A method for improving internal detection performance based on high-sensitivity fluxgate probe

Abstract

The present invention relates to the field of deformation monitoring technology, and discloses a method for improving the internal detection performance based on a high-sensitivity fluxgate probe, wherein a data acquisition device is arranged on a pipeline detection robot; the first three-axis magnetic field intensity data of the defect-free area is collected, and the baseline magnetic field data is generated according to the first three-axis magnetic field intensity data; the second three-axis magnetic field intensity data, phase difference and position encoding in the defective area are collected; the second three-axis magnetic field intensity data is denoised and baseline corrected to obtain abnormal three-axis magnetic field intensity data; defect feature extraction is performed based on the abnormal magnetic field signal data set to obtain feature parameters, and the feature parameters are normalized and input into a pre-trained neural network model to obtain defect depth, defect width and defect length. The present invention can realize accurate measurement of cracks and quantitative detection of crack depth, improve the performance of the current internal detector, and upgrade the internal detector from qualitative detection to quantitative detection.

Description

Technical Field

[0001] The present invention relates to the field of deformation monitoring technology, and in particular to a method for improving detection performance based on a high-sensitivity fluxgate probe. Background Art

[0002] As the core facility for energy transportation, the structural integrity of oil and gas pipelines is directly related to safe operation and environmental protection. However, due to long-term influence of factors such as corrosion, mechanical stress and environmental erosion, pipelines are prone to cracks, corrosion and other defects, which may cause leakage or even explosion accidents. Therefore, pipeline in-pipeline detection technology has become a key means to ensure pipeline safety.

[0003] Traditional pipeline in-pipeline inspection technologies mainly rely on methods such as magnetic particle inspection, ultrasonic inspection and eddy current inspection, but these technologies have limitations: magnetic particle inspection requires magnetic powder to be coated on the pipeline surface, is only suitable for outer wall defect detection, and cannot quantitatively evaluate the defect size; ultrasonic inspection relies on coupling agents, and the signal is attenuated in complex pipeline environments (such as high viscosity media or inner wall scaling), and the detection efficiency is low; eddy current inspection is not sensitive enough to deep defects and is affected by the unevenness of the electromagnetic properties of the material.

[0004] Therefore, how to effectively denoise the magnetic field data and combine it with advanced algorithm models to improve the accuracy of defect detection has become an urgent problem to be solved. Summary of the invention

[0005] In view of this, the present invention proposes a method for improving the internal detection performance based on a high-sensitivity fluxgate probe, which can achieve accurate measurement of cracks and quantitative detection of crack depth, improve the performance of current internal detectors, and upgrade the internal detectors from qualitative detection to quantitative detection.

[0006] The present invention proposes a method for improving the detection performance based on a high-sensitivity fluxgate probe, comprising:

[0007] An annular fluxgate probe array is arranged on the pipeline detection robot, and an adjustable magnetic field excitation coil is arranged around the annular fluxgate probe array;

[0008] Controlling the pipeline inspection robot to move in a defect-free area in the pipeline, collecting first three-axis magnetic field intensity data of the defect-free area, and generating baseline magnetic field data according to the first three-axis magnetic field intensity data;

[0009] Control the pipeline inspection robot to move in the defective area in the pipeline, start the adjustable magnetic field excitation coil, and collect the second three-axis magnetic field intensity data, phase difference and position code in the defective area;

[0010] De-noising the second three-axis magnetic field intensity data to obtain de-noised three-axis magnetic field intensity data, and performing baseline correction to obtain abnormal three-axis magnetic field intensity data;

[0011] A second three-axis magnetic field gradient tensor is obtained based on the abnormal three-axis magnetic field intensity data, and the abnormal three-axis magnetic field intensity data, the second three-axis magnetic field gradient tensor, the phase difference and the position code are integrated into an abnormal magnetic field signal data set;

[0012] Defect features are extracted based on the abnormal magnetic field signal data set to obtain feature parameters, which are normalized and input into the pre-trained neural network model to obtain defect depth, defect width and defect length.

[0013] Further, the method includes: generating the baseline magnetic field data according to the first three-axis magnetic field data comprises: performing mean calculation according to multiple groups of first three-axis magnetic field intensity data, expressed as:

[0014] ;

[0015] in, For the coordinate axis The first three-axis magnetic field strength data on is the number of data sets, is the baseline magnetic field data.

[0016] Further, it includes: denoising the second three-axis magnetic field intensity data to obtain denoised three-axis magnetic field intensity data, and performing baseline correction to obtain abnormal three-axis magnetic field intensity data, specifically: firstly filtering out the power frequency noise of the second three-axis magnetic field intensity data, then suppressing the high-frequency noise, and finally compensating for the motion noise to obtain denoised three-axis magnetic field intensity data, and performing baseline correction on the denoised three-axis magnetic field intensity data to obtain abnormal three-axis magnetic field intensity data.

[0017] Further, it includes: filtering the second three-axis magnetic field intensity data for power frequency noise, specifically: using an infinite impulse response filter to perform zero-phase filtering;

[0018] The high-frequency noise is then suppressed, specifically: a wavelet threshold denoising method is used to decompose the second three-axis magnetic field intensity data after the power frequency noise is filtered out into five layers of wavelet coefficients, soft threshold processing is performed on each layer, and after processing, the signal is reconstructed to obtain the three-axis magnetic field intensity data after high-frequency denoising;

[0019] The compensation of motion noise is specifically performed as follows: collecting the three-axis acceleration of the pipeline inspection robot, obtaining motion noise data by the least square method, and finally subtracting the motion noise data from the three-axis magnetic field intensity data after high-frequency denoising to obtain the denoised three-axis magnetic field intensity data, which is expressed as:

[0020] + ;

[0021] in, is the motion noise data, is the three-axis acceleration of the pipeline inspection robot, is the first fitting coefficient, is the second fitting coefficient, is the third fitting coefficient;

[0022] ;

[0023] in, is the denoised three-axis magnetic field intensity data, It is the three-axis magnetic field intensity data after high-frequency denoising;

[0024] The de-noised three-axis magnetic field intensity data is baseline corrected to obtain abnormal three-axis magnetic field intensity data, which is expressed as:

[0025] ;

[0026] in, Abnormal three-axis magnetic field intensity data.

[0027] Further, it includes: obtaining a second three-axis magnetic field gradient tensor based on the abnormal three-axis magnetic field intensity data, expressed as:

[0028] ;

[0029] in, is the second triaxial magnetic field gradient tensor.

[0030] Furthermore, it includes: extracting defect features based on the abnormal magnetic field signal data set to obtain characteristic parameters, specifically: extracting defect features based on the abnormal magnetic field signal data set to obtain the peak amplitude of the abnormal magnetic field intensity data, the maximum gradient value of the second three-axis magnetic field gradient tensor, the abnormal area width, the phase lag and the multi-frequency response ratio.

[0031] Further, it includes: the specific content of obtaining the peak amplitude of the abnormal magnetic field strength data is: using a sliding window method to obtain candidate peaks, threshold screening the candidate peaks, and finally peak merging to obtain the peak amplitude of the abnormal magnetic field strength data.

[0032] Further, it includes: the specific content of obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor is: taking the modulus length of the second three-axis magnetic field gradient tensor, searching for the maximum gradient value, and obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor and the corresponding position.

[0033] Further, it includes: the specific content of obtaining the abnormal area width is: according to a preset threshold, marking all continuous areas where the abnormal magnetic field strength data is greater than the preset threshold, finding the left and right boundaries from the area, and finally obtaining the abnormal area width.

[0034] Further, the method comprises: obtaining the phase lag comprises: obtaining the phase lag by using the FFT method and the zero-crossing detection method on the excitation signal generated by the adjustable magnetic field excitation coil and the abnormal magnetic field strength data;

[0035] Further, it includes: the specific content of obtaining the multi-frequency response ratio is: band-pass filtering the abnormal magnetic field intensity data, and extracting the amplitudes at the highest Hz and the lowest Hz for comparison to obtain the multi-frequency response ratio.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention is based on the improvement of the detection performance of the high-sensitivity fluxgate probe. By arranging a circular fluxgate probe array on the pipeline detection robot and configuring an adjustable magnetic field excitation coil around it, the present invention can realize the fine capture of weak magnetic field abnormal signals, greatly improving the detection sensitivity and resolution of pipeline defects; the first three-axis magnetic field data of the defect-free area is first collected before the defect and the baseline magnetic field data is generated, which can provide an accurate reference for the subsequent baseline correction, thereby ensuring the accuracy of the abnormal magnetic field signal extraction; the present invention adopts a multi-level noise processing scheme: first, an infinite impulse response (IIR) filter is used for zero-phase filtering to effectively filter out the power frequency noise; then, the wavelet threshold denoising method is used to decompose the signal, and soft threshold processing is performed on the wavelet coefficients of each layer, and then the signal is reconstructed to suppress high-frequency noise; finally, by collecting the three-axis acceleration data of the pipeline detection robot, the motion noise is modeled and compensated in combination with the least squares method, thereby further improving The purity of the signal; the baseline correction process compares the noise-processed data with the previously collected baseline magnetic field data, so as to accurately separate the abnormal magnetic field signal and ensure that the detection result is not affected by environmental and motion interference; the present invention uses the abnormal three-axis magnetic field intensity data after baseline correction, and further calculates the second three-axis magnetic field gradient tensor to reflect the spatial distribution characteristics of the magnetic field change; through the steps of sliding window, threshold screening and peak merging, the peak amplitude of the abnormal magnetic field signal can be accurately extracted; at the same time, by taking the modulus length of the gradient tensor and searching for the maximum gradient value, the key parameters of the magnetic field gradient are obtained; in addition, it also combines multiple parameters such as abnormal area width, phase lag and multi-frequency response ratio to fully reflect the multi-dimensional characteristics of pipeline defects and provide sufficient information for subsequent intelligent judgment; multi-level noise filtering and motion noise compensation technology enable the present invention to effectively deal with the interference caused by complex environment in actual pipeline detection, such as power frequency electromagnetic noise, environmental high-frequency interference and dynamic noise caused by robot movement. Baseline correction ensures that the consistency and stability of the data can be maintained even when the detection conditions change, thereby greatly improving the overall reliability and stability of the detection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0038] Figure 1 The present invention is a flowchart of a method for improving detection performance based on a high-sensitivity fluxgate probe according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0040] See also Figure 1 As shown, an embodiment of the present invention provides a method for improving detection performance based on a high-sensitivity fluxgate probe, comprising:

[0041] S1: Arrange an annular fluxgate probe array on the pipeline inspection robot, and arrange an adjustable magnetic field excitation coil around the annular fluxgate probe array;

[0042] S2: Control the pipeline inspection robot to move in a defect-free area in the pipeline, collect first three-axis magnetic field intensity data of the defect-free area, and generate baseline magnetic field data according to the first three-axis magnetic field intensity data;

[0043] S3: Control the pipeline inspection robot to move in the defective area in the pipeline, start the adjustable magnetic field excitation coil, and collect the second three-axis magnetic field intensity data, phase difference and position code in the defective area;

[0044] S4: performing denoising processing on the second three-axis magnetic field intensity data to obtain denoised three-axis magnetic field intensity data, and performing baseline correction to obtain abnormal three-axis magnetic field intensity data;

[0045] S5: obtaining a second three-axis magnetic field gradient tensor based on the abnormal three-axis magnetic field intensity data, and integrating the abnormal three-axis magnetic field intensity data, the second three-axis magnetic field gradient tensor, the phase difference and the position code into an abnormal magnetic field signal data set;

[0046] S6: Defect features are extracted based on the abnormal magnetic field signal data set to obtain feature parameters, which are normalized and input into the pre-trained neural network model to obtain defect depth, defect width and defect length.

[0047] Specifically, the annular fluxgate probe array installed on the pipeline inspection robot adopts a circular layout, which can cover the inner wall of the pipeline 360° without blind spots, ensuring that the magnetic field change information in all directions of the pipeline is captured. At the same time, on the outside of this probe array, multiple adjustable magnetic field excitation coils are evenly arranged. These excitation coils can generate magnetic fields of different intensities and frequencies. By adjusting the working parameters of the excitation coil, the distribution and amplitude of the excitation magnetic field can be accurately controlled to meet the inspection needs of pipelines of different materials and structures.

[0048] It should be noted that the layout of the annular fluxgate probe array enables the detection equipment to realize all-round and multi-angle data collection in the pipeline, ensuring the coverage and data integrity during the detection process, thereby improving the accuracy of defect detection. The adjustable magnetic field excitation coil is used to generate a controllable excitation magnetic field, which can stimulate the magnetic response in the pipeline material, especially forming a more obvious magnetic field anomaly at the defect. By adjusting the parameters of the excitation magnetic field in a timely manner, the magnetic signal in the defect area can be strengthened, making the originally weak magnetic field changes easier to be captured and analyzed by the probe. This structural design not only improves the equipment's ability to capture subtle magnetic field changes, but also can adapt to the complex detection environment of different pipeline materials and structures through the adjustment of the excitation magnetic field, thereby improving the robustness and sensitivity of the overall detection system. Through the synergistic effect of the annular probe and the excitation coil, the detection robot can obtain higher quality magnetic field data and feedback defect information in a timely manner. This design effectively shortens the data acquisition and processing time, and at the same time provides an accurate basis for subsequent defect location and quantitative analysis, greatly improving the detection efficiency and safety level.

[0049] In some embodiments of the present application, the content of generating the baseline magnetic field data according to the first three-axis magnetic field data is: performing mean calculation according to multiple sets of first three-axis magnetic field intensity data, expressed as:

[0050] ;

[0051] in, For the coordinate axis The first three-axis magnetic field strength data on is the number of data sets, is the baseline magnetic field data.

[0052] Specifically, in the defect-free area of ​​the pipeline, multiple sets of first three-axis magnetic field intensity data are continuously collected, and then the mean value of each set of data is calculated on each coordinate axis, and finally a set of stable baseline magnetic field data is obtained. This mean value calculation method can effectively filter out accidental instantaneous interference and random noise, ensuring that the obtained baseline data truly reflects the magnetic field distribution of the pipeline in a defect-free state.

[0053] It should be noted that the generated baseline magnetic field data represents the magnetic field characteristics of the pipeline under normal conditions. It serves as a reference for subsequent defect detection and abnormal signal analysis, helping to distinguish normal signals from abnormal changes caused by defects. By averaging multiple sets of data, it can effectively eliminate instantaneous noise caused by equipment fluctuations, environmental interference, random errors and other factors, making the baseline data more stable and accurate, providing a reliable basis for subsequent signal comparison.

[0054] In some embodiments of the present application, the second three-axis magnetic field intensity data is denoised to obtain denoised three-axis magnetic field intensity data, and a baseline correction is performed to obtain abnormal three-axis magnetic field intensity data. The specific content is: first, the second three-axis magnetic field intensity data is subjected to power frequency noise filtering, and then the high-frequency noise is suppressed, and finally the motion noise is compensated to obtain denoised three-axis magnetic field intensity data, and the denoised three-axis magnetic field intensity data is subjected to baseline correction to obtain abnormal three-axis magnetic field intensity data.

[0055] In some embodiments of the present application, the second three-axis magnetic field intensity data is subjected to power frequency noise filtering, specifically: an infinite impulse response filter is used to perform zero-phase filtering;

[0056] In some embodiments of the present application, when filtering the power frequency noise of the second three-axis magnetic field intensity data, an infinite impulse response (IIR) filter is specifically used for zero phase filtering. In data processing, an IIR filter with power frequency noise (such as 50Hz or 60Hz) characteristics is used, and then the collected magnetic field data is first forward filtered, and then the same filtering process is performed in reverse order, thereby eliminating the phase delay introduced by the filter. This process ensures that the amplitude of the filtered data is effectively suppressed from the power frequency noise, while keeping the phase information of the original signal undistorted.

[0057] It should be noted that power frequency noise is common in industrial environments. By using a specially designed IIR filter, this periodic noise can be effectively filtered out, making the collected magnetic field data cleaner and facilitating the subsequent detection and processing of abnormal signals. Zero-phase filtering eliminates the phase offset problem that may be caused by traditional IIR filters through forward and reverse filtering. This is crucial for subsequent processing links that rely on phase information for defect analysis and location, ensuring that the time domain characteristics of the data truly reflect the original signal. After eliminating the influence of power frequency noise, subsequent data analysis can more accurately capture the tiny magnetic field changes caused by pipeline defects, improving the sensitivity and accuracy of defect detection.

[0058] Subsequently, high-frequency noise is suppressed, specifically: the wavelet threshold denoising method is used to decompose the second three-axis magnetic field intensity data after power frequency noise is filtered out into five layers of wavelet coefficients, and soft threshold processing is applied to each layer. After processing, the signal is reconstructed to obtain the three-axis magnetic field intensity data after high-frequency denoising;

[0059] In some embodiments of the present application, first, the filtered signal is decomposed into five layers of wavelet coefficients through wavelet transform, each layer corresponds to signal components in different frequency ranges, so as to distinguish high-frequency noise from signal features at multiple scales. Then, soft threshold processing is applied to the wavelet coefficients of each layer, that is, the coefficients below the set threshold are smoothly attenuated, while those larger coefficients reflecting the real signal information are retained or only slightly modified. Finally, the signal is reconstructed using the processed wavelet coefficients of each layer, so as to obtain the three-axis magnetic field intensity data after the high-frequency noise has been effectively suppressed.

[0060] It should be noted that after the initial filtering of the power frequency noise, other high-frequency noise components may still exist in the signal. The wavelet threshold denoising method can be used to process these high-frequency noises more finely, thereby significantly reducing the noise level. By decomposing the signal into five layers of wavelet coefficients, noise and useful signals can be distinguished at different scales, which helps to effectively suppress noise while retaining key feature information. Compared with hard threshold processing, soft threshold processing can more smoothly attenuate noise components and avoid the loss of important signal details due to overly strict thresholds, thereby ensuring the accuracy of subsequent defect feature extraction. After high-frequency noise suppression, the clarity and signal-to-noise ratio of the signal are greatly improved, making subtle abnormal magnetic field changes more obvious, providing a more reliable data basis for defect identification and quantification, and ultimately improving the accuracy and robustness of the entire detection system.

[0061] The motion noise is compensated by collecting the three-axis acceleration of the pipeline inspection robot, obtaining the motion noise data through the least square method, and finally subtracting the motion noise data from the three-axis magnetic field intensity data after high-frequency denoising to obtain the denoised three-axis magnetic field intensity data, which is expressed as:

[0062] + ;

[0063] in, is the motion noise data, is the three-axis acceleration of the pipeline inspection robot, is the first fitting coefficient, is the second fitting coefficient, is the third fitting coefficient;

[0064] ;

[0065] in, is the denoised three-axis magnetic field intensity data, It is the three-axis magnetic field intensity data after high-frequency denoising;

[0066] In some embodiments of the present application, first, the three-axis acceleration data generated by the pipeline inspection robot during operation is collected in real time, and these data can accurately reflect the dynamic changes in the robot's motion state. Then, these acceleration data are fitted and calculated by applying the least squares method to establish a mathematical model between the motion state and the magnetic field interference, thereby obtaining noise data related to the robot's motion. Finally, the calculated motion noise data is subtracted from the three-axis magnetic field intensity data obtained after high-frequency noise suppression to obtain the final three-axis magnetic field intensity data after motion noise compensation and removal of motion-induced interference.

[0067] It should be noted that during the operation of the pipeline inspection robot, additional noise will be introduced due to dynamic movements such as vibration, acceleration or deceleration. These motion noises may interfere with the true reflection of the magnetic field data. By compensating for motion noise, this interference can be effectively reduced, making the detection data closer to the actual state. After compensating for motion noise, the denoised magnetic field intensity data can more truly reflect the changes in the magnetic properties of the material in the pipeline, helping the detection system to more accurately identify magnetic field anomalies caused by defects, and improving the sensitivity and accuracy of detection. By comprehensively considering the interference caused by the robot movement, and using mathematical modeling and data compensation technology, the impact of external movement on the detection data is minimized, further enhancing the adaptability and stability of the detection system under complex working conditions. The motion noise model is established using the least squares method, and the acceleration data is accurately fitted, which can accurately quantify and deduct motion noise, so that the final magnetic field data is purer. This is crucial for subsequent defect feature extraction and intelligent judgment to ensure the authenticity and consistency of the signal.

[0068] The denoised three-axis magnetic field intensity data is baseline corrected to obtain abnormal three-axis magnetic field intensity data, which is expressed as:

[0069] ;

[0070] in, Abnormal three-axis magnetic field intensity data.

[0071] In some embodiments of the present application, it is first necessary to compare these data with the baseline magnetic field data collected and generated in advance in the defect-free area. Specifically, the baseline correction removes the stable components of the background magnetic field by calculating the difference between the detection data and the baseline data, and only retains the abnormal magnetic field signal generated by the pipeline defect. After this processing, the abnormal three-axis magnetic field intensity data finally obtained can more accurately reflect the magnetic field changes in the defective area.

[0072] It should be noted that baseline correction can eliminate the stable components of the background magnetic field, making the local magnetic field anomalies caused by defects more obvious, thereby facilitating subsequent defect identification and positioning. By eliminating irrelevant background information, slight changes in abnormal signals can be amplified, thereby improving the ability to detect subtle defects. The corrected data forms a clear contrast with the standard baseline data, making the test results more targeted and accurate, and helping to determine the existence and extent of defects. The obtained abnormal three-axis magnetic field intensity data provides accurate input data for the subsequent calculation of the magnetic field gradient tensor and extraction of defect characteristic parameters, thereby ensuring the reliability and accuracy of the entire detection process.

[0073] In some embodiments of the present application, a second three-axis magnetic field gradient tensor is obtained based on the abnormal three-axis magnetic field intensity data, and is expressed as:

[0074] ;

[0075] in, is the second triaxial magnetic field gradient tensor.

[0076] Specifically, we first perform spatial differential operations on the abnormal magnetic field intensity data after baseline correction and denoising, calculate its rate of change in the second and third axis directions, and then construct the magnetic field gradient tensor. Each component of this tensor describes in detail the local changes of the magnetic field in different directions, and can reflect the amplitude and trend of the abnormal changes in the magnetic field.

[0077] It should be noted that the magnetic field gradient tensor can intuitively display the sharp changes in the local magnetic field caused by pipeline defects, making the defective area show obvious gradient anomalies in the data, thereby providing an effective basis for defect location. By extracting gradient information, the signal of weak magnetic field changes can be amplified, and the normal area and the defective area can be further distinguished, thereby improving the detection sensitivity and accuracy. The obtained gradient tensor, as an important parameter of defect characteristics, provides rich and detailed data support for subsequent defect feature extraction, size quantification and intelligent judgment, which helps to achieve more accurate defect assessment and risk analysis.

[0078] In some embodiments of the present application, defect features are extracted based on the abnormal magnetic field signal data set to obtain characteristic parameters, specifically: defect features are extracted based on the abnormal magnetic field signal data set to obtain the peak amplitude of the abnormal magnetic field intensity data, the maximum gradient value of the second three-axis magnetic field gradient tensor, the abnormal area width, the phase lag and the multi-frequency response ratio.

[0079] It should be noted that by extracting characteristic parameters of multiple dimensions, a comprehensive evaluation of pipeline defects can be performed from multiple angles such as intensity, gradient, spatial distribution, phase and frequency response. This multi-dimensional feature extraction method enables the physical characteristics of defects to be fully reflected, ensuring that the detection results are more accurate. Each characteristic parameter plays a role in specific defect attributes. The peak amplitude reflects the abnormal intensity of the magnetic field at the defect, the maximum gradient value reveals the location of the sharp change of the magnetic field, the width of the abnormal area indicates the extension range of the defect, and the phase lag and multi-frequency response ratio provide dynamic and spectral information of the defect response. Combining these parameters can significantly improve the sensitivity and accuracy of defect identification and positioning. The normalized characteristic parameters are input into the pre-trained neural network model to realize the automatic determination of defects (such as quantification of defect depth, width, and length), effectively improve the intelligent level of online detection and data processing, and reduce the subjective errors of manual intervention. The multi-feature extraction method can effectively distinguish normal signals from defect signals in complex field environments, enhance the robustness of the system to noise and interference, and provide solid data support for the safe operation of pipelines.

[0080] In some embodiments of the present application, the specific content of obtaining the peak amplitude of abnormal magnetic field strength data is: using a sliding window method to obtain candidate peaks, threshold screening the candidate peaks, and finally peak merging to obtain the peak amplitude of abnormal magnetic field strength data.

[0081] Specifically, the abnormal magnetic field intensity data after noise processing and baseline correction is divided into several continuous small segments (i.e., "windows"), and the maximum value or mutation point of the local signal is detected in each window to identify preliminary candidate peaks. This step can effectively capture the signal changes in the local area, ensure that no possible defect response is missed, and perform threshold judgment on the preliminary candidate peaks, retaining only those peaks whose amplitude exceeds the preset threshold, thereby eliminating false peaks caused by environmental noise or accidental fluctuations. This process ensures that only real signals related to pipeline defects are extracted, effectively improving the accuracy of detection. Considering that the same defect may produce multiple similar peaks in a continuous sliding window, these adjacent and overlapping candidate peaks are finally merged to form a single abnormal peak amplitude. This not only avoids repeated counting, but also more accurately reflects the abnormal characteristics of the magnetic field in the defect area.

[0082] It should be noted that the most representative peak amplitude in the abnormal magnetic field signal can be accurately extracted through the sliding window method and subsequent threshold screening and peak merging. This parameter directly reflects the intensity of the magnetic field response at the defect, providing key data for qualitative and quantitative analysis of defects. Threshold screening of candidate peaks can effectively eliminate false peaks caused by noise or signal fluctuations, while peak merging processing avoids interference from repeated counting, ensuring that the final peak data is more stable and reliable, thereby improving the anti-interference ability of the overall detection system. By accurately extracting and quantifying abnormal peaks, it can provide a high-quality data basis for the subsequent normalization of defect characteristic parameters and the input of the neural network model, thereby achieving more accurate judgment of defect depth, width and length, and improving the intelligence and automation level of the detection system.

[0083] In some embodiments of the present application, the specific content of obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor is: taking the modulus length of the second three-axis magnetic field gradient tensor, searching for the maximum gradient value, and obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor and the corresponding position.

[0084] Specifically, in the process of calculating the second three-axis magnetic field gradient tensor, the modulus of the gradient tensor of each measurement point is first calculated, that is, the gradient components in the three-axis direction are combined into a vector, and its Euclidean norm is calculated to obtain the overall intensity of the magnetic field change at the point. Next, by traversing the modulus data of all measurement points, a search algorithm (such as a global search or a local extreme value extraction method) is used to find the maximum value of the gradient modulus in the entire area, and the specific location of the maximum value is recorded. The final result of this process is the maximum gradient value of the second three-axis magnetic field gradient tensor and its corresponding spatial coordinates.

[0085] It should be noted that the maximum gradient value usually appears in the area where the magnetic field changes most dramatically, and these areas often correspond to the edge or mutation of the pipeline defect. By determining the maximum gradient value and its position, the defect range and boundary can be calibrated more accurately. Taking the modulus length of the gradient tensor and searching for the maximum value can quantitatively describe the intensity of the local magnetic field anomaly, provide key parameters for defect severity assessment, and help distinguish between minor anomalies and serious defects. After obtaining the maximum gradient value and its corresponding position, it can serve as a strong support for subsequent defect feature extraction and intelligent judgment, improving the accuracy and robustness of defect identification in the entire detection process. This parameter not only intuitively reflects the change in the magnetic field gradient caused by the defect, but also provides important quantitative data for the input neural network model, thereby helping to achieve accurate prediction of defect depth, width and length.

[0086] In some embodiments of the present application, the specific content of obtaining the abnormal area width is: according to a preset threshold, mark all continuous areas where the abnormal magnetic field strength data is greater than the preset threshold, find the left and right boundaries from the area, and finally obtain the abnormal area width.

[0087] Specifically, in the process of obtaining the width of the abnormal area, the processed abnormal magnetic field intensity data is first scanned according to a preset threshold value to identify all points with values ​​higher than the threshold value. Then, these points higher than the threshold value are clustered to classify adjacent and continuous points into the same abnormal area. Subsequently, the distribution of data points in each abnormal area is analyzed to determine the left and right boundary positions of the area. Finally, the width of the abnormal area is calculated based on the distance between the left and right boundaries, thereby quantitatively describing the lateral extension range of the defect.

[0088] It should be noted that by setting a threshold to distinguish abnormal signals from normal signals, the area affected by the defect can be accurately marked, thereby providing an intuitive and clear definition for defect location. The width of the abnormal area, as a key parameter of the defect, can quantitatively reflect the degree of lateral expansion of the defect, which is helpful for the subsequent comprehensive evaluation of the size, shape and severity of the defect. The continuous area marking and boundary extraction method can effectively eliminate isolated noise interference, making the extracted width data more stable and accurate, providing reliable basic data for the entire defect detection process. The accurately calculated abnormal area width can be normalized and input into the intelligent model as part of the subsequent feature extraction, which helps to realize the automatic determination of parameters such as defect depth, width and length, thereby improving the automation level and diagnostic accuracy of the detection system.

[0089] In some embodiments of the present application, the specific content of obtaining the phase lag is: using the FFT method and the zero-crossing detection method to obtain the phase lag for the excitation signal generated by the adjustable magnetic field excitation coil and the abnormal magnetic field strength data;

[0090] Specifically, in the process of obtaining the phase lag, the periodic excitation signal generated by the adjustable magnetic field excitation coil is first used as a reference signal, and the abnormal magnetic field strength data in the pipeline is collected at the same time. In order to ensure the accuracy of the calculation, the collected magnetic field strength signal needs to be properly preprocessed, such as removing power frequency noise, high-frequency noise and motion noise, so as to reduce the impact of external interference on the phase calculation. Subsequently, the two sets of signals are Fourier transformed (FFT) respectively to extract their phase information on the main frequency component. In order to further improve the accuracy of the phase lag calculation, the zero-crossing detection method is used to accurately identify the zero-crossing point of the signal, and combined with the phase information obtained by FFT calculation, finally a high-resolution phase lag is obtained.

[0091] It should be noted that the phase lag reflects the change in the propagation characteristics of the magnetic field signal caused by the defect. Defects of different types, depths or materials may cause different degrees of phase lag. Therefore, this parameter helps to analyze the characteristics of the defect and improve the ability to distinguish the defect types. The FFT method can extract the phase information of the main frequency components with high precision, and the zero-crossing detection method can further improve the resolution of the phase lag calculation, making the detection of defects more accurate and reducing the errors caused by noise interference. As one of the important features of the abnormal magnetic field signal data set, the phase lag can be combined with information such as the change in magnetic field intensity and magnetic field gradient to more comprehensively describe the defect characteristics and provide an important basis for the subsequent estimation of the depth, width and length of the defect.

[0092] In some embodiments of the present application, the multi-frequency response ratio is obtained by performing bandpass filtering on the abnormal magnetic field intensity data, and extracting the amplitudes at the highest Hz and the lowest Hz for comparison to obtain the multi-frequency response ratio.

[0093] Specifically, in the process of obtaining the multi-frequency response ratio, the abnormal magnetic field intensity data is first band-pass filtered to extract the signal components within a specific frequency range. The parameter setting of the band-pass filter needs to be based on the detection requirements of the system, and usually multiple frequency bands are selected to eliminate the interference of irrelevant low-frequency or high-frequency noise on the signal. Subsequently, the amplitude information of the highest frequency component (highest Hz) and the lowest frequency component (lowest Hz) is extracted from the filtered signal, and their ratio is calculated, that is, the multi-frequency response ratio. This ratio can be used to measure the distribution of abnormal magnetic field signals on different frequency components, thereby reflecting the frequency response characteristics of pipeline defects.

[0094] It should be noted that the multi-frequency response ratio can reflect the energy distribution characteristics of the magnetic field anomaly signal in different frequency ranges. Different types of defects (such as corrosion, cracks, metal inclusions, etc.) will cause differences in the response of the magnetic field in different frequency bands. For example, shallow surface defects may have a stronger response at higher frequencies, while deep defects may have a higher proportion of energy in low-frequency components. Therefore, by calculating the multi-frequency response ratio, different types of defects can be effectively distinguished and the recognition accuracy of defects can be improved. Since defects of different sizes and depths have different response characteristics at different frequencies, the multi-frequency response ratio can be used as an important parameter for estimating defect size and depth. For example, when the depth of the defect increases, the high-frequency component of the magnetic field anomaly signal may gradually attenuate, while the low-frequency component is relatively enhanced, resulting in a change in the multi-frequency response ratio. Therefore, combined with other characteristic parameters (such as magnetic field gradient tensor, abnormal magnetic field intensity peak, etc.), the accuracy of defect size and depth estimation can be further improved.

[0095] In some embodiments of the present application, the characteristic parameters are normalized and input into the pre-trained neural network model to obtain the defect depth, defect width and defect length. Specifically, in the process of defect size estimation, the characteristic parameters extracted from the abnormal magnetic field signal data set (such as the peak amplitude of the abnormal magnetic field intensity data, the maximum gradient value of the magnetic field gradient tensor, the width of the abnormal area, the phase lag, the multi-frequency response ratio, etc.) need to be normalized. The purpose of normalization is to eliminate the differences in the numerical magnitude of different features and make them distributed in the same numerical range, so as to avoid the dominant influence of certain features on model training due to excessive values, and at the same time improve the convergence speed and prediction stability of the neural network model.

[0096] The normalized feature data is then input into a pre-trained neural network model, which is trained based on historical data and can learn the complex mapping relationship between feature parameters and defect dimensions (depth, width, length). Through the forward calculation of the model, the key geometric dimensions of the defect, namely defect depth, defect width and defect length, are finally output, providing a basis for pipeline health status assessment and defect severity analysis.

[0097] It should be noted that the nonlinear mapping capability of the neural network model can be used to improve the accuracy of defect size prediction. Traditional defect size calculation methods often rely on specific empirical formulas or simple linear regression, which are difficult to adapt to complex magnetic field signal changes. The neural network model can learn the nonlinear mapping relationship between high-dimensional features and can be trained based on a large amount of historical defect data to make the prediction results more accurate, especially suitable for complex morphological estimation of pipeline defects. For example, for the same abnormal magnetic field characteristics, different defect depths, widths, and lengths may result in different change patterns of magnetic field signals. The neural network can effectively extract these patterns and make accurate size inferences.

[0098] The present invention improves the detection performance based on a high-sensitivity fluxgate probe. The present invention arranges a circular fluxgate probe array on the pipeline detection robot and configures an adjustable magnetic field excitation coil around it. The present invention can realize the fine capture of weak magnetic field abnormal signals, greatly improving the detection sensitivity and resolution of pipeline defects; the first three-axis magnetic field data of the defect-free area is collected before the defect and the baseline magnetic field data is generated, which can provide an accurate reference for the subsequent baseline correction, thereby ensuring the accuracy of the abnormal magnetic field signal extraction; the present invention adopts a multi-level noise processing scheme: first, an infinite impulse response (IIR) filter is used to perform zero-phase filtering to effectively filter out the power frequency noise; then, a wavelet threshold denoising method is used to decompose the signal, and soft threshold processing is performed on the wavelet coefficients of each layer, and then the signal is reconstructed to suppress high-frequency noise; finally, the three-axis acceleration data of the pipeline detection robot is collected, and the motion noise is modeled and compensated in combination with the least squares method, so as to further improve the purity of the signal; baseline correction The positive process compares the noise-processed data with the previously collected baseline magnetic field data, so as to accurately separate the abnormal magnetic field signal and ensure that the detection result is not affected by environmental and motion interference; the present invention uses the abnormal three-axis magnetic field intensity data after baseline correction, and further calculates the second three-axis magnetic field gradient tensor to reflect the spatial distribution characteristics of the magnetic field change; through the steps of sliding window, threshold screening and peak merging, the peak amplitude of the abnormal magnetic field signal can be accurately extracted; at the same time, by taking the modulus length of the gradient tensor and searching for the maximum gradient value, the key parameters of the magnetic field gradient are obtained; in addition, it also combines multiple parameters such as abnormal area width, phase lag and multi-frequency response ratio to fully reflect the multi-dimensional characteristics of pipeline defects and provide sufficient information for subsequent intelligent judgment; multi-level noise filtering and motion noise compensation technology enable the present invention to effectively deal with the interference caused by complex environment in actual pipeline detection, such as power frequency electromagnetic noise, environmental high-frequency interference and dynamic noise caused by robot movement. Baseline correction ensures that the consistency and stability of the data can be maintained even when the detection conditions change, thereby greatly improving the overall reliability and stability of the detection system.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for improving detection performance based on a high-sensitivity fluxgate probe, characterized in that: include: An annular fluxgate probe array is arranged on the pipeline detection robot, and an adjustable magnetic field excitation coil is arranged around the annular fluxgate probe array; Controlling the pipeline inspection robot to move in a defect-free area in the pipeline, collecting first three-axis magnetic field intensity data of the defect-free area, and generating baseline magnetic field data according to the first three-axis magnetic field intensity data; Control the pipeline inspection robot to move in the defective area in the pipeline, start the adjustable magnetic field excitation coil, and collect the second three-axis magnetic field intensity data, phase difference and position code in the defective area; De-noising the second three-axis magnetic field intensity data to obtain de-noised three-axis magnetic field intensity data, and performing baseline correction to obtain abnormal three-axis magnetic field intensity data; A second three-axis magnetic field gradient tensor is obtained based on the abnormal three-axis magnetic field intensity data, and the abnormal three-axis magnetic field intensity data, the second three-axis magnetic field gradient tensor, the phase difference and the position code are integrated into an abnormal magnetic field signal data set; Defect features are extracted based on the abnormal magnetic field signal data set to obtain feature parameters, which are normalized and input into the pre-trained neural network model to obtain defect depth, defect width and defect length.

2. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 1 is characterized in that: include: The content of generating the baseline magnetic field data according to the first three-axis magnetic field data is: performing mean calculation according to multiple groups of first three-axis magnetic field intensity data, expressed as: ； in, For the coordinate axis The first three-axis magnetic field strength data on is the number of data sets, is the baseline magnetic field data.

3. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 2 is characterized in that: include: The second three-axis magnetic field intensity data is denoised to obtain denoised three-axis magnetic field intensity data, and a baseline correction is performed to obtain abnormal three-axis magnetic field intensity data. Specifically, the second three-axis magnetic field intensity data is firstly subjected to power frequency noise filtering, and then high-frequency noise is suppressed, and finally motion noise is compensated to obtain denoised three-axis magnetic field intensity data, and the denoised three-axis magnetic field intensity data is subjected to baseline correction to obtain abnormal three-axis magnetic field intensity data.

4. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 3 is characterized in that: include: The filtering of the second three-axis magnetic field intensity data for power frequency noise is specifically performed by: using an infinite impulse response filter to perform zero-phase filtering; The high-frequency noise is then suppressed, specifically: a wavelet threshold denoising method is used to decompose the second three-axis magnetic field intensity data after the power frequency noise is filtered out into five layers of wavelet coefficients, soft threshold processing is performed on each layer, and after processing, the signal is reconstructed to obtain the three-axis magnetic field intensity data after high-frequency denoising; The compensation of motion noise is specifically performed as follows: collecting the three-axis acceleration of the pipeline inspection robot, obtaining motion noise data by the least square method, and finally subtracting the motion noise data from the three-axis magnetic field intensity data after high-frequency denoising to obtain the denoised three-axis magnetic field intensity data, which is expressed as: + ； in, is the motion noise data, is the three-axis acceleration of the pipeline inspection robot, is the first fitting coefficient, is the second fitting coefficient, is the third fitting coefficient; ； in, is the denoised three-axis magnetic field intensity data, It is the three-axis magnetic field intensity data after high-frequency denoising; The de-noised three-axis magnetic field intensity data is baseline corrected to obtain abnormal three-axis magnetic field intensity data, which is expressed as: ； in, Abnormal three-axis magnetic field intensity data.

5. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 4 is characterized in that: include: The second three-axis magnetic field gradient tensor is obtained based on the abnormal three-axis magnetic field intensity data, and is expressed as: ； in, is the second triaxial magnetic field gradient tensor.

6. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 5 is characterized in that: include: The defect feature extraction is performed based on the abnormal magnetic field signal data set to obtain feature parameters, specifically: defect feature extraction is performed based on the abnormal magnetic field signal data set to obtain the peak amplitude of the abnormal magnetic field intensity data, the maximum gradient value of the second three-axis magnetic field gradient tensor, the abnormal area width, the phase lag and the multi-frequency response ratio.

7. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 6 is characterized in that: include: The specific content of obtaining the peak amplitude of the abnormal magnetic field strength data is: using a sliding window method to obtain candidate peaks, performing threshold screening on the candidate peaks, and finally performing peak merging to obtain the peak amplitude of the abnormal magnetic field strength data.

8. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 7 is characterized in that: include: The specific content of obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor is: taking the modulus length of the second three-axis magnetic field gradient tensor, searching for the maximum gradient value, and obtaining the maximum gradient value of the second three-axis magnetic field gradient tensor and the corresponding position.

9. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 8, characterized in that: include: The abnormal area width is specifically as follows: according to a preset threshold, marking all continuous areas where the abnormal magnetic field strength data is greater than the preset threshold, finding the left and right boundaries from the area, and finally obtaining the abnormal area width.

10. The method for improving detection performance based on a high-sensitivity fluxgate probe according to claim 9, characterized in that: include: The specific content of the phase lag is: the phase lag is obtained by using the FFT method and the zero-crossing detection method on the excitation signal generated by the adjustable magnetic field excitation coil and the abnormal magnetic field strength data; The multi-frequency response ratio specifically includes: performing band-pass filtering on the abnormal magnetic field strength data, and extracting the amplitudes at the highest Hz and the lowest Hz for comparison to obtain the multi-frequency response ratio.

Images