Method for automatically adjusting eddy current detection system by using binocular video measurement
The binocular video measurement system collects weld surface images in real time and performs three-dimensional reconstruction and feature recognition, automatically adjusting eddy current detection parameters, solving the sensitivity and accuracy problems in complex weld detection, eliminating false defect signals, and achieving efficient and reliable non-destructive testing.
Patent Information
- Application Number
- CN202510833462.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-09-19
AI Technical Summary
Existing weld array eddy current detection technology has low detection sensitivity and accuracy when facing welds with complex geometric features, and is seriously interfered by false defect signals. Manual adjustment is inefficient, making it difficult to achieve optimal detection status.
A binocular video measurement system is used to collect stereo images of the weld surface in real time. Through 3D point cloud reconstruction and visual feature recognition, the optimal eddy current detection parameters are generated. The weld geometric mutation characteristics and eddy current imaging signals are matched in a unified spatial coordinate system, and the detection parameters are automatically adjusted to eliminate false defect signals.
It achieves a high degree of fit between eddy current detection parameters and weld surface conditions, improves detection sensitivity and accuracy, reduces interference from false defect signals, forms an automated detection process with closed-loop feedback, and improves detection reliability and efficiency.
Smart Images

Figure CN120668777A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of non-destructive testing, and in particular to a method for automatically adjusting an eddy current testing system by utilizing binocular video measurement. Background Art
[0002] Array eddy current testing (ACET) technology is widely used for nondestructive testing in aerospace, rail transit, energy equipment, and other fields. By stimulating eddy current fields with multiple coils, it can efficiently identify surface and near-surface defects such as cracks, corrosion, and inclusions. It is particularly well-suited for quality inspection of conductive materials, such as steel structures and rail welds. However, existing ACET technology for welds faces numerous challenges in practical application. Weld surface geometric features, such as roughness, reinforcement, undercuts, dents, and misalignment, are complex and highly variable. Array eddy current testing requires high surface flatness and consistency. Consequently, some rough welds can easily cause dynamic variations in the probe lift-off distance during inspection, leading to eddy current field distortion. Because varying surface topography alters the distribution and intensity of the eddy current field, it is difficult to precisely adapt the inspection parameters to actual operating conditions, significantly reducing detection sensitivity and accuracy. Furthermore, sudden changes in weld surface geometry can easily generate false defect signals. These false signals can confuse inspectors with true defect signals, significantly interfering with the accuracy of the inspection results. In order to improve the detection sensitivity, the detection system needs to be adjusted. The current detection method requires manual adjustment of the detection system according to the on-site conditions. However, this manual adjustment method is not only inefficient, but also makes it difficult to make the eddy current detection system work stably in the optimal state, which seriously affects the accuracy and reliability of the eddy current detection results. Summary of the Invention
[0003] To solve the above problems, the present invention provides a method for automatically adjusting an eddy current detection system using binocular video measurement. The present invention is implemented as follows:
[0004] A method for automatically adjusting an eddy current testing system using binocular video measurement, used in weld array eddy current testing, comprises the following steps:
[0005] S1, through the integrated binocular video sensor set in front of the array eddy current detection sensor, the surface stereo image of the current detection area of the weld to be tested is collected in real time;
[0006] S2. Reconstructing a three-dimensional point cloud based on the surface stereo image to extract surface roughness parameters of the current inspection area of the weld to be tested;
[0007] S3, performing visual feature recognition analysis on the surface stereo image and / or the reconstructed three-dimensional point cloud to obtain surface geometric mutation features of the current inspection area of the weld to be tested;
[0008] The surface geometric mutation features include weld reinforcement, undercut, depression, staggered edge and their position and size parameters;
[0009] S4. Generating an optimal working parameter set for the eddy current testing system through a preset eddy current parameter mapping model according to the surface roughness parameters and the surface geometric mutation characteristics;
[0010] The optimal working parameter set includes excitation frequency, probe gain and phase angle;
[0011] S5. Before the array eddy current detection sensor arrives at the current detection area, dynamically loading the optimal working parameter set described in step S4 to the array eddy current detection instrument;
[0012] S6. Perform array eddy current detection and obtain eddy current imaging signals;
[0013] S7, matching the position of the surface geometric mutation feature identified in step S3 with the eddy current imaging signal obtained in step S6 in a unified spatial coordinate system;
[0014] S8. Based on the matching results, locate the abnormal signal area in the eddy current imaging signal that coincides with the surface geometric mutation feature;
[0015] S9, marking the abnormal signal area as a pseudo defect signal or performing signal suppression processing, and outputting a corrected eddy current imaging detection result.
[0016] Furthermore, the preset eddy current parameter mapping model in step S4 is trained and generated by the following steps:
[0017] Collect binocular vision data and eddy current imaging signals of weld samples with different surface roughness and surface geometric mutation characteristics;
[0018] The optimal working parameter set corresponding to each sample is calibrated through experiments;
[0019] The machine learning regression model is trained with surface roughness parameters, type, location and size parameters of geometric mutation features as input and the optimal working parameter set as output;
[0020] Verify whether the detection signal-to-noise ratio of the optimal working parameter set output by the model on unknown samples reaches the preset threshold.
[0021] Furthermore, the triggering condition for dynamically loading the optimal working parameter set in step S5 is:
[0022] The parameter loading timing is calculated in real time according to the displacement speed of the array eddy current detection sensor and the leading distance of the binocular video sensor.
[0023] Furthermore, the position parameters of the surface geometric mutation features in step S7 are represented by spatial coordinates (x, y, z) in a three-dimensional point cloud coordinate system;
[0024] The size parameters include the bite depth, the height difference of the staggered edges, the area of the recessed area and the depth gradient value.
[0025] Furthermore, the signal suppression processing in step S9 includes any of the following methods:
[0026] Setting the amplitude of the eddy current imaging signal in the overlapping abnormal signal area to zero;
[0027] Marking pseudo-defect areas in eddy current imaging test results;
[0028] The gain threshold of the local eddy current imaging signal is dynamically adjusted based on the size parameters of the geometric mutation characteristics to suppress abnormal signals below the threshold.
[0029] Furthermore, the input of the eddy current parameter mapping model also includes the conductivity of the weld material and the real-time detection speed;
[0030] The output optimal working parameter set further includes a weight coefficient of each frequency channel in multi-frequency eddy current testing.
[0031] An eddy current detection system automatically adjusted using binocular video measurement, comprising:
[0032] Binocular video module: includes a binocular video sensor and light source integrated in front of the array eddy current detection sensor, used to collect stereo images of the weld surface to be tested in real time;
[0033] 3D reconstruction and feature recognition module: connects to the binocular video module and performs the following operations:
[0034] Reconstruct the three-dimensional point cloud of the weld surface to be measured based on the stereo image;
[0035] Extract the surface roughness parameters of the weld to be tested;
[0036] Identify the geometric mutation features of the weld surface and their location and size parameters;
[0037] Dynamic parameter control module: connected to the 3D reconstruction and feature recognition modules, used to:
[0038] According to the surface roughness parameters and surface geometric mutation characteristics, the optimal working parameter set is generated through the preset eddy current parameter mapping model;
[0039] Before the array eddy current detection sensor arrives at the detection area, the optimal working parameter set is dynamically loaded into the array eddy current detection instrument;
[0040] Array eddy current detection module: includes array eddy current detection sensors and excitation acquisition circuits, performs array eddy current detection based on the loaded optimal working parameter set and outputs eddy current imaging signals;
[0041] Data fusion and correction module: connects the 3D reconstruction and feature recognition module and the array eddy current detection module, and is used to:
[0042] Match the position of the surface geometric mutation feature with the eddy current imaging signal in a unified coordinate system;
[0043] Locate abnormal signal areas that coincide with surface geometric abrupt changes;
[0044] Mark or suppress false defect signals and output corrected eddy current imaging test results;
[0045] Synchronous control unit: coordinates the acquisition timing of the binocular video module and array eddy current detection module.
[0046] Furthermore, the synchronization control unit includes:
[0047] Displacement encoder, real-time feedback array eddy current detection sensor position and calculation of detection speed;
[0048] The FPGA chip triggers parameter loading based on the detection speed and the leading distance of the binocular video module.
[0049] Compared with the prior art, the present invention has the following beneficial effects:
[0050] The present invention uses binocular vision to collect stereo images of the weld surface in real time, and with the help of three-dimensional point cloud reconstruction and visual feature recognition technology, it can accurately obtain various geometric features of the weld surface, including parameter information such as roughness, height, and undercut. Based on the acquired image data, the optimal working parameter set is generated using a preset eddy current parameter mapping model, realizing dynamic adjustment of the eddy current detection system parameters. Compared with the traditional manual adjustment method, this technology makes the detection parameters highly consistent with the weld surface condition, greatly improving the detection sensitivity and accuracy. In terms of pseudo-defect signal processing, the present invention uses a deep learning algorithm to match and analyze the geometric mutation characteristics of the weld surface with the eddy current imaging signal in a unified spatial coordinate system, integrating visual detection and eddy current detection information, accurately locating the pseudo-defect signal area, effectively eliminating interference signals caused by surface morphology, and at the same time enhancing the recognition ability of real defect signals and improving defect recognition accuracy. From binocular visual perception, eddy current parameter adjustment, eddy current scanning, and detection result correction, an automated process of closed-loop feedback detection is formed, reducing manual intervention and achieving reliable and fast non-destructive testing results. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the examples of the present invention or the technical solutions in the prior art or the drawings required for the description of the prior art, a brief introduction is given. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0052] Figure 1 A schematic diagram of a detection structure for this implementation.
[0053] Figure 2 Schematic diagram of the detection system module of this embodiment.
[0054] Figure 3 Schematic diagram of the detection method of the present invention.
[0055] 10-Binocular video sensor; 20-Array eddy current detection sensor. DETAILED DESCRIPTION
[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention.
[0057] In industrial nondestructive testing, weld inspection presents a variety of complex working conditions, such as high surface roughness, complex geometric structures, and material heterogeneity, all of which directly affect detection accuracy and reliability. Therefore, when using eddy current detection sensors for weld surface defect detection, it is necessary to adjust the detection parameters in real time according to the weld surface conditions.
[0058] In this embodiment, the accuracy of detection is improved by combining multimodal sensor fusion with AI-assisted analysis. Specifically, a method for automatically adjusting an eddy current detection system using binocular video measurement is used in weld array eddy current detection, comprising the following steps:
[0059] S1, through the integrated binocular video sensor set in front of the array eddy current detection sensor, the surface stereo image of the current detection area of the weld to be tested is collected in real time;
[0060] S2. Reconstructing a three-dimensional point cloud based on the surface stereo image to extract surface roughness parameters of the current inspection area of the weld to be tested;
[0061] S3, performing visual feature recognition analysis on the surface stereo image and / or the reconstructed three-dimensional point cloud to obtain surface geometric mutation features of the current inspection area of the weld to be tested;
[0062] The surface geometric mutation features include weld reinforcement, undercut, depression, staggered edge and their position and size parameters;
[0063] S4. Generating an optimal working parameter set for the eddy current testing system through a preset eddy current parameter mapping model according to the surface roughness parameters and the surface geometric mutation characteristics;
[0064] The optimal working parameter set includes excitation frequency, probe gain and phase angle;
[0065] S5. Before the array eddy current detection sensor arrives at the current detection area, dynamically loading the optimal working parameter set described in step S4 to the array eddy current detection instrument;
[0066] S6. Perform array eddy current detection and obtain eddy current imaging signals;
[0067] S7, matching the position of the surface geometric mutation feature identified in step S3 with the eddy current imaging signal obtained in step S6 in a unified spatial coordinate system;
[0068] S8. Based on the matching results, locate the abnormal signal area in the eddy current imaging signal that coincides with the surface geometric mutation feature;
[0069] S9, marking the abnormal signal area as a pseudo defect signal or performing signal suppression processing, and outputting a corrected eddy current imaging detection result.
[0070] Example 1: Hot-rolled rail welds on railways are inspected. Typical defects include gray spots, incomplete penetration, and cracks. A binocular video sensor is used to capture geometric variations on the weld surface. The binocular video sensor can use a ring-shaped red LED as a supplemental light source to suppress reflections.
[0071] The acquired images are reconstructed into three-dimensional point clouds and subjected to roughness analysis, with particular attention paid to the base material area and the weld center. Visual feature recognition and analysis are performed on the surface stereo images and / or the reconstructed three-dimensional point clouds to obtain the surface geometric mutation features of the current inspection area of the weld to be tested. A deep learning model for recognition based on the surface geometric mutation features is constructed, and YOLOv5s is used for network training. 2,000 annotated weld defect images are collected in advance as YOLOv5s network training data.
[0072] After identifying the surface geometry mutation features, the model is processed and the roughness Ra value is defined as the output result. Parameter mapping is performed with eddy current testing. The excitation frequency, probe gain, and phase angle are quickly and automatically configured according to the different roughness values before the test is performed. Furthermore, the preset eddy current parameter mapping model in step S4 is trained and generated through the following steps:
[0073] Collect binocular vision data and eddy current imaging signals of weld samples with different surface roughness and surface geometric mutation characteristics;
[0074] The optimal working parameter set corresponding to each sample is calibrated through experiments;
[0075] The machine learning regression model is trained with surface roughness parameters, type, location and size parameters of geometric mutation features as input and the optimal working parameter set as output;
[0076] Verify whether the detection signal-to-noise ratio of the optimal working parameter set output by the model on unknown samples reaches the preset threshold.
[0077] Furthermore, the triggering condition for dynamically loading the optimal working parameter set in step S5 is:
[0078] The parameter loading timing is calculated in real time according to the displacement speed of the array eddy current detection sensor and the leading distance of the binocular video sensor.
[0079] Furthermore, the position parameters of the surface geometric mutation features described in step S7 are represented by spatial coordinates (x, y, z) in a three-dimensional point cloud coordinate system. In this embodiment, the spatial coordinate system is based on the starting point of the weld to be measured, with X along the direction of the weld, and the spatial coordinate system is used as the global reference system. The camera coordinate system and the eddy current detection sensor coordinate system can also be linked. The camera coordinate system takes the optical center of the binocular lens as the origin, and the Z axis is along the optical axis. It is converted into a spatial coordinate system through calibration. The eddy current sensor coordinate system takes the center of the array eddy current detection sensor as the origin, and the Z axis is perpendicular to the measured surface. The eddy current detection sensor's posture matrix is associated with the spatial coordinate system.
[0080] The size parameters include the bite depth, the height difference of the staggered edges, the area of the recessed area and the depth gradient value.
[0081] Furthermore, the signal suppression processing in step S9 includes any of the following methods:
[0082] The amplitude of the eddy current imaging signal in the overlapping abnormal signal area is set to zero; for example, to confirm the deterministic interference caused by geometric mutations (such as misalignment and residual height mutations), the amplitude zeroing method has high real-time performance and can quickly improve the signal-to-noise ratio by setting the eddy current signal amplitude to zero.
[0083] The pseudo-defect area is marked in the eddy current imaging test results; then the pseudo-defect area can be covered with a red semi-transparent polygon. This method has high information retention and can be used for review processing.
[0084] The gain threshold of the local eddy current imaging signal is dynamically adjusted based on the size parameters of the geometric mutation characteristics to suppress abnormal signals below the threshold.
[0085] Furthermore, the input of the eddy current parameter mapping model also includes the conductivity of the weld material and the real-time detection speed; the conductivity of the material is measured offline by an eddy current conductivity meter, and the encoder is used to provide real-time feedback.
[0086] The output optimal working parameter set further includes a weight coefficient of each frequency channel in multi-frequency eddy current testing.
[0087] An eddy current detection system automatically adjusted using binocular video measurement, comprising:
[0088] Binocular video module: includes a binocular video sensor and light source integrated in front of the array eddy current detection sensor, used to collect stereo images of the weld surface to be tested in real time;
[0089] 3D reconstruction and feature recognition module: connects to the binocular video module and performs the following operations:
[0090] Reconstruct the three-dimensional point cloud of the weld surface to be measured based on the stereo image;
[0091] Extract the surface roughness parameters of the weld to be tested;
[0092] Identify the geometric mutation features of the weld surface and their location and size parameters;
[0093] Dynamic parameter control module: connected to the 3D reconstruction and feature recognition modules, used to:
[0094] According to the surface roughness parameters and surface geometric mutation characteristics, the optimal working parameter set is generated through the preset eddy current parameter mapping model;
[0095] Before the array eddy current detection sensor arrives at the detection area, the optimal working parameter set is dynamically loaded into the array eddy current detection instrument;
[0096] Array eddy current detection module: includes array eddy current detection sensors and excitation acquisition circuits, performs array eddy current detection based on the loaded optimal working parameter set and outputs eddy current imaging signals;
[0097] Data fusion and correction module: connects the 3D reconstruction and feature recognition module and the array eddy current detection module, and is used to:
[0098] Match the position of the surface geometric mutation feature with the eddy current imaging signal in a unified coordinate system;
[0099] Locate abnormal signal areas that coincide with surface geometric abrupt changes;
[0100] Mark or suppress false defect signals and output corrected eddy current imaging test results;
[0101] Synchronous control unit: coordinates the acquisition timing of the binocular video module and array eddy current detection module.
[0102] Furthermore, the synchronization control unit includes:
[0103] Displacement encoder, real-time feedback array eddy current detection sensor position and calculation of detection speed;
[0104] The FPGA chip triggers parameter loading based on the detection speed and the leading distance of the binocular video module.
[0105] Example 2: Using pressure vessel steel structure weld inspection as an example, array eddy current testing was performed on the circumferential welds of the pressure vessel cylinder. The weld material is low-alloy steel, and the weld surface has certain geometric features such as excess height and undercut. Traditional inspection methods are ineffective in this scenario, prone to misjudgment and omission.
[0106] This embodiment uses binocular video measurement to automatically adjust the eddy current detection system for weld array eddy current detection. The system includes a binocular video module, a three-dimensional reconstruction and feature recognition module, a dynamic parameter control module, an array eddy current detection module, a data fusion and correction module, and a synchronization control unit. Before detection, the binocular video sensor and the distance between the light source and the array eddy current detection sensor in the binocular video module are debugged. Preferably, the distance is debugged to 10 cm in front to ensure that the weld surface image can be collected in advance. At the same time, the position of the array eddy current detection sensor is fed back in real time through the displacement encoder, and the FPGA chip calculates and sets the parameter loading timing based on the sensor displacement speed and the front distance of the binocular video sensor.
[0107] S1. Image Acquisition: The binocular video module starts operating. Preferably, the binocular video sensor acquires a real-time stereo image of the surface of the weld under test at a rate of 30 frames per second. As the inspection device moves, it continuously acquires stereo images of the surface of the weld under test, providing a data foundation for subsequent analysis and processing.
[0108] S2. 3D point cloud reconstruction and feature extraction: The 3D reconstruction and feature recognition module receives the surface stereo image acquired in step S1, identifies the feature points in the image, calculates the coordinates of the feature points in 3D space, thereby constructing a 3D point cloud model of the weld surface, and extracts the surface roughness parameters of the current inspection area of the weld to be tested from the 3D point cloud model.
[0109] S3. Obtain surface geometric mutation features: Quantify the surface roughness by calculating indicators such as the height variance of the point cloud, and use edge detection algorithms to identify geometric mutation features such as weld height, undercut depth, and misalignment height difference, as well as their position and size parameters.
[0110] S4. Generate an optimal operating parameter set: The dynamic parameter control module obtains the conductivity of the low-alloy steel in this embodiment from the material database, combines it with the real-time detection speed fed back by the displacement encoder, inputs the position and size parameters obtained in step S3 into the eddy current parameter mapping model, and uses the excitation frequency, probe gain, phase angle, and weight coefficients of each frequency channel in multi-frequency eddy current detection as output to generate the optimal operating parameter set for the eddy current detection system. In a preferred embodiment, the output excitation frequency is 10 kHz, the probe gain is 40 dB, the phase angle is 30°, and the weight coefficients of each frequency channel are 0.3, 0.5, and 0.2.
[0111] S5. Dynamic parameter loading: The synchronous control unit calculates the parameter loading sequence in real time based on the displacement speed of the array eddy current detection sensor and the leading distance of the binocular video sensor. When the calculation shows that the array eddy current detection sensor is about to reach the detection area, the dynamic parameter control module is triggered to dynamically load the optimal operating parameter set generated in step S4 into the array eddy current detection instrument, ensuring that the array eddy current detection module uses the optimal parameters for detection when entering the detection area.
[0112] S6. Array Eddy Current Testing: The Array Eddy Current Testing module performs array eddy current testing based on the loaded optimal parameter working set. The array eddy current testing sensor moves across the weld surface, where multiple coils excite the eddy current field. This excitation capture circuit collects the eddy current imaging signal and transmits it to the Data Fusion and Correction Module.
[0113] S7. Signal matching: The data fusion and correction module matches the position of the surface geometric mutation features identified by the 3D reconstruction and feature recognition module (represented by the spatial coordinates (x, y, z) in the 3D point cloud coordinate system) with the eddy current imaging signal obtained by the array eddy current detection module in a unified spatial coordinate system.
[0114] S8. Abnormal signal area positioning: Through coordinate transformation and feature matching algorithm, the abnormal signal area in the eddy current imaging signal that coincides with the surface geometric mutation feature is positioned.
[0115] S9. Output of test results: For overlapping abnormal signal regions, signal suppression processing is performed by setting the eddy current imaging signal amplitude in that region to zero, eliminating interference from false defect signals. Finally, the corrected eddy current imaging test results are output, clearly displaying the true defect information within the weld of this embodiment. Experimental verification has shown that this embodiment significantly reduces the false positive rate and significantly improves detection efficiency. This method significantly outperforms traditional detection methods, effectively ensuring the accuracy and efficiency of pressure vessel weld quality inspection.
[0116] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.
Claims
1. A method for automatically adjusting an eddy current testing system using binocular video measurement, used in weld array eddy current testing, characterized in that: The following steps are involved: S1, through the integrated binocular video sensor set in front of the array eddy current detection sensor, the surface stereo image of the current detection area of the weld to be tested is collected in real time; S2. Reconstructing a three-dimensional point cloud based on the surface stereo image to extract surface roughness parameters of the current inspection area of the weld to be tested; S3, performing visual feature recognition analysis on the surface stereo image and / or the reconstructed three-dimensional point cloud to obtain surface geometric mutation features of the current inspection area of the weld to be tested; The surface geometric mutation features include weld reinforcement, undercut, depression, staggered edge and their position and size parameters; S4, generating an optimal working parameter set for the eddy current testing system through a preset eddy current parameter mapping model according to the surface roughness parameters and surface geometric mutation characteristics described in step S2; The optimal working parameter set includes excitation frequency, probe gain and phase angle; S5. Before the array eddy current detection sensor arrives at the current detection area, dynamically loading the optimal working parameter set described in step S4 to the array eddy current detection instrument; S6. Perform array eddy current detection and obtain eddy current imaging signals; S7, matching the position of the surface geometric mutation feature identified in step S3 with the eddy current imaging signal obtained in step S6 in a unified spatial coordinate system; S8. Based on the matching results, locate the abnormal signal area in the eddy current imaging signal that coincides with the surface geometric mutation feature; S9, marking the abnormal signal area as a pseudo defect signal or performing signal suppression processing, and outputting a corrected eddy current imaging detection result.
2. The method for automatically adjusting an eddy current detection system using binocular video measurement according to claim 1, wherein: The preset eddy current parameter mapping model in step S4 is trained and generated by the following steps: Collect binocular vision data and eddy current imaging signals of weld samples with different surface roughness and surface geometric mutation characteristics; The optimal working parameter set corresponding to each sample is calibrated through experiments; The machine learning regression model is trained with surface roughness parameters, type, location and size parameters of geometric mutation features as input and the optimal working parameter set as output; Verify whether the detection signal-to-noise ratio of the optimal working parameter set output by the model on unknown samples reaches the preset threshold.
3. The method for automatically adjusting an eddy current detection system using binocular video measurement according to claim 1, wherein: The triggering condition for dynamically loading the optimal working parameter set in step S5 is: The parameter loading timing is calculated in real time according to the displacement speed of the array eddy current detection sensor and the leading distance of the binocular video sensor.
4. The method for automatically adjusting an eddy current detection system using binocular video measurement according to claim 1, wherein: The position parameters of the surface geometric mutation features in step S7 are represented by spatial coordinates (x, y, z) in a three-dimensional point cloud coordinate system; The size parameters include the bite depth, the height difference of the staggered edges, the area of the recessed area and the depth gradient value.
5. The method for automatically adjusting an eddy current detection system using binocular video measurement according to claim 1, wherein: The signal suppression process in step S9 includes any of the following methods: Setting the amplitude of the eddy current imaging signal in the overlapping abnormal signal area to zero; Marking pseudo-defect areas in eddy current imaging test results; The gain threshold of the local eddy current imaging signal is dynamically adjusted based on the size parameters of the geometric mutation characteristics to suppress abnormal signals below the threshold.
6. The method for automatically adjusting an eddy current detection system using binocular video measurement according to claim 2, wherein: The input of the eddy current parameter mapping model also includes the conductivity of the weld material and the real-time detection speed; The output optimal working parameter set further includes a weight coefficient of each frequency channel in multi-frequency eddy current testing.
7. An eddy current detection system automatically adjusted using binocular video measurement, characterized in that: include: Binocular video module: includes a binocular video sensor and light source integrated in front of the array eddy current detection sensor, used to collect stereo images of the weld surface to be tested in real time; 3D reconstruction and feature recognition module: connects to the binocular video module and performs the following operations: Reconstruct the three-dimensional point cloud of the weld surface to be measured based on the stereo image; Extract the surface roughness parameters of the weld to be tested; Identify the geometric mutation features of the weld surface and their location and size parameters; Dynamic parameter control module: connected to the 3D reconstruction and feature recognition modules, used to: According to the surface roughness parameters and surface geometric mutation characteristics, the optimal working parameter set is generated through the preset eddy current parameter mapping model; Before the array eddy current detection sensor arrives at the detection area, the optimal working parameter set is dynamically loaded into the array eddy current detection instrument; Array eddy current detection module: includes array eddy current detection sensors and excitation acquisition circuits, performs array eddy current detection based on the loaded optimal working parameter set and outputs eddy current imaging signals; Data fusion and correction module: connects the 3D reconstruction and feature recognition module and the array eddy current detection module, and is used to: Match the position of the surface geometric mutation feature with the eddy current imaging signal in a unified coordinate system; Locate abnormal signal areas that coincide with surface geometric abrupt changes; Mark or suppress false defect signals and output corrected eddy current imaging test results; Synchronous control unit: coordinates the acquisition timing of the binocular video module and array eddy current detection module.
8. The automatic adjustment eddy current detection system using binocular video measurement according to claim 7, characterized in that: The synchronization control unit includes: Displacement encoder, real-time feedback array eddy current detection sensor position and calculation of detection speed; The FPGA chip triggers parameter loading based on the detection speed and the leading distance of the binocular video module.
Citation Information
Cited By
Eddy current sensor pose adjusting method and system, medium and computer equipment
CN121535726A
