis method based on magnetic flux leakage detection provided in the first embodiment of the present invention;

[0031] Figure 2This is a schematic diagram of a pipe weld defect analysis device based on magnetic flux leakage detection provided in the second embodiment of the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the energy and chemical industries, the safety and reliability of pipeline transportation are directly related to the national economic lifeline and public safety. Among them, the quality of pipeline welds is particularly critical, as weld defects can lead to leaks, fractures, or even catastrophic accidents.

[0034] Traditional methods rely on a single excitation method, making them ill-suited for complex pipeline structures and variable working environments. This results in insufficient detection accuracy, especially when dealing with deep or minute defects, where signal capture and resolution capabilities significantly decrease, making it difficult to accurately distinguish the type and spatial location of defects. Therefore, existing technologies lead to insufficient accuracy in pipeline weld defect detection, easily resulting in misjudgments or missed detections.

[0035] To address the aforementioned issues, the following specific embodiments will provide a detailed introduction and explanation of a pipeline weld defect analysis method based on magnetic flux leakage detection provided in this application.

[0036] Reference Figure 1 The first embodiment of the present invention provides a method for analyzing defects in pipe welds based on magnetic flux leakage detection, comprising the following steps:

[0037] S101, the first composite excitation signal is generated by the multi-frequency excitation source and adjusted and optimized to obtain the initial leakage magnetic signal set;

[0038] S102, Based on the initial leakage magnetic field signal set and the calibrated pipe material parameters, calculate the attenuation coefficient of the leakage magnetic field signal and determine the scattering characteristics of the signal;

[0039] S103, if the attenuation coefficient of the leakage magnetic signal exceeds the preset attenuation threshold, then adjust the excitation signal and generate an optimized leakage magnetic signal set.

[0040] S104, extract the high-frequency transient features and low-frequency stable features from the signal time series and spatial distribution characteristics in the optimized leakage magnetic signal set to generate a preliminary feature vector of weld defects;

[0041] S105, using a pre-established signal propagation path model, the preliminary feature vector is mapped to the three-dimensional defect space location to generate a spatial distribution map containing the defect location;

[0042] S106, If there are ambiguous areas in the spatial distribution map, then the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect;

[0043] S107, Based on the depth and direction of the defect, analyze the spatial distribution characteristics of the preliminary feature vector, determine the defect type, generate a detection report containing the defect type and spatial location, and dynamically adjust the excitation signal parameters to generate a second composite excitation signal;

[0044] S108, based on the second composite excitation signal, recalculate the signal attenuation coefficient, extract the optimized scattering features, and obtain high-precision weld defect location results.

[0045] In step S101, a first composite excitation signal is generated using a multi-frequency excitation source and adjusted and optimized to obtain an initial leakage magnetic signal set, including:

[0046] S1011 generates high-frequency and low-frequency signals within a preset frequency range, and synthesizes a composite excitation signal containing high and low frequencies;

[0047] S1012, Based on the composite excitation signal, combined with the pipe wall diameter and wall thickness parameters, calculate and adjust the amplitude adjustment coefficient to obtain the amplitude-adjusted excitation signal data;

[0048] S1013, calculate the signal propagation direction based on the amplitude-adjusted excitation signal data, and optimize the transmission angle to obtain optimized excitation signal data;

[0049] S1014, Obtain the magnetic flux leakage signals of different depth regions of the pipe weld from the optimized excitation signal data, convert the collected analog signals into digital signals, and obtain the initial magnetic flux leakage signal set data.

[0050] In step S1011, a high-frequency signal and a low-frequency signal are generated within a preset frequency range to synthesize a composite excitation signal containing both high and low frequencies.

[0051] It should be noted that the Fourier transform can represent a signal in the time domain as a superposition of different frequencies, revealing the frequency structure of the signal and the amplitude and phase information of each frequency component, obtaining a frequency domain feature vector, describing the response characteristics of the signal at different frequencies, and providing basic data for defect identification and classification.

[0052] In one implementation, a composite signal can be designed by pre-setting a frequency range. Assuming pipeline inspection needs to cover a frequency range of 10Hz to 100kHz, firstly, a multi-frequency excitation source generates a signal covering this range. Then, a Fast Fourier Transform (FFT) is used to extract low-frequency (10Hz to 1kHz) and high-frequency (10kHz to 100kHz) signals from the generated signal. These two frequency components are then weighted and superimposed according to a certain amplitude ratio, ultimately synthesizing a composite excitation signal covering a wide bandwidth. The low-frequency component is mainly used for deep defect detection to ensure penetration, while the high-frequency component is mainly used for surface crack detection to improve resolution, thus achieving effective signal generation and synthesis.

[0053] In step S1012, based on the composite excitation signal and combined with the pipe wall diameter and wall thickness parameters, the amplitude adjustment coefficient is calculated and adjusted to obtain the amplitude-adjusted excitation signal data.

[0054] It should be noted that the amplitude adjustment coefficient is calculated by combining the pipe diameter and wall thickness parameters. Specifically, it is based on the pipe diameter... and wall thickness With preset threshold , After comparison, the amplitude adjustment factor is calculated, which can be obtained using the formula: , Calculation, where , These are the adjustment amplitude ratios for high-frequency and low-frequency signals, respectively. , These are parameters representing the high-frequency and low-frequency modes in the excitation signal, such as amplitude, frequency, or corresponding filter characteristic parameters, used to distinguish and process signals with different frequency components. , The gain parameter is used, and the corresponding amplitude is increased only when the parameter exceeds a threshold. Amplitude adjustment is performed separately for different frequency bands of the composite signal; that is, the amplitude ratio of the high-frequency signal is increased according to the pipe diameter, and the amplitude ratio of the low-frequency signal is increased according to the wall thickness. This meets the technical requirements of low-frequency detection of deep defects and high-frequency detection of surface defects. The adjustment operation adopts the form of multiplication factor. After separating the signal into different frequency bands through Fourier transform in the frequency domain, the amplitude of each frequency band is multiplied by the corresponding adjustment coefficient, and then the time-domain composite signal is synthesized through inverse Fourier transform, thereby ensuring the integrity of the signal and the coordination of the phase relationship, and avoiding waveform distortion. The signal sampling rate and parameter calibration must meet the system design requirements to ensure the effectiveness and stability of the adjusted signal in detection.

[0055] In one implementation, regarding amplitude adjustment, assuming a pipe diameter of 500mm and a wall thickness of 10mm, preset diameter thresholds are set at 300mm and wall thickness thresholds at 8mm. Since the diameter exceeds the thresholds, the proportion of high-frequency signal amplitude is increased to 60% to enhance surface defect detection capabilities; since the wall thickness exceeds the thresholds, the proportion of low-frequency signal amplitude is increased to 40% to improve response to deep defects. The adjusted signals are more adapted to the pipe's physical characteristics, reducing the risk of false positives and false negatives.

[0056] In step S1013, the signal propagation direction is calculated based on the amplitude-adjusted excitation signal data, and the transmission angle is optimized to obtain the optimized excitation signal data.

[0057] It should be noted that adjusting the excitation source emission angle based on the pipe weld depth alters the propagation efficiency of high-frequency and low-frequency signals in the excitation signal. Specifically, if the pipe weld depth is greater than a preset threshold, the excitation source emission angle is adjusted to enhance the propagation efficiency of the low-frequency signal; conversely, if the pipe weld depth is less than the preset threshold, the emission angle is adjusted to enhance the propagation efficiency of the high-frequency signal, resulting in the optimized excitation signal data. The signal propagation direction is derived from the amplitude and phase spatial distribution data of the excitation signal, combined with finite element simulation analysis of the phase gradient and amplitude distribution. Calculations are performed in the frequency domain, and the time-domain propagation path is obtained through inverse transformation, achieving accurate directional positioning and optimization.

[0058] In one implementation, the emission angle is optimized based on the weld depth. Assuming a weld depth of 5mm and a preset threshold of 3mm, if the depth exceeds the threshold, the emission angle of the excitation source is adjusted to 30 degrees to enhance the propagation efficiency of low-frequency signals along the pipe axis, which is beneficial for detecting deep weld defects. If the depth is less than the threshold, the angle is adjusted to 45 degrees to enhance the radial propagation of high-frequency signals, which is beneficial for identifying surface defects.

[0059] In step S1014, the magnetic flux leakage signals of different depth regions of the pipe weld are obtained from the optimized excitation signal data, and the collected analog signals are converted into digital signals to obtain the initial magnetic flux leakage signal set data.

[0060] It should be noted that the specific steps for converting the acquired analog signal into a digital signal are as follows: using a magnetic flux leakage signal acquisition device to acquire an analog magnetic flux leakage signal located at the pipe weld that can reflect the defect, then periodically measuring the analog signal at a certain sampling frequency to ensure that the high-frequency transient characteristics and low-frequency stable part of the signal can be captured, and finally mapping the sampled analog amplitude value to a finite number of digital levels to form discrete digital values, thus obtaining a digital magnetic flux leakage signal.

[0061] In one implementation, a high-sensitivity magnetic sensor is used to collect leakage magnetic field signals, and the leakage magnetic field data excited by the excitation signal is optimized. Assuming the sampling rate of the acquisition device is 1MHz, the signal is converted into a digital signal through an analog-to-digital converter. The leakage magnetic field signal intensity in a certain weld area is 0.1mT. After conversion into a digital signal, the location and depth of defects can be accurately analyzed. For example, based on the collected signal set, a 1mm deep surface crack and a 5mm deep internal defect can be clearly distinguished, improving the sensitivity and reliability of defect identification.

[0062] In step S102, based on the initial magnetic flux leakage signal set and the calibrated pipe material parameters, the attenuation coefficient of the magnetic flux leakage signal is calculated to determine the scattering characteristics of the signal, including:

[0063] S1021, Based on the initial leakage magnetic field signal set, combined with the pre-input pipe diameter, wall thickness and weld width, a high-precision three-dimensional finite element mesh is generated to obtain three-dimensional mesh data;

[0064] S1022, Based on the three-dimensional mesh data and the preset magnetic permeability and electrical conductivity parameters of the pipe material, simulate the electromagnetic interaction boundary between the excitation source and the pipe surface, calculate the electromagnetic field distribution, and obtain the initial data of the leakage magnetic field signal.

[0065] S1023, Based on the initial data of the leakage magnetic signal, calculate the propagation path of the signal in the complex pipeline structure, and determine the attenuation coefficient and scattering characteristics by combining the geometric features of the path.

[0066] In step S1021, based on the initial leakage magnetic field signal set and combined with the pre-input pipe diameter, wall thickness and weld width, a high-precision three-dimensional finite element mesh is generated to obtain three-dimensional mesh data;

[0067] It should be noted that the finite element mesh divides the geometric region of the computational model into many small and simple elements (such as triangles, tetrahedrons, hexahedrons, etc.), which are called "cells," and nodes are the vertices of the elements. Through this partitioning, complex physical problems can be transformed into numerical computation problems on discrete elements.

[0068] In one implementation, a finite element simulation method is used based on an initial magnetic flux leakage signal set. Taking ANSYS software as an example, a high-precision three-dimensional finite element mesh is generated. Taking a pipe diameter of 0.5m, wall thickness of 0.01m, and weld width of 0.005m as an example, the mesh element size is set to 0.002m to ensure that the mesh in the weld area is refined to 0.001m, generating approximately 5 million tetrahedral elements to ensure geometric accuracy.

[0069] In step S1022, based on the three-dimensional mesh data and the preset magnetic permeability and conductivity parameters of the pipe material, the electromagnetic interaction boundary between the excitation source and the pipe surface is simulated, the electromagnetic field distribution is calculated, and the initial data of the leakage magnetic field signal is obtained.

[0070] It should be noted that Maxwell's equations are mainly used to describe the electromagnetic properties of pipe materials, such as permeability and conductivity, to establish mathematical models of actual electromagnetic field environments, and to accurately simulate the electromagnetic field distribution and signal propagation process in the pipe weld area.

[0071] In one implementation, taking ANSYS software as an example, once the three-dimensional mesh data is generated, the relative magnetic permeability μ of the pipe material is used as the basis for the calculation. r = 200 and conductivity σ = 5 × 10 6 The S / m parameter assigns electromagnetic properties to the corresponding mesh region in the software, and sets the excitation source, such as a frequency of 50Hz and an AC current density of 1×10⁻⁶. 6 Using a coil with an amplitude of A / m² and Dirichlet boundary conditions, the finite element method is employed to solve Maxwell's equations to simulate the electromagnetic interaction boundary between the excitation source and the pipe surface. Then, numerical solutions to Maxwell's equations are used, combined with frequency domain analysis to obtain the spatial distribution of the electromagnetic field and initial data of leakage magnetic signals, enabling high-precision detection and evaluation of pipe weld defects.

[0072] In step S1023, based on the initial data of the leakage magnetic field signal, the propagation path of the signal in the complex pipeline structure is calculated, and the attenuation coefficient and scattering characteristics are determined by combining the geometric features of the path.

[0073] It should be noted that the scattering characteristics are specifically manifested as the amplitude and direction changes of the leakage magnetic signal in the pipeline defect area, reflecting the scattering mode of the signal at the defect. These characteristics are obtained through finite element simulation and are usually manifested as the scattering intensity distribution at different grid points. The high-dimensional feature space contains information such as signal strength and angle deviation. After dimensionality reduction by principal component analysis, the significant scattering characteristics caused by the defect can be highlighted, which can be used to accurately locate the spatial position and shape of the defect.

[0074] In one implementation, a three-dimensional path with spatial refraction, reflection, and scattering characteristics is generated by calculating the propagation path using finite element simulation and material geometry information. Its geometric features include path length, refraction bending angle, and scattering angle. Taking ANSYS software as an example, based on the electromagnetic field distribution and initial leakage magnetic field signal data of the pipe weld area, the attenuation process of the magnetic flux density is analyzed in conjunction with the geometric features of the path. Finally, the signal attenuation coefficient (e.g., 0.15 / m) and the scattering characteristics extracted through magnetic field gradient and Fourier transform are determined, completing a comprehensive quantitative analysis of the detected signal.

[0075] In step S103, if the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated, including:

[0076] S1031, if the attenuation coefficient of the leakage magnetic signal exceeds the preset attenuation threshold, the amplitude and transmission angle of the signal are adjusted to generate an adjusted parameter set, and the amplitude change and angle adjustment are recorded to obtain the first optimized parameter set.

[0077] S1032, Based on the first optimized parameter set, generate a leakage magnetic signal set, perform high-frequency sampling to capture transient features, and obtain the first leakage magnetic signal set;

[0078] S1033, extract time series data from the first leakage magnetic signal set, separate transient features and spatial distribution characteristics to obtain the first feature dataset;

[0079] S1034, optimize the first feature dataset to generate optimized time series and spatial distribution characteristics, and obtain the optimized leakage magnetic signal set.

[0080] In step S1031, if the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, the amplitude and transmission angle of the signal are adjusted to generate an adjusted parameter set, and the amplitude change and angle adjustment are recorded to obtain the first optimized parameter set.

[0081] In one implementation, if the attenuation coefficient of the leakage magnetic signal exceeds a preset threshold (e.g., 0.8), the amplitude of the excitation signal (e.g., from 100mT to 120mT) and the emission angle (e.g., from 45° to 50°) are adjusted to generate a parameter set containing an amplitude change of 20mT and an angle adjustment of 5°, thus forming the first optimized parameter set. This process ensures the accuracy and sensitivity of subsequent leakage magnetic signal acquisition by iteratively updating and recording the signal parameters.

[0082] In step S1032, a leakage magnetic signal set is generated based on the first optimized parameter set, and transient features are captured by high-frequency sampling to obtain the first leakage magnetic signal set.

[0083] In one implementation, based on the first set of optimized parameters, the signal generation module generates a set of leakage magnetic signals through an electromagnetic exciter and uses a 10kHz high-frequency sampling method to capture transient magnetic field changes caused by pipeline defects, such as signal spikes and abrupt changes, thereby obtaining a first set of leakage magnetic signals containing time series data and spatial distribution characteristics, laying the foundation for subsequent defect feature extraction and analysis.

[0084] In step S1033, time series data is extracted from the first leakage magnetic signal set, and transient features and spatial distribution characteristics are separated to obtain the first feature dataset;

[0085] In one implementation, a first set of leakage magnetic signals containing time-series data is obtained through high-frequency real-time sampling, which can reflect the location and characteristics of defects. Then, feature extraction is performed on the signal set, such as wavelet transform, to separate transient features (such as rapid changes in defect edges) and spatial distribution characteristics (such as defect depth and width).

[0086] In step S1034, the first feature dataset is optimized to generate optimized time series and spatial distribution characteristics, thereby obtaining the optimized leakage magnetic signal set;

[0087] In one implementation, the first feature dataset is subjected to dimensionality reduction and noise filtering to optimize the smoothness of the time series and the clarity of transient peaks, while enhancing the accuracy and stability of spatial distribution characteristics. This generates an optimized set of magnetic flux leakage signals containing better time series and spatial distribution characteristics, providing high-quality input for subsequent weld defect identification and evaluation.

[0088] In step S104, high-frequency transient features and low-frequency stable features are extracted from the signal time series and spatial distribution characteristics of the optimized leakage magnetic field signal set to generate a preliminary feature vector of weld defects, including:

[0089] S1041, Obtain time series data and spatial distribution data from the optimized leakage magnetic signal set, perform convolution operation on the time series data, extract high-frequency transient features, and obtain a first high-frequency feature set;

[0090] S1042, Based on the first high-frequency feature set, separate the low-frequency stable features of the spatial distribution data to obtain the first low-frequency feature set;

[0091] S1043, if the feature dimensions of the first high-frequency feature set and the first low-frequency feature set meet the preset first dimension threshold, then the first high-frequency feature set and the first low-frequency feature set are merged to generate a first feature vector.

[0092] S1044, Dimensionality reduction is performed on the first feature vector to generate a preliminary feature vector for weld defects.

[0093] In step S1041, time series data and spatial distribution data are obtained from the optimized leakage magnetic signal set, and convolution operation is performed on the time series data to extract high-frequency transient features and obtain a first high-frequency feature set.

[0094] In one implementation, a sensor array is uniformly arranged on the weld surface at high spatial resolution (e.g., 2 mm) to collect spatiotemporal distribution data. This data is then combined with a time-series signal acquired at a high sampling frequency (e.g., 100 kHz) to form an original data matrix containing the temporal and spatial variations of the signal. When performing convolution operations on the time-series data, a convolution kernel of appropriate size (e.g., 3×3) is first designed to capture high-frequency transient features. Then, the convolution kernel slides along the time-series data, and convolution operations are performed at each sliding position to extract local feature responses. A multi-layer convolutional network structure is used to progressively extract feature channels of different scales (e.g., 128 channels), each corresponding to a transient mode. Finally, after processing, a first high-frequency feature set reflecting the high-frequency changes of the signal is generated to characterize the dynamic characteristics of weld defects.

[0095] In step S1042, based on the first high-frequency feature set, the low-frequency stable features of the spatial distribution data are separated to obtain the first low-frequency feature set;

[0096] In one implementation, when inspecting the same pipe weld, low-frequency features may correspond to a large range of material thickness variations or uniform corrosion areas on the weld surface. First, the acquired magnetic field spatial distribution data is subjected to a Fourier transform, converting it from the time or spatial domain to the frequency domain. By setting a low-frequency threshold, frequency components below this threshold are filtered out; the features corresponding to these low-frequency components are the low-frequency features. Then, the filtered low-frequency components are subjected to an inverse Fourier transform, converting them back to the spatial domain, thus obtaining the first low-frequency feature set. The processed first low-frequency feature set may contain 64 feature channels, describing the stable spatial distribution pattern of the magnetic field.

[0097] In step S1043, if the feature dimensions of the first high-frequency feature set and the first low-frequency feature set meet the preset first dimension threshold, then the first high-frequency feature set and the first low-frequency feature set are merged to generate a first feature vector.

[0098] It should be noted that the specific steps for merging the first high-frequency feature set and the first low-frequency feature set include: firstly, determining whether the feature dimensions of the two meet a preset threshold (e.g., the total dimension after merging does not exceed 256); if so, concatenating the first high-frequency feature set (e.g., 128 dimensions) and the first low-frequency feature set (e.g., 64 dimensions) in terms of feature dimensions to form a unified first feature vector (e.g., 192 dimensions); this fused vector integrates transient high-frequency features and stable low-frequency features, which can more comprehensively describe the characteristics of weld defects, facilitating subsequent dimensionality reduction processing and defect identification.

[0099] In step S1044, the first feature vector is subjected to dimensionality reduction processing to generate a preliminary feature vector for weld defects.

[0100] It should be noted that the specific steps of dimensionality reduction include: first, calculating the covariance matrix of the first feature dataset and analyzing the correlation between features of each dimension; then, by solving the eigenvalues ​​and eigenvectors of the covariance matrix, selecting the first few principal components that can retain most of the data variance (such as more than 90%); next, using these principal components to construct a low-dimensional subspace, projecting the high-dimensional features onto this subspace to achieve dimensionality compression, and finally obtaining the second feature vector set after dimensionality reduction.

[0101] In one implementation, the 192-dimensional feature vector is reduced to a preliminary 32-dimensional feature vector, retaining key defect information such as crack depth or corrosion area. This dimensionality reduction operation reduces computational cost while preserving crucial features, facilitating subsequent defect classification or localization.

[0102] In step S105, the preliminary feature vector is mapped to the three-dimensional defect spatial location using a pre-established signal propagation path model, generating a spatial distribution map containing the defect location.

[0103] It should be noted that the training process of the pre-established signal propagation path model includes:

[0104] S1051, Initialize the parameters of the signal propagation path model and the initial propagation path data;

[0105] S1052, Based on the initial propagation path data and the pre-stored pipe geometric and material properties, perform finite element simulation and signal propagation simulation to obtain the preliminary propagation trajectory and preliminary scattering characteristics;

[0106] S1053, Based on the preliminary propagation trajectory and the preliminary scattering characteristics, construct a first scattering feature set;

[0107] S1054, Dimensionality reduction processing is performed on the first scattering feature set to generate a second feature vector;

[0108] S1055, Based on the second feature vector, update the spatial location of the defect and adjust the signal propagation path;

[0109] S1056, when the number of training iterations is greater than or equal to the preset maximum number of training iterations, the training is considered complete, and the optimized signal propagation path model is obtained.

[0110] In step S1051, the parameters of the signal propagation path model and the initial propagation path data are initialized;

[0111] It should be noted that the parameters for initializing the signal propagation path model mainly include the geometric parameters of the pipe and weld (such as pipe diameter, wall thickness, weld width, and mesh element size), material parameters (such as relative permeability, conductivity, sound velocity, and density), excitation source parameters (such as excitation current density, frequency, excitation location, and boundary conditions), as well as the attenuation coefficient and scattering characteristics during signal propagation. These parameters together constitute the finite element simulation environment, ensuring accurate simulation of the propagation path and scattering behavior of the magnetic field signal in the weld and pipe materials. The initial propagation path data is based on a composite excitation signal set, combined with the geometric parameters, material properties, and boundary conditions of the pipe and weld, and describes the preliminary data of the signal propagation path in the pipe and weld material, i.e., the initial propagation path data.

[0112] In step S1052, based on the initial propagation path data and the pre-stored pipe geometric and material properties, finite element simulation and signal propagation simulation are performed to obtain the preliminary propagation trajectory and preliminary scattering characteristics.

[0113] It should be noted that, based on the initial propagation path data, combined with pre-stored pipe geometric properties (such as pipe diameter, wall thickness, weld width, and mesh generation accuracy) and material properties (such as magnetic permeability, electrical conductivity, sound velocity, and density), a high-precision three-dimensional finite element model is constructed using finite element simulation software. By setting excitation source parameters and boundary conditions, the propagation process of the signal in the weld and pipe material is simulated, the propagation trajectory of the signal and the reflection and scattering phenomena caused by defects during the propagation process are calculated, and then the preliminary propagation trajectory and preliminary scattering characteristics are extracted.

[0114] In step S1053, a first scattering feature set is constructed based on the preliminary propagation trajectory and the preliminary scattering features;

[0115] It should be noted that, based on the preliminary propagation trajectory and preliminary scattering characteristics, the steps for constructing the first scattering feature set are as follows: First, the weld area is divided into multiple small grid units, and the signal scattering intensity, direction and propagation path information corresponding to each unit are extracted to form a high-dimensional data set; then, these data are organized and encoded according to spatial location and signal characteristics to construct the first scattering feature set containing multi-channel and multi-dimensional information.

[0116] In step S1054, the first scattering feature set is subjected to dimensionality reduction processing to generate a second feature vector;

[0117] In one solution, principal component analysis (PCA) is used to reduce the dimensionality of the first scattering feature set. The specific steps include calculating the covariance matrix of the feature set, extracting the principal components, and selecting several principal components whose cumulative contribution rate reaches a preset threshold (e.g., above 90%). This compresses the high-dimensional first scattering feature set into a low-dimensional second eigenvector. For example, reducing the 256-dimensional first scattering feature set to a 48-dimensional second eigenvector not only preserves key defect information but also significantly reduces the data dimensionality, facilitating subsequent defect localization and classification analysis.

[0118] In step S1055, the defect spatial location is updated and the signal propagation path is adjusted based on the second feature vector;

[0119] It should be noted that, based on the second feature vector, the specific steps for updating the defect spatial location and adjusting the signal propagation path are as follows: First, the dimensionality-reduced feature vector is mapped to the defect space through three-dimensional mapping to determine the specific location coordinate set of the defect; then, combined with the propagation trajectory and scattering characteristics obtained from finite element simulation, a three-dimensional distribution map containing the defect location is generated using spatial interpolation to update the defect spatial location; finally, based on the updated defect distribution information, the transmission parameters and refraction and reflection laws in the path are corrected to improve the accuracy of the subsequent signal propagation path and the accuracy of defect location, forming a closed-loop optimization process.

[0120] In step S1056, when the number of training iterations is greater than or equal to the preset maximum number of training iterations, the training is determined to be complete, and the optimized signal propagation path model is obtained.

[0121] It should be noted that when the number of training iterations is greater than or equal to the preset maximum number of training iterations, the training process is considered complete. Based on the parameters that are continuously updated and optimized during the training period, the system uses the final composite excitation signal set to recalculate the signal propagation path and attenuation coefficient through finite element simulation, and extracts the final scattering features to construct an optimized signal propagation path model.

[0122] In one implementation, the preliminary feature vector is mapped to the defect space using a trained signal propagation path model to determine the three-dimensional location coordinate set of the defect. This can be represented as follows: During detection, the 48-dimensional feature vector may be mapped onto the pipe surface to generate a set of coordinates with a spatial resolution accurate to 1 mm, marking the specific location of the crack or corrosion. Based on the three-dimensional location coordinate set of the defect, the Kriging spatial interpolation algorithm is used to interpolate the magnetic field strength data of 50 locations covering the weld area, generating a three-dimensional defect spatial distribution map with a size of 200 mm × 100 mm × 50 mm with a spatial resolution of 1 mm × 1 mm. This clearly shows the spatial distribution characteristics of the crack extending 5 mm along the pipe axis and varying in depth within the range of 3 mm to 7 mm.

[0123] In step S106, if there are ambiguous regions in the spatial distribution map, the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect, including:

[0124] S1061, Generate preliminary feature vectors for the fuzzy regions in the spatial distribution map;

[0125] S1062, Based on the preliminary feature vector, combined with the scattering characteristics and the attenuation coefficient, update the probability distribution to obtain the optimized feature vector;

[0126] S1063, Based on the optimized feature vector, perform three-dimensional reconstruction and combine it with spatial positioning technology to determine the defect depth and defect direction.

[0127] In step S1061, preliminary feature vectors are generated for the fuzzy regions in the spatial distribution map;

[0128] It should be noted that the existence of ambiguous regions is due to the presence of uncertain or interfering information in the preliminary feature vector. For example, the roughness of the pipe surface and the complexity of the defect shape make it difficult to accurately extract the scattering characteristics and attenuation coefficient of the ultrasonic signal, resulting in unclear or ambiguous areas in the defect location distribution map. Therefore, a Bayesian inference algorithm is needed to resolve these unclear or ambiguous areas. Combining the scattering characteristics and attenuation coefficient of the signal, the acquired ultrasonic or magnetic flux leakage detection signal is first subjected to a Fourier transform to obtain the amplitude and phase information in the frequency domain. This allows the extraction of scattering characteristics reflecting the intensity and direction of signal reflection in the defect area. Simultaneously, the energy loss of the signal during propagation is calculated to form the attenuation coefficient. Subsequently, the scattering characteristics and the attenuation coefficient are combined to obtain a preliminary feature vector containing the defect properties.

[0129] In one implementation, for ambiguous areas in the spatial distribution map, ultrasonic signals from the area are collected, and Fourier transform is used to convert the 10MHz time-domain signal into frequency-domain features, extracting a preliminary feature vector containing 128 dimensions, which serves as the basis for subsequent optimization of defect depth and direction using Bayesian inference algorithms.

[0130] In step S1062, based on the preliminary feature vector, combined with the scattering characteristics and the attenuation coefficient, the probability distribution is updated to obtain the optimized feature vector;

[0131] It should be noted that, in combination with the scattering characteristics and the attenuation coefficient, the optimization process of updating the probability distribution is as follows: Bayesian inference algorithm is used to continuously update, thereby optimizing the feature vector to accurately describe the depth and direction of the defect, and obtaining the probability distribution describing the uncertainty of the defect's location, depth and direction.

[0132] In one implementation, if there are ambiguous regions in the initial feature vector, such as signal interference caused by rough pipe surfaces or complex defect shapes, a Bayesian inference algorithm can be used for optimization. For example, assuming that some dimensions of the initial feature vector are blurred due to noise interference, Bayesian inference can use known acoustic properties of the steel and the distribution of defect types to calculate the posterior probability and generate an optimized feature vector. This optimization process can refine the 128-dimensional vector into 64 dimensions, highlighting key defect-related features and facilitating subsequent localization.

[0133] In step S1063, three-dimensional reconstruction is performed based on the optimized feature vector, and the defect depth and defect direction are determined by combining spatial positioning technology.

[0134] It should be noted that the specific steps of the three-dimensional reconstruction include: based on the optimized preliminary feature vector (such as dimensionality reduction to 64 dimensions through Bayesian inference), using the triangulation method and combining the geometric information of the pipeline, calculating the specific location coordinates of the defect in three-dimensional space.

[0135] In one implementation, the depth and direction of defects are determined through 3D reconstruction based on optimized feature vectors and spatial positioning technology. For example, if a defect is determined to be 5mm deep and 15 degrees axially offset from the pipe, its 3D location is accurately described using spatial coordinate mapping. Further, data fusion technology integrates the defect depth, direction, and distribution map to generate a high-precision defect location distribution map. This distribution map typically uses a gridded approach (e.g., dividing the pipe surface into 1mm × 1mm areas) to visually display the specific location and spatial distribution of cracks or corrosion.

[0136] In step S107, based on the depth and direction of the defect, the spatial distribution characteristics of the preliminary feature vector are analyzed to determine the defect type, generate a detection report containing the defect type and spatial location, and dynamically adjust the excitation signal parameters to generate a second composite excitation signal, including:

[0137] S1071, Analyze the inspection report, obtain defect location distribution data, process the defect coordinates and depth information in the defect location distribution data, and obtain first defect distribution data;

[0138] S1072, Calculate the frequency range of the excitation signal based on the first defect distribution data, and determine the first frequency range data by analyzing the material properties and signal attenuation characteristics in the first defect distribution data;

[0139] S1073, Perform signal propagation efficiency simulation on the first frequency range data, adjust the amplitude of the excitation signal, and obtain the first amplitude adjustment parameter;

[0140] S1074, based on the first amplitude adjustment parameter and the first frequency range data, calculate the signal propagation path in the first defect distribution data, perform angle calibration, and generate the second composite excitation signal.

[0141] In step S1071, the detection report is analyzed to obtain defect location distribution data. The defect coordinates and depth information in the defect location distribution data are processed to obtain first defect distribution data.

[0142] It should be noted that the specific steps for generating the inspection report include: combining defect classification results and spatial positioning technology to determine the type, depth, and direction of the defect; integrating multi-dimensional information about the defect to form a high-precision defect location distribution map; generating a heat map of the defect distribution to visually display the spatial distribution of defects in the pipeline weld, facilitating maintenance personnel to quickly locate and determine the problem area; and analyzing the safety risks of the defect by combining pipeline operating status data (such as pressure and temperature) to generate a comprehensive inspection report that includes the defect type, location, and impact assessment, providing a scientific basis for subsequent maintenance decisions.

[0143] In addition, the process of processing the defect coordinates and depth information in the defect location distribution data includes: clustering and analyzing the defect coordinates and depth to determine the concentrated distribution of the defect depth region; generating a high-precision three-dimensional defect distribution map through spatial interpolation and smoothing; and optimizing defect location and visualization by combining material properties and signal attenuation characteristics, ultimately forming the first defect distribution data that facilitates subsequent maintenance decisions.

[0144] In one implementation method, after extracting multiple defect coordinates such as (10,5,3), (12,6,4), and (15,5,2) and their corresponding depth information from the inspection report, the K-means clustering algorithm is used to set the number of clusters to 2. These defect coordinates and depth data are analyzed to determine the concentrated areas of defects at depths of 2-3 mm and 4 mm, thereby forming the first defect distribution data to reflect the spatial distribution characteristics of the defects.

[0145] In step S1072, the frequency range of the excitation signal is calculated based on the first defect distribution data, and the first frequency range data is determined by analyzing the material properties and signal attenuation characteristics in the first defect distribution data.

[0146] In one implementation method, based on the first defect distribution data, the frequency range of the excitation signal is calculated using a frequency response function. Based on the first defect distribution data, for example, cluster analysis reveals that the defects in the weld are mainly concentrated in the 2-4 mm depth region. Combining the acoustic impedance of the stainless steel material and the signal attenuation characteristics of ultrasonic waves within this depth range, the frequency response function is used to calculate the signal propagation efficiency at different frequencies, determining the optimal frequency range of the excitation signal to be 1-5 MHz. Specifically, higher frequencies (e.g., 3-5 MHz) are used to improve resolution for shallow defects, while lower frequencies (e.g., 1-3 MHz) are used to enhance penetration for deep defects, thereby achieving effective coverage and accurate detection of defects.

[0147] In step S1073, the signal propagation efficiency of the first frequency range data is simulated, and the amplitude of the excitation signal is adjusted to obtain the first amplitude adjustment parameter;

[0148] In one implementation method, based on the first frequency range data, the propagation and attenuation of the signal in the material and defects are simulated using the finite element method through signal propagation efficiency simulation, and the energy loss of the signal at different frequencies and paths is evaluated. It is found that the excitation signal suffers an energy loss of about 20% during propagation due to reflection from material boundaries. Therefore, the amplitude of the excitation signal is dynamically adjusted from the initial 100mV to 120mV to compensate for the energy loss and improve the signal detection effect in the defect area, thereby completing the determination of the first amplitude adjustment parameter.

[0149] In step S1074, the signal propagation path in the first defect distribution data is calculated based on the first amplitude adjustment parameter and the first frequency range data, and angle calibration is performed to generate the second composite excitation signal.

[0150] In one implementation method, based on a first amplitude adjustment parameter (e.g., adjusting the amplitude from 100mV to 120mV) and a first frequency range data (e.g., 3MHz), a signal propagation path model based on finite element simulation is applied to the defect coordinates (e.g., (10,5,3), (12,6,4)mm) obtained by clustering in the first defect distribution data. The propagation trajectory and attenuation characteristics of the signal in the material are numerically calculated. After judging the signal loss by combining the propagation efficiency simulation results, the incident angle of the ultrasonic probe is then adjusted from the initial 30 degrees to 45 degrees through an angle calibration mechanism to optimize the signal coverage. Finally, a second composite excitation signal set with higher detection efficiency and penetration power for the defect depth region is generated.

[0151] In step S108, based on the second composite excitation signal, the signal attenuation coefficient is recalculated, the optimized scattering features are extracted, and high-precision weld defect location results are obtained, including:

[0152] S1081, Based on the second composite excitation signal set, combined with the preset mesh division accuracy and boundary condition settings, perform finite element simulation to calculate the propagation path of the signal in the weld material and obtain the first propagation path data;

[0153] S1082, Based on the first propagation path data and combined with material properties, calculate the energy loss during signal propagation to obtain the first attenuation coefficient data;

[0154] S1083, if the first attenuation coefficient data is lower than the preset attenuation threshold, then extract the scattering features of the first propagation path data to obtain the first scattering feature data.

[0155] S1084, Based on the first scattering feature data and combined with the distribution information of the defect depth region, calculate the precise coordinates of the defect in the weld to obtain the high-precision weld defect location result.

[0156] In step S1081, based on the second composite excitation signal set and combined with the preset mesh generation accuracy and boundary condition settings, finite element simulation is performed to calculate the propagation path of the signal in the weld material and obtain the first propagation path data.

[0157] In one implementation, during finite element simulation based on a composite excitation signal set in a weld defect detection scenario, the following steps are taken: Assuming carbon steel material parameters (sound velocity 5900 m / s, density 7850 kg / m³), and combining a preset mesh refinement accuracy (e.g., 0.5 mm refined mesh) and free boundary conditions, numerical simulation is performed using finite element software to calculate the propagation path of the excitation signal in the weld. By solving the wave equation, the propagation, reflection, and refraction processes of the signal within the material are simulated, obtaining signal intensity and propagation time information, and ultimately generating first propagation path data reflecting the signal path and energy distribution.

[0158] In step S1082, based on the first propagation path data and combined with material properties, the energy loss during signal propagation is calculated to obtain the first attenuation coefficient data;

[0159] In one implementation, based on the first propagation path data and combined with the acoustic impedance, density, and elastic modulus of the material, the finite element method is used to calculate the propagation process of the signal in the weld material. By analyzing the curve of signal intensity change with propagation distance, the energy attenuation on different path segments is calculated. For example, the simulation shows that the signal intensity attenuates from the initial 100% to 80% after propagating for 5mm. Based on this, the attenuation coefficient is calculated to be 0.045dB / mm using the formula, forming the first attenuation coefficient data.

[0160] In step S1083, if the first attenuation coefficient data is lower than a preset attenuation threshold, the scattering features of the first propagation path data are extracted to obtain the first scattering feature data.

[0161] In one implementation, if the first attenuation coefficient data is lower than a preset attenuation threshold (e.g., 0.3 dB / mm), the signal propagation trajectory is extracted from the first propagation path data. Finite element simulation is used to calculate its propagation and scattering characteristics in the material. Then, when using a k-means clustering algorithm to extract scattering features from the first propagation path data, features such as signal intensity and angle deviation in the propagation path can be used as input. The cluster number is set to 3 to distinguish different types of scattering modes. For example, the clustering result may indicate that defects near the weld surface produce high-frequency scattering, while internal defects cause low-frequency scattering, thus obtaining first scattering feature data reflecting the local characteristics of the defects.

[0162] In step S1084, based on the first scattering feature data and combined with the distribution information of the defect depth region, the precise coordinates of the defect in the weld are calculated to obtain the high-precision weld defect location result.

[0163] In one implementation, when locating defects based on the first scattering feature data, geometric positioning can be performed, combined with the distribution information of the defect depth region. Assuming the inspection report shows that the defect depth is concentrated between 1-4 mm, analysis of the scattering feature data determines that the scattering signal of a certain defect mainly comes from a region with a depth of 2 mm. Using a geometric optics model, the incident and reflection angles of the signal are calculated, and the precise coordinates of the defect are derived, such as (11, 5, 2) mm. Finally, the location result of the first weld defect is obtained.

[0164] In summary, this invention discloses a method for analyzing pipeline weld defects based on magnetic flux leakage detection, comprising:

[0165] The first composite excitation signal is generated by a multi-frequency excitation source and adjusted and optimized to obtain the initial leakage magnetic signal set;

[0166] Based on the initial magnetic flux leakage signal set and the calibrated pipe material parameters, the attenuation coefficient of the magnetic flux leakage signal is calculated, and the scattering characteristics of the signal are determined.

[0167] If the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated.

[0168] High-frequency transient features and low-frequency stable features are extracted from the signal time series and spatial distribution characteristics of the optimized leakage magnetic field signal set to generate a preliminary feature vector of weld defects.

[0169] By using a pre-established signal propagation path model, the preliminary feature vectors are mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location.

[0170] If there are ambiguous areas in the spatial distribution map, the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect;

[0171] Based on the depth and direction of the defect, the spatial distribution characteristics of the preliminary feature vector are analyzed to determine the defect type, generate a detection report containing the defect type and spatial location, and dynamically adjust the excitation signal parameters to generate a second composite excitation signal.

[0172] Based on the second composite excitation signal, the signal attenuation coefficient is recalculated, the optimized scattering features are extracted, and high-precision weld defect location results are obtained.

[0173] Reference Figure 2 The second embodiment of the present invention provides a pipe weld defect analysis device based on magnetic flux leakage detection, comprising:

[0174] Initial leakage flux signal generation module: Generates the first composite excitation signal through a multi-frequency excitation source, and adjusts and optimizes it to obtain the initial leakage flux signal set;

[0175] Magnetic leakage signal characteristic determination module: Based on the initial magnetic leakage signal set and the calibrated pipe material parameters, calculate the attenuation coefficient of the magnetic leakage signal and determine the scattering characteristics of the signal;

[0176] Leakage magnetic signal attenuation optimization module: If the attenuation coefficient of the signal exceeds the preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated;

[0177] Weld defect feature extraction module: Extracts high-frequency transient features and low-frequency stable features from the signal time series and spatial distribution characteristics of the optimized leakage magnetic signal set to generate a preliminary feature vector of weld defects;

[0178] Defect location spatial distribution module: By using a pre-established signal propagation path model, the preliminary feature vector is mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location;

[0179] Defect detection shape determination module: If there are ambiguous areas in the spatial distribution map, the scattering features and the attenuation coefficient are used to optimize the preliminary feature vector to determine the depth and direction of the defect;

[0180] The second composite excitation signal module: Based on the depth and direction of the defect, analyzes the spatial distribution characteristics of the preliminary feature vector, determines the defect type, generates a detection report containing the defect type and spatial location, and dynamically adjusts the excitation signal parameters to generate the second composite excitation signal;

[0181] Weld defect report generation module: Based on the second composite excitation signal, recalculate the signal attenuation coefficient, extract the optimized scattering features, and obtain high-precision weld defect location results.

[0182] It should be noted that the deep learning-based multi-touch signal processing and magnetic flux leakage detection-based pipeline weld defect analysis device provided in this embodiment of the invention are used to execute all the process steps of the magnetic flux leakage detection-based pipeline weld defect analysis method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0183] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an initial leakage magnetic field signal generation program. When the processor executes the computer program, it implements the steps described in the various embodiments of the pipe weld defect analysis methods above, for example... Figure 1 The step S101 is shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 2 The initial leakage magnetic field signal generation module is shown.

[0184] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0185] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0186] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0187] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0188] The above specific embodiments have further described in detail the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and do not limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for analyzing defects in pipe welds based on magnetic flux leakage detection, characterized in that, include: The first composite excitation signal is generated by a multi-frequency excitation source and adjusted and optimized to obtain the initial leakage magnetic signal set; Based on the initial magnetic flux leakage signal set and the calibrated pipe material parameters, the attenuation coefficient of the magnetic flux leakage signal is calculated, and the scattering characteristics of the signal are determined. If the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated. High-frequency transient features and low-frequency stable features are extracted from the signal time series and spatial distribution characteristics of the optimized leakage magnetic field signal set to generate a preliminary feature vector of weld defects. By using a pre-established signal propagation path model, the preliminary feature vectors are mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location. If there are ambiguous areas in the spatial distribution map, the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect; Based on the depth and direction of the defect, the spatial distribution characteristics of the preliminary feature vector are analyzed to determine the defect type, generate a detection report containing the defect type and spatial location, and dynamically adjust the excitation signal parameters to generate a second composite excitation signal. Based on the second composite excitation signal, the signal attenuation coefficient is recalculated, the optimized scattering features are extracted, and high-precision weld defect location results are obtained. Wherein, if the attenuation coefficient of the leakage magnetic signal exceeds a preset attenuation threshold, adjusting the excitation signal and generating an optimized leakage magnetic signal set includes: If the attenuation coefficient of the leakage magnetic signal exceeds the preset attenuation threshold, the amplitude and transmission angle of the signal are adjusted to generate an adjusted parameter set. The amplitude change and angle adjustment are recorded to obtain the first optimized parameter set. Based on the first optimized parameter set, a leakage magnetic signal set is generated, and transient features are captured by high-frequency sampling to obtain the first leakage magnetic signal set. Time series data are extracted from the first leakage magnetic field signal set, and transient features and spatial distribution characteristics are separated to obtain the first feature dataset; The first feature dataset is optimized to generate optimized time series and spatial distribution characteristics, thus obtaining the optimized leakage magnetic signal set. The step of extracting high-frequency transient features and low-frequency stable features from the signal time series and spatial distribution characteristics of the optimized leakage magnetic field signal set to generate a preliminary feature vector of weld defects includes: From the optimized leakage magnetic signal set, time series data and spatial distribution data are obtained, and convolution operation is performed on the time series data to extract high-frequency transient features and obtain a first high-frequency feature set. Based on the first high-frequency feature set, the low-frequency stable features of the spatially distributed data are separated to obtain the first low-frequency feature set; If the feature dimensions of the first high-frequency feature set and the first low-frequency feature set meet the preset first dimension threshold, then the first high-frequency feature set and the first low-frequency feature set are merged to generate a first feature vector. The first feature vector is subjected to dimensionality reduction processing to generate a preliminary feature vector for weld defects.

2. The method according to claim 1, characterized in that, The process of generating a first composite excitation signal through a multi-frequency excitation source, adjusting and optimizing it, and obtaining an initial leakage magnetic signal set includes: High-frequency and low-frequency signals are generated within a preset frequency range to synthesize a composite excitation signal containing both high and low frequencies. Based on the composite excitation signal, combined with the pipe wall diameter and wall thickness parameters, the amplitude adjustment coefficient is calculated and adjusted to obtain the amplitude-adjusted excitation signal data; The signal propagation direction is calculated based on the amplitude-adjusted excitation signal data, and the emission angle is optimized to obtain the optimized excitation signal data. The magnetic flux leakage signals of different depth regions of the pipeline weld are obtained from the optimized excitation signal data. The collected analog signals are converted into digital signals to obtain the initial magnetic flux leakage signal set data.

3. The method according to claim 1, characterized in that, The step of calculating the attenuation coefficient of the magnetic flux leakage signal and determining the scattering characteristics of the signal based on the initial magnetic flux leakage signal set and the calibrated pipe material parameters includes: Based on the initial leakage magnetic field signal set, combined with the pre-input pipe diameter, wall thickness and weld width, a high-precision three-dimensional finite element mesh is generated to obtain three-dimensional mesh data; Based on the three-dimensional mesh data and the preset magnetic permeability and electrical conductivity parameters of the pipe material, the electromagnetic interaction boundary between the excitation source and the pipe surface is simulated, the electromagnetic field distribution is calculated, and the initial data of the leakage magnetic field signal is obtained. Based on the initial data of the leakage magnetic field signal, the propagation path of the signal in the complex pipeline structure is calculated, and the attenuation coefficient and scattering characteristics are determined by combining the geometric characteristics of the path.

4. The method according to claim 1, characterized in that, The training process of the pre-established signal propagation path model includes: Initialize the parameters of the signal propagation path model and the initial propagation path data; Based on the initial propagation path data and the pre-stored pipe geometry and material properties, finite element simulation and signal propagation simulation are performed to obtain the preliminary propagation trajectory and preliminary scattering characteristics. Based on the preliminary propagation trajectory and the preliminary scattering characteristics, a first scattering feature set is constructed; The first scattering feature set is subjected to dimensionality reduction processing to generate a second feature vector; Based on the second feature vector, the spatial location of the defect is updated and the signal propagation path is adjusted; When the number of training iterations is greater than or equal to the preset maximum number of training iterations, the training is considered complete, and the optimized signal propagation path model is obtained.

5. The method according to claim 1, characterized in that, If there are ambiguous regions in the spatial distribution map, the preliminary feature vector is optimized by combining the scattering characteristics and the attenuation coefficient to determine the depth and direction of the defect, including: Generate preliminary feature vectors for the fuzzy regions in the spatial distribution map; Based on the preliminary feature vector, combined with the scattering characteristics and the attenuation coefficient, the probability distribution is updated to obtain the optimized feature vector; Based on the optimized feature vectors, three-dimensional reconstruction is performed, and combined with spatial positioning technology, the defect depth and defect direction are determined.

6. The method according to claim 1, characterized in that, The dynamic adjustment of excitation signal parameters to generate a second composite excitation signal includes: Analyze the inspection report to obtain defect location distribution data, and process the defect coordinates and depth information in the defect location distribution data to obtain first defect distribution data; Based on the first defect distribution data, the frequency range of the excitation signal is calculated, and the first frequency range data is determined by analyzing the material properties and signal attenuation characteristics in the first defect distribution data. The signal propagation efficiency of the first frequency range data is simulated, and the amplitude of the excitation signal is adjusted to obtain the first amplitude adjustment parameter; Based on the first amplitude adjustment parameter and the first frequency range data, the signal propagation path in the first defect distribution data is calculated, and angle calibration is performed to generate the second composite excitation signal.

7. The method according to claim 1, characterized in that, The step of recalculating the signal attenuation coefficient based on the second composite excitation signal, extracting the optimized scattering features, and obtaining high-precision weld defect location results includes: Based on the second composite excitation signal set, combined with the preset mesh generation accuracy and boundary condition settings, finite element simulation is performed to calculate the propagation path of the signal in the weld material and obtain the first propagation path data; Based on the first propagation path data and combined with material properties, the energy loss during signal propagation is calculated to obtain the first attenuation coefficient data; If the first attenuation coefficient data is lower than the preset attenuation threshold, the scattering features of the first propagation path data are extracted to obtain the first scattering feature data. Based on the first scattering feature data and the distribution information of the defect depth region, the precise coordinates of the defect in the weld are calculated to obtain the high-precision weld defect location result.

8. A pipe weld defect analysis device, characterized in that, For implementing the method as described in any one of claims 1-7, comprising: Initial leakage flux signal generation module: Generates the first composite excitation signal through a multi-frequency excitation source, and adjusts and optimizes it to obtain the initial leakage flux signal set; Magnetic leakage signal characteristic determination module: Based on the initial magnetic leakage signal set and the calibrated pipe material parameters, calculate the attenuation coefficient of the magnetic leakage signal and determine the scattering characteristics of the signal; Leakage magnetic signal attenuation optimization module: If the attenuation coefficient of the signal exceeds the preset attenuation threshold, the excitation signal is adjusted and an optimized leakage magnetic signal set is generated; Weld defect feature extraction module: Extracts high-frequency transient features and low-frequency stable features from the signal time series and spatial distribution characteristics of the optimized leakage magnetic signal set to generate a preliminary feature vector of weld defects; Defect location spatial distribution module: By using a pre-established signal propagation path model, the preliminary feature vector is mapped to the three-dimensional defect spatial location to generate a spatial distribution map containing the defect location; Defect detection shape determination module: If there are ambiguous areas in the spatial distribution map, the scattering features and the attenuation coefficient are used to optimize the preliminary feature vector to determine the depth and direction of the defect; The second composite excitation signal module: Based on the depth and direction of the defect, analyzes the spatial distribution characteristics of the preliminary feature vector, determines the defect type, generates a detection report containing the defect type and spatial location, and dynamically adjusts the excitation signal parameters to generate the second composite excitation signal; Weld defect report generation module: Based on the second composite excitation signal, recalculate the signal attenuation coefficient, extract the optimized scattering features, and obtain high-precision weld defect location results.

Images