Metal pipeline defect identification and classification method

Abstract

The application discloses a metal pipeline defect identification and classification method, relates to the technical field of metal pipeline defects, and comprises the following steps: collecting a magnetic flux leakage detection signal of a target metal pipeline to obtain an original magnetic flux leakage signal sequence; performing denoising and normalization processing on the original magnetic flux leakage signal sequence to obtain a standardized magnetic flux leakage signal sequence; extracting time domain features, frequency domain features and statistical features from the standardized magnetic flux leakage signal sequence to obtain a defect feature vector; constructing a training sample set according to the defect feature vector and a corresponding preset defect category label, and dividing the training sample set into a training subset and a verification subset; training a support vector machine classifier by using the training subset, optimizing parameters of the support vector machine classifier through the verification subset, and obtaining a defect identification and classification model. The application can be widely applied to pipeline integrity detection scenes in the fields of oil and gas transportation, chemical production and the like, and meets the application requirements of industrial-grade defect classification.

Description

Technical Field

[0001] This invention relates to the field of metal pipe defect technology, and in particular to a method for identifying and classifying metal pipe defects. Background Technology

[0002] Magnetic flux leakage (MFL) testing technology, with its advantages of being non-contact and highly sensitive, has become the mainstream method for non-destructive testing of defects in metal pipelines in the oil and gas, chemical, and other fields. Traditional methods often employ a "manual interpretation + shallow classifier" approach. Among these, support vector machines (SVMs) are widely used for automatic classification of MFL defects due to their strong generalization ability with small samples. Related research often focuses on single-dimensional features, such as extracting only time-domain peak values ​​and pulse widths, or superimposing a few frequency-domain parameters such as the dominant frequency. This has improved detection efficiency to some extent and is suitable for the early detection needs of small-scale pipelines.

[0003] However, existing classification schemes based on support vector machines have significant limitations in multi-dimensional feature fusion: most studies have not fully integrated multi-dimensional information such as time-domain waveform details, frequency-domain energy distribution, and signal statistical characteristics, resulting in feature vectors that do not comprehensively characterize the essential attributes of defects. For example, the time-domain amplitudes of magnetic flux leakage signals from small corrosion pits and microcracks are similar, but they differ significantly in indicators such as the proportion of high-frequency components and statistical kurtosis in the frequency domain. The lack of single-dimensional features can easily lead to confusion between the two types of defects. Existing magnetic flux leakage defect identification technologies generally suffer from insufficient classification accuracy due to inadequate multi-dimensional feature fusion. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for identifying and classifying defects in metal pipes to solve the core problem of low defect identification accuracy caused by the single dimension of feature representation in existing solutions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for identifying and classifying defects in metal pipes, which includes acquiring the magnetic flux leakage detection signal of the target metal pipe to obtain the original magnetic flux leakage signal sequence; The original magnetic flux leakage signal sequence is denoised and normalized to obtain a standardized magnetic flux leakage signal sequence. Time-domain features, frequency-domain features, and statistical features are extracted from the standardized magnetic flux leakage signal sequence and combined to obtain a defect feature vector. A training sample set is constructed based on the defect feature vector and the corresponding preset defect category label, and the training sample set is divided into a training subset and a validation subset. The support vector machine classifier is trained using the training subset, and the parameters of the support vector machine classifier are optimized using the validation subset to obtain a defect identification classification model. The defect feature vector corresponding to the metal pipe under test is input into the defect identification and classification model, and the corresponding defect category determination result is output.

[0007] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the original magnetic flux leakage signal sequence is subjected to denoising and normalization processing to obtain a standardized magnetic flux leakage signal sequence, specifically including: The original magnetic flux leakage signal sequence was denoised using a wavelet threshold denoising method to remove environmental noise and random interference, resulting in a denoised magnetic flux leakage signal sequence. The amplitude of the denoised magnetic flux leakage signal sequence is mapped to the [0,1] interval to eliminate the amplitude deviation caused by detector lift-off value fluctuations and differences in pipe material, thus obtaining a standardized magnetic flux leakage signal sequence.

[0008] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the method involves extracting time-domain features, frequency-domain features, and statistical features from the standardized magnetic flux leakage signal sequence and combining them to obtain a defect feature vector, specifically including: The peak value, peak-to-peak value, rising edge slope, falling edge slope, and pulse width are extracted from the standardized magnetic flux leakage signal sequence as time-domain features; The standardized magnetic flux leakage signal sequence is subjected to a fast Fourier transform to extract the main frequency, spectral energy, and proportion of high-frequency components as frequency domain features. The mean, variance, skewness, and kurtosis are extracted from the standardized magnetic flux leakage signal sequence as statistical features. The time-domain features, frequency-domain features, and statistical features are combined in sequence to obtain the defect feature vector.

[0009] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the method includes: constructing a training sample set based on the defect feature vector and the corresponding preset defect category label, and dividing the training sample set into a training subset and a validation subset, specifically including: Collect magnetic flux leakage detection data of metal pipes containing known defect types, and extract defect feature vectors corresponding to various defects according to the method described in claims 2 to 3; Each set of defect feature vectors is labeled with a corresponding preset defect category label, which includes four categories: no defects, corrosion defects, crack defects, and pore defects. All labeled data are integrated into a training sample set, and the training sample set is divided into a training subset and a validation subset in a 7:3 ratio.

[0010] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the method includes: training a support vector machine classifier using the training subset and optimizing the parameters of the support vector machine classifier using the validation subset to obtain a defect identification and classification model, specifically including: The radial basis function kernel function is selected as the kernel function of the support vector machine classifier. The optimal parameter combination is determined by traversing different penalty factors C and kernel parameters γ using a grid search method, with the classification accuracy of the validation subset as the evaluation index. The support vector machine classifier is trained using the training subset and the optimal parameter combination to obtain a defect identification classification model.

[0011] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, in the grid search method, the traversal range of the penalty factor C is [0.1, 100], and the step size is 0.1; the traversal range of the kernel parameter γ is [0.01, 10], and the step size is 0.01.

[0012] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, in the process of training the support vector machine classifier using the training subset, the performance of the support vector machine classifier is evaluated by using 5-fold cross-validation to select the parameter combination with the highest verification accuracy.

[0013] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the step of acquiring the magnetic flux leakage detection signal of the target metal pipe to obtain the original magnetic flux leakage signal sequence specifically includes: The leakage magnetic field detector equipped with a Hall sensor is controlled to move at a constant speed along the axial direction of the target metal pipe; A fixed sampling frequency is set to perform circumferential scanning detection on the target metal pipe, and leakage magnetic signals at various positions around the pipe are collected to obtain the original leakage magnetic signal sequence, ensuring that the spatial interval between adjacent sampling points is consistent.

[0014] In a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the traveling speed of the magnetic flux leakage detector is 0.5 m / s, and the sampling frequency is 10 kHz.

[0015] As a preferred embodiment of the metal pipe defect identification and classification method of the present invention, the defect identification and classification model, after classifying the defect feature vector corresponding to the metal pipe under test, outputs a defect category determination result with a classification accuracy of not less than 95%.

[0016] The beneficial effects of this invention are as follows: By employing wavelet threshold denoising combined with amplitude normalization in the signal preprocessing steps, environmental noise and random interference in the magnetic flux leakage signal are effectively filtered out. Simultaneously, the amplitude deviation caused by detector lift-off value fluctuations and differences in pipeline material is eliminated, providing a stable and reliable signal foundation for subsequent feature extraction. This avoids interference with classification accuracy caused by noise and amplitude inconsistencies from the source. By simultaneously extracting the time-domain waveform features, frequency-domain energy distribution features, and signal statistical characteristics of the standardized magnetic flux leakage signal sequence and combining them into a full-dimensional defect feature vector, the essential attributes of different types of defects are comprehensively characterized. This overcomes the problem of insufficient single-dimensional feature representation capabilities in existing technologies, especially solving the confusion problem caused by overlapping features between similar defects such as small corrosion pits and microcracks. Through a parameter optimization step combining grid search with validation subset accuracy feedback, the kernel function parameters and penalty factor of the support vector machine classifier are accurately matched, fully leveraging the generalization advantage of support vector machines in small sample scenarios. Combined with a standardized training sample set construction process, high-precision and high-stability identification and classification of metal pipeline defects are finally achieved. This can be widely applied to pipeline integrity detection scenarios in fields such as oil and gas transportation and chemical production, meeting the application requirements of industrial-grade defect classification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for a method for identifying and classifying defects in metal pipes.

[0019] Figure 2 This is a flowchart of the signal preprocessing steps.

[0020] Figure 3 Flowchart of the training steps for the defect identification and classification model. Detailed Implementation

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0024] Reference Figures 1-3 This is one embodiment of the present invention, which provides a method for identifying and classifying defects in metal pipes, including the following steps: The original magnetic flux leakage signal sequence is denoised using a wavelet threshold denoising method to remove environmental noise and random interference, resulting in a denoised magnetic flux leakage signal sequence. The amplitude of the denoised magnetic flux leakage signal sequence is mapped to the [0,1] interval to eliminate amplitude deviations caused by detector lift-off value fluctuations and differences in pipe material, thus obtaining a standardized magnetic flux leakage signal sequence.

[0025] Furthermore, by employing wavelet threshold denoising to denoise the original magnetic flux leakage signal sequence, random noise introduced by electromagnetic interference and mechanical vibration in the field detection environment can be effectively filtered out, while retaining the effective magnetic flux leakage component caused by the defect and avoiding noise signals from obscuring the details of the defect features. Subsequently, the amplitude of the denoised magnetic flux leakage signal sequence is mapped to the [0,1] interval, which can eliminate amplitude deviations caused by non-defect factors such as fluctuations in the lift-off value of the magnetic flux leakage detector and inhomogeneity of the pipeline base material. This ensures that the features extracted subsequently are only related to the attributes of the defect itself, providing a stable and consistent input benchmark for the classification model and significantly improving the model's anti-interference ability and generalization.

[0026] Peak value, peak-to-peak value, rising edge slope, falling edge slope, and pulse width are extracted from the standardized magnetic flux leakage signal sequence as time-domain features; fast Fourier transform is performed on the standardized magnetic flux leakage signal sequence to extract the main frequency, spectral energy, and high-frequency component ratio as frequency-domain features; mean, variance, skewness, and kurtosis are extracted from the standardized magnetic flux leakage signal sequence as statistical features.

[0027] Furthermore, by simultaneously extracting time-domain features, frequency-domain features, and statistical features from the standardized magnetic flux leakage signal sequence and combining them into a defect feature vector, a multi-dimensional characterization of defect information is achieved. Among them, the time-domain features accurately reflect the geometric contour and signal waveform of the defect, the frequency-domain features reveal the distribution law of magnetic field energy caused by the defect, and the statistical features quantify the distribution characteristics and anomaly degree of the signal. After the three complement each other and are integrated, the differences in magnetic flux leakage response of different types of defects such as corrosion defects, crack defects, and pore defects can be fully distinguished. In particular, it solves the confusion problem caused by the overlap of single-dimensional features of similar defects such as small corrosion pits and microcracks, and greatly improves the characterization ability of the feature vector.

[0028] Collect magnetic flux leakage detection data of metal pipes containing known defect types, and extract defect feature vectors corresponding to various defects according to the method described in claims 2 to 3; label each set of defect feature vectors with a corresponding preset defect category label, the preset defect category label including four categories: no defects, corrosion defects, crack defects and hole defects; integrate all labeled data into a training sample set, and divide the training sample set into a training subset and a validation subset in a 7:3 ratio.

[0029] Furthermore, by collecting magnetic flux leakage detection data of metal pipes containing known defect types, a fully labeled training sample set was constructed in strict accordance with the aforementioned preprocessing and feature extraction process. This sample set was then scientifically divided into a training subset and a validation subset in a 7:3 ratio. This ensured both the diversity and representativeness of the training data and enabled unbiased evaluation of the model's performance through an independent validation set. This avoided the risk of overfitting during training, laying a data foundation for the reliable training of the classification model and ensuring that the model could accurately learn the feature mapping patterns of different types of defects.

[0030] A radial basis function kernel is selected as the kernel function of the support vector machine classifier; a grid search method is used to traverse different penalty factors C and kernel parameters γ, and the classification accuracy of the validation subset is used as the evaluation index to determine the optimal parameter combination; the support vector machine classifier is trained using the training subset and the optimal parameter combination to obtain a defect identification classification model.

[0031] Furthermore, by selecting the radial basis function kernel function as the kernel function of the support vector machine classifier, and combining it with the grid search method to optimize the penalty factor C and kernel parameter γ using the accuracy of the validation subset classification as the evaluation index, the parameter combination of the classifier can be accurately matched. This fully leverages the nonlinear classification advantage of support vector machines in small sample scenarios, making the classification decision boundary more closely match the feature distribution of different types of defects. While improving classification accuracy, it also takes into account the generalization performance of the model, solving the problems of traditional parameter selection relying on experience and poor adaptability.

[0032] In the grid search method, the traversal range of the penalty factor C is [0.1, 100], and the step size is 0.1; the traversal range of the kernel parameter γ is [0.01, 10], and the step size is 0.01.

[0033] Furthermore, by limiting the traversal range of the penalty factor C to [0.1, 100] with a step size of 0.1, and the traversal range of the kernel parameter γ to [0.01, 10] with a step size of 0.01, the search space is significantly reduced while ensuring the accuracy of parameter optimization. This avoids computational redundancy caused by an excessively large parameter range and prevents the risk of missing the optimal solution due to an excessively small range. It significantly improves the efficiency and reliability of the parameter optimization process, enabling the model to converge to the optimal performance state within a reasonable time.

[0034] During the training of the support vector machine classifier using the aforementioned training subset, the performance of the support vector machine classifier is evaluated using a 5-fold cross-validation method to select the parameter combination with the highest validation accuracy.

[0035] Furthermore, by introducing 5-fold cross-validation during training to evaluate the performance of the support vector machine classifier, the training subset can be further divided into multiple mutually exclusive sub-training and validation sets. By taking the average accuracy through multiple rounds of validation, the random error caused by a single partition is effectively reduced, ensuring that the selected parameter combination performs stably on different data subsets, further improving the robustness and generalization ability of the model, and avoiding overfitting of the model to a specific data distribution.

[0036] The magnetic flux leakage detector equipped with a Hall sensor is controlled to move at a constant speed along the axial direction of the target metal pipe; a fixed sampling frequency is set to perform circumferential scanning detection on the target metal pipe, and magnetic flux leakage signals at various positions along the circumference of the pipe are collected to obtain the original magnetic flux leakage signal sequence, ensuring that the spatial interval between adjacent sampling points is consistent.

[0037] Furthermore, by controlling the magnetic flux leakage detector equipped with a Hall sensor to move at a constant speed along the axial direction of the target metal pipe and setting a fixed sampling frequency to perform circumferential scanning detection of the pipe, the spatial interval between adjacent sampling points is completely consistent, avoiding signal distortion or uneven spatial resolution caused by detector speed fluctuations and unstable sampling frequencies. This ensures that the acquired original magnetic flux leakage signal sequence can truly reflect the defect information at each position in the circumference of the pipe, providing a high-quality data source for subsequent signal processing and feature extraction.

[0038] The magnetic flux leakage detector travels at a speed of 0.5 m / s, and the sampling frequency is 10 kHz.

[0039] Furthermore, by setting the travel speed of the magnetic flux leakage detector to 0.5 m / s and the sampling frequency to 10 kHz, a spatial sampling interval of approximately 0.05 mm along the pipeline axis was achieved after matching the two. This not only meets the detection resolution requirements for minute defects but also avoids data redundancy and processing pressure caused by excessively high sampling frequencies. It ensures both detection accuracy and detection efficiency, making it suitable for the online detection needs of long-distance pipelines in industrial sites.

[0040] After the defect identification and classification model classifies the defect feature vector corresponding to the metal pipe under test, the classification accuracy of the output defect category determination result is not less than 95%.

[0041] Furthermore, by setting the classification accuracy of the defect identification and classification model to no less than 95%, the lower limit of the model's performance indicators is clarified, ensuring that it has practical application value in scenarios with extremely high safety requirements such as oil and gas transportation and chemical production. It can reliably distinguish between types such as no defects, corrosion defects, crack defects, and pore defects, providing accurate data support for pipeline integrity assessment and maintenance decisions, and effectively reducing the risk of major safety accidents such as pipeline leaks and bursts.

[0042] In summary, this invention effectively filters out environmental noise and random interference in magnetic flux leakage signals by employing wavelet threshold denoising combined with amplitude normalization as a signal preprocessing step. It also eliminates amplitude deviations caused by detector lift-off fluctuations and differences in pipe material, providing a stable and reliable signal foundation for subsequent feature extraction. This avoids interference with classification accuracy caused by noise and amplitude inconsistencies from the source. By simultaneously extracting the time-domain waveform features, frequency-domain energy distribution features, and signal statistical characteristics of the standardized magnetic flux leakage signal sequence and combining them into a full-dimensional defect feature vector, the invention comprehensively characterizes the essential attributes of different types of defects, overcoming the insufficient single-dimensional feature representation capability in existing technologies. In particular, it solves the problem of confusion caused by overlapping features between similar defects such as small corrosion pits and microcracks. Through a parameter optimization step combining grid search with validation subset accuracy feedback, the kernel function parameters and penalty factor of the support vector machine classifier are accurately matched, fully leveraging the generalization advantage of support vector machines in small sample scenarios. Combined with a standardized training sample set construction process, this invention ultimately achieves high-precision and high-stability identification and classification of metal pipe defects. It can be widely applied to pipeline integrity detection scenarios in fields such as oil and gas transportation and chemical production, meeting the application requirements of industrial-grade defect classification.

[0043] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying and classifying defects in metal pipes, characterized in that: include, The magnetic flux leakage signal of the target metal pipe is collected to obtain the original magnetic flux leakage signal sequence; The original magnetic flux leakage signal sequence is denoised and normalized to obtain a standardized magnetic flux leakage signal sequence. Time-domain features, frequency-domain features, and statistical features are extracted from the standardized magnetic flux leakage signal sequence and combined to obtain a defect feature vector. A training sample set is constructed based on the defect feature vector and the corresponding preset defect category label, and the training sample set is divided into a training subset and a validation subset. The support vector machine classifier is trained using the training subset, and the parameters of the support vector machine classifier are optimized using the validation subset to obtain a defect identification classification model. The defect feature vector corresponding to the metal pipe under test is input into the defect identification and classification model, and the corresponding defect category determination result is output.

2. The method for identifying and classifying defects in metal pipes as described in claim 1, characterized in that: The original magnetic flux leakage signal sequence is denoised and normalized to obtain a standardized magnetic flux leakage signal sequence, specifically including: The original magnetic flux leakage signal sequence was denoised using a wavelet threshold denoising method to remove environmental noise and random interference, resulting in a denoised magnetic flux leakage signal sequence. The amplitude of the denoised magnetic flux leakage signal sequence is mapped to the [0,1] interval to eliminate the amplitude deviation caused by detector lift-off value fluctuations and differences in pipe material, thus obtaining a standardized magnetic flux leakage signal sequence.

3. The method for identifying and classifying defects in metal pipes as described in claim 2, characterized in that: Time-domain features, frequency-domain features, and statistical features are extracted from the standardized magnetic flux leakage signal sequence and combined to obtain a defect feature vector, specifically including: The peak value, peak-to-peak value, rising edge slope, falling edge slope, and pulse width are extracted from the standardized magnetic flux leakage signal sequence as time-domain features; The standardized magnetic flux leakage signal sequence is subjected to a fast Fourier transform to extract the main frequency, spectral energy, and proportion of high-frequency components as frequency domain features. The mean, variance, skewness, and kurtosis are extracted from the standardized magnetic flux leakage signal sequence as statistical features. The time-domain features, frequency-domain features, and statistical features are combined in sequence to obtain the defect feature vector.

4. The method for identifying and classifying defects in metal pipes as described in claim 3, characterized in that: A training sample set is constructed based on the defect feature vector and the corresponding preset defect category label, and the training sample set is divided into a training subset and a validation subset, specifically including: Collect magnetic flux leakage detection data of metal pipes containing known defect types, and extract defect feature vectors corresponding to various defects according to the method described in claims 2 to 3; Each set of defect feature vectors is labeled with a corresponding preset defect category label, which includes four categories: no defects, corrosion defects, crack defects, and pore defects. All labeled data are integrated into a training sample set, and the training sample set is divided into a training subset and a validation subset in a 7:3 ratio.

5. The method for identifying and classifying defects in metal pipes as described in claim 4, characterized in that: The support vector machine classifier is trained using the training subset, and its parameters are optimized using the validation subset to obtain a defect identification classification model, specifically including: The radial basis function kernel function is selected as the kernel function of the support vector machine classifier. The optimal parameter combination is determined by traversing different penalty factors C and kernel parameters γ using a grid search method, with the classification accuracy of the validation subset as the evaluation index. The support vector machine classifier is trained using the training subset and the optimal parameter combination to obtain a defect identification classification model.

6. The method for identifying and classifying defects in metal pipes as described in claim 5, characterized in that: In the grid search method, the traversal range of the penalty factor C is [0.1, 100], and the step size is 0.1; the traversal range of the kernel parameter γ is [0.01, 10], and the step size is 0.

01.

7. The method for identifying and classifying defects in metal pipes as described in claim 6, characterized in that: During the training of the support vector machine classifier using the aforementioned training subset, the performance of the support vector machine classifier is evaluated using a 5-fold cross-validation method to select the parameter combination with the highest validation accuracy.

8. The method for identifying and classifying defects in metal pipes as described in claim 7, characterized in that: The process of acquiring the magnetic flux leakage detection signal of the target metal pipe to obtain the original magnetic flux leakage signal sequence specifically includes: The leakage magnetic field detector equipped with a Hall sensor is controlled to move at a constant speed along the axial direction of the target metal pipe; A fixed sampling frequency is set to perform circumferential scanning detection on the target metal pipe, and leakage magnetic signals at various positions around the pipe are collected to obtain the original leakage magnetic signal sequence, ensuring that the spatial interval between adjacent sampling points is consistent.

9. The method for identifying and classifying defects in metal pipes as described in claim 8, characterized in that: The magnetic flux leakage detector travels at a speed of 0.5 m / s, and the sampling frequency is 10 kHz.

10. The method for identifying and classifying defects in metal pipes as described in claim 9, characterized in that: After classifying the defect feature vector corresponding to the metal pipe under test, the defect identification and classification model outputs a defect category determination result with a classification accuracy of no less than 95%.

Images