Method and system for detecting secondary machining defect of internal thread of copper pipe

Abstract

The invention relates to the technical field of copper pipe defect detection, in particular to a copper pipe internal thread secondary machining defect detection method and system, and the method comprises the steps: obtaining thread surface feature data based on multi-mode optical parameter measurement, and the multi-mode optical parameters at least comprise a polarized light inhibition reflection parameter and a multi-band spectral reflection parameter; temperature gradient parameters in the machining process are synchronously obtained, and deformation error correction is conducted on thread surface feature data in combination with dynamic compensation; according to the corrected thread surface feature data, a multi-dimensional defect judgment model is pre-trained to execute hierarchical defect detection; and integrating the coupling relationship between the multi-mode optical parameters and the electromagnetic parameters to output a defect measurement result, wherein the coupling relationship realizes multi-physical quantity fusion judgment through a preset weight distribution matrix. According to the invention, the problems of optical detection signal distortion and high microcrack omission ratio caused by surface high specular reflection in copper pipe internal thread secondary processing defect detection are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of copper tube defect detection, and in particular to a method and system for detecting secondary processing defects of internal threads of copper tubes. Background Art

[0002] Secondary machining of copper tube internal threads is a critical process in the manufacture of precision pipe fittings for applications such as refrigeration and hydraulics. The quality of this machining directly impacts the fitting's sealing performance and service life. Due to the complex thread structure and the presence of interference factors such as high temperature and vibration in the machining environment, defects such as surface microcracks and subsurface inclusions are easily generated during machining.

[0003] Existing technologies primarily rely on single optical detection methods (such as laser triangulation or structured light 3D scanning) for defect identification. This technology determines defects by analyzing the light intensity distribution or geometric features of the threaded surface. However, the high specular reflectivity caused by the polished copper tube surface severely interferes with the optical measurement signal, causing the optical signatures of defects such as surface microcracks to be drowned out by reflected noise. This significantly reduces detection accuracy and makes it impossible to effectively distinguish between machining vibration marks and actual material defects. This technical limitation forces production lines to rely on manual re-inspection, severely limiting detection efficiency and consistency.

[0004] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present disclosure and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] The present invention provides a method and system for detecting secondary processing defects of internal threads of copper pipes, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is: A method for detecting secondary processing defects of internal threads of copper pipes, the method comprising: Acquiring thread surface feature data based on multimodal optical parameter measurement, wherein the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; Synchronously acquiring temperature gradient parameters during the machining process, and performing deformation error correction on the thread surface feature data in combination with dynamic compensation; According to the corrected thread surface feature data, the pre-trained multi-dimensional defect determination model performs hierarchical defect detection; The defect measurement result is output by integrating the coupling relationship between the multimodal optical parameters and the electromagnetic parameters, wherein the coupling relationship realizes multi-physical quantity fusion judgment through a preset weight distribution matrix.

[0007] Furthermore, a hierarchical defect detection is performed, including: Obtaining a reflection intensity distribution based on the multimodal optical parameter measurement, and generating a surface defect probability distribution based on a multi-band feature correlation analysis; Setting a first determination threshold, and marking abnormal areas exceeding the first determination threshold according to the surface defect probability distribution map; Performing electromagnetic field phase response measurement in the abnormal area to obtain a sub-surface defect phase offset distribution related to the sub-surface defect of the internal thread of the copper tube; The surface defect probability distribution and the sub-surface defect phase shift distribution are weightedly fused to output the determination result of the defect type and depth level.

[0008] Furthermore, the surface defect probability distribution is generated based on the multi-band feature correlation analysis, including: performing gradient amplitude calculation on the reflection intensity distribution in the ultraviolet band, and extracting high gradient change areas as potential crack features; performing local contrast enhancement on the reflection intensity distribution in the visible light band to identify low reflectivity regions associated with material texture anomalies; Performing a spatial convolution operation on the potential crack feature and the low reflectivity area to generate an initial defect probability distribution; Based on the difference in spectral absorptivity in the near-infrared band, effective defect areas that meet preset absorption characteristics are screened in the initial defect probability distribution to generate the surface defect probability distribution.

[0009] Furthermore, a sub-surface defect phase shift distribution related to the sub-surface defect of the internal thread of the copper tube is obtained, including: Applying an alternating electromagnetic field excitation to the abnormal area, wherein the alternating electromagnetic field excitation frequency is dynamically adjusted according to the copper tube wall thickness and the preset detection depth; Collecting eddy current induction signals from the abnormal area, and obtaining real-time phase response data through orthogonal phase-locked amplification and demodulation; Performing spatial domain normalization processing on the real-time phase response data to eliminate phase distortion caused by edge effects and generate the sub-surface defect phase offset distribution; The phase shift amount in the subsurface defect phase shift distribution is converted into a defect depth level parameter based on the phase response experimental data of standard sample defects at different depths.

[0010] Furthermore, weighted fusion of the surface defect probability distribution and the sub-surface defect phase shift distribution is performed, including: Taking the peak intensity of the surface defect probability distribution as the optical characteristic component, and extracting the gradient change rate of the sub-surface defect phase offset distribution as the electromagnetic characteristic component; Performing dynamic weighted calculation on the optical characteristic component and the electromagnetic characteristic component to generate a comprehensive judgment value; When the temperature of the detection area exceeds a preset threshold, the weight coefficient of the electromagnetic characteristic component is increased; when the surface roughness index exceeds a critical value, the weight coefficient of the optical characteristic component is reduced; Comparing the comprehensive judgment value with a preset defect classification standard, if the comprehensive judgment value is in the first interval, outputting the surface crack defect type; If the comprehensive judgment value is in the second interval, the sub-surface inclusion defect type is output; If the comprehensive judgment value spans multiple intervals at the same time, the manual review mechanism is triggered.

[0011] Furthermore, the coupling relationship realizes multi-physical quantity fusion judgment through a preset weight distribution matrix, including: The weight distribution matrix is composed of an optical weight coefficient, an electromagnetic weight coefficient, and a deformation compensation weight coefficient, wherein the initial value of each weight coefficient is allocated based on the properties of the copper material and the sum is 1; Multiplying the optical eigenvector, the electromagnetic eigenvector, and the deformation eigenvector by the corresponding weight coefficients in the weight distribution matrix respectively and then superimposing them to generate a fused eigenvector; Matching the fused feature vector with a preset defect feature space, and outputting the defect type and depth level; The defect feature space is constructed by jointly distributing the optical, electromagnetic and deformation features of historical defect samples.

[0012] Furthermore, matching the fused feature vector with a preset defect feature space includes: Calculating, based on each of the weight coefficients in the weight distribution matrix, a weighted Euclidean distance between the fused feature vector and each defect category reference vector in the defect feature space; Performing density attenuation compensation on the weighted Euclidean distance according to the sample distribution density of each defect category in the defect feature space; The modified weighted Euclidean distance is mapped into a probability distribution of defect categories through a probability conversion model, and the defect type corresponding to the maximum probability is selected as the preliminary judgment result; When the maximum probability value is lower than the confidence threshold, a manual review process is triggered and the defect feature space is updated with the current fused feature vector.

[0013] Furthermore, the deformation error correction of the thread surface feature data is performed in combination with dynamic compensation, including: Collect the axial temperature distribution data of the copper tube in real time and calculate the maximum temperature difference and temperature change rate; Obtaining a radial deformation compensation amount and an axial expansion compensation amount of the copper tube based on the maximum temperature difference and the temperature change rate; Performing compensation correction on the optically measured thread tooth height based on the radial deformation compensation amount, wherein the compensation value is determined by the product relationship between the thermal expansion coefficient of the copper material and the radial deformation compensation amount; The optically measured thread lead angle is compensated and corrected based on the axial telescopic compensation amount, and the compensation value is determined by the linear relationship between the deformation angle conversion coefficient and the axial telescopic compensation amount.

[0014] A copper tube internal thread secondary processing defect detection system, the system comprising: A data acquisition module, which acquires thread surface feature data based on multimodal optical parameter measurement, where the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; The error correction module synchronously obtains the temperature gradient parameters during the processing and combines dynamic compensation to correct the deformation error of the thread surface feature data; The defect detection module performs hierarchical defect detection based on the pre-trained multi-dimensional defect judgment model based on the corrected thread surface feature data; The result output module outputs the defect measurement results based on the coupling relationship between multi-modal optical parameters and electromagnetic parameters. The coupling relationship is realized through a preset weight distribution matrix to realize the fusion judgment of multiple physical quantities.

[0015] Furthermore, the defect detection module includes: Correlation analysis unit, which obtains reflection intensity distribution based on multimodal optical parameter measurement and generates surface defect probability distribution based on multi-band feature correlation analysis; a threshold judgment unit, which sets a first judgment threshold and marks an abnormal area exceeding the first judgment threshold according to a surface defect probability distribution map; The offset measurement unit performs electromagnetic field phase response measurement in the abnormal area to obtain the sub-surface defect phase offset distribution related to the sub-surface defect of the internal thread of the copper tube; The fusion judgment unit performs weighted fusion on the surface defect probability distribution and the sub-surface defect phase offset distribution, and outputs the judgment results of the defect type and depth level.

[0016] The technical solution of the present invention can achieve the following technical effects: Multimodal optical fusion technology can effectively overcome the interference of mirror reflection, improve the accuracy of micro-crack detection, achieve accurate distinction between processing vibration marks and material defects, and reduce reliance on manual re-inspection to improve detection efficiency.

[0017] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 The figure is a flow chart of a method for detecting defects in secondary processing of internal threads of copper tubes; Figure 2 A schematic diagram of a process for performing hierarchical defect detection; Figure 3 This is a flow chart of multi-physical quantity fusion judgment; Figure 4 Schematic diagram of the process of deformation error correction. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Embodiment 1; like Figure 1 As shown, the present application provides a method for detecting secondary processing defects of internal threads of copper tubes, the method comprising: S10: Acquire thread surface feature data based on multimodal optical parameter measurement, where the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; S20: synchronously obtain the temperature gradient parameters during the processing, and perform deformation error correction on the thread surface feature data in combination with dynamic compensation; S30: Based on the corrected thread surface feature data, the pre-trained multi-dimensional defect determination model performs hierarchical defect detection; S40: Outputting defect measurement results based on the coupling relationship between multi-modal optical parameters and electromagnetic parameters, wherein the coupling relationship is used to achieve multi-physical quantity fusion judgment through a preset weight distribution matrix.

[0023] Specifically, the integrated multimodal optical measuring device is first used to scan the internal thread surface of the copper tube to obtain its surface feature data. The multimodal optical measuring device includes a polarization light module and a multi-band spectral imaging module. The polarization light module is used to suppress strong reflection interference, highlight the thread detail features, and obtain a polarization-suppressed reflection image sequence; the multi-band spectral imaging module is used to capture the change in material reflectivity under different bands, reflecting the material composition differences and roughness characteristics of the processed surface; during the processing, the temperature gradient parameters of the copper tube surface are obtained through a thermal imaging sensor or a built-in temperature sensor, and the temperature change curves before, during and after processing are recorded. Since temperature changes may cause microscopic expansion or deformation of the material, the measured thread feature data will be offset. Therefore, a dynamic compensation algorithm is used to compensate for the original optical data. The data is corrected to improve data stability and accuracy; the corrected thread surface feature data is input into the pre-trained multi-dimensional defect judgment model. The multi-dimensional defect judgment model is based on the convolutional neural network (CNN) structure and combined with the hierarchical feature extraction strategy to perform multi-layer classification of defects, including cracks, chipping, processing residues, pitch anomalies and other types, and output preliminary judgment results and confidence scores; the multi-modal optical parameters and electromagnetic parameters (such as the conductivity change signal obtained by the eddy current sensor) are fused and analyzed, and multiple parameters are weighted and combined through the preset weight distribution matrix to construct a multi-physical field coupling relationship model to improve the comprehensiveness and robustness of the detection, and finally output quantitative defect measurement results, including defect location, size, type and possible cause analysis.

[0024] Through the technical solution of the present invention, the accuracy of microcrack detection is improved, the accurate distinction between machining vibration marks and material defects is achieved, and the reliance on manual re-inspection is reduced to improve detection efficiency.

[0025] Further, if Figure 2 As shown, a hierarchical defect detection is performed, including: The reflection intensity distribution is obtained based on multimodal optical parameter measurement, and the surface defect probability distribution is generated based on multi-band feature correlation analysis; Setting a first determination threshold, and marking abnormal areas exceeding the first determination threshold according to a surface defect probability distribution map; Conduct electromagnetic field phase response measurements in abnormal areas to obtain subsurface defect phase shift distributions related to subsurface defects in the internal threads of copper tubes. The surface defect probability distribution and sub-surface defect phase shift distribution are weightedly fused to output the determination results of defect type and depth level.

[0026] As a preferred embodiment of the above embodiment, a multimodal optical measuring device (such as a spectrometer, a laser scanner, etc.) is used to scan the surface of the internal thread of the copper tube to measure the light reflection intensity of multiple bands. Through the spectral information of different bands, comprehensive data on the surface condition can be obtained. The optical parameter measurement results are converted into a reflection intensity distribution map through a data processing algorithm to reflect the surface quality of the internal thread of the copper tube; machine learning or data analysis algorithms (such as PCA principal component analysis, cluster analysis, etc.) are used to perform correlation analysis on the multi-band reflection intensity data. This analysis can identify the correlation between different bands and then generate a surface defect probability distribution map. The surface defect probability distribution map can display the defect probability of each position and help identify flaws or anomalies on the surface; based on the surface defect probability distribution map, a first judgment threshold is set, and all areas exceeding the first judgment threshold will be marked as abnormal areas. The first judgment threshold is usually set based on experience or experimental data, and the purpose is to screen out parts with more serious surface defects and abnormal areas. The marks can be highlighted by color or other visual means; for the marked abnormal areas, electromagnetic field phase response measurement is implemented. The electromagnetic field phase response measurement uses the characteristics of the interaction between electromagnetic waves and matter to detect sub-surface defects of the internal threads of copper tubes. Electromagnetic waves (such as microwaves, radio frequencies, etc.) penetrate the surface of the copper tube, record the phase changes of the reflected or transmitted waves, and obtain the sub-surface defect phase offset distribution; for the surface defect probability distribution and the sub-surface defect phase offset distribution, a weighted fusion algorithm (such as weighted average, convolutional neural network, etc.) is used to comprehensively analyze the results of the two. The purpose of weighted fusion is to improve the accuracy of detection and obtain more comprehensive defect information, and finally output the determination results of the defect type (such as scratches, cracks, corrosion, etc.) and its depth level (such as surface defects, shallow defects, deep defects, etc.); through the processing of the above steps, the defect type and depth level of the internal threads of the copper tube are output. This information can be presented in the form of reports, charts, etc. for subsequent quality control, repair or further analysis.

[0027] Furthermore, the surface defect probability distribution is generated based on the multi-band feature correlation analysis, including: The gradient amplitude of the reflection intensity distribution in the ultraviolet band is calculated, and the high gradient change area is extracted as the potential crack feature; Perform local contrast enhancement on the reflectance intensity distribution in the visible light band to identify low reflectivity areas associated with material texture anomalies; Perform spatial convolution operation on potential crack features and low reflectivity areas to generate initial defect probability distribution; Based on the difference in spectral absorption rate in the near-infrared band, effective defect areas that meet the preset absorption characteristics are screened in the initial defect probability distribution to generate a surface defect probability distribution.

[0028] As a preferred embodiment of the above, multi-band image data of the copper tube internal thread processing surface is obtained, and the multi-band image data includes reflection images of the ultraviolet band, visible light band and near infrared band. The image acquisition can use a multispectral industrial camera to ensure that the image is aligned and corrected by spatial registration; by calculating the gradient amplitude of the ultraviolet band reflection image, the Sobel operator is used to extract the image edge to obtain the high gradient change area, which often corresponds to surface microcracks or processing marks. After extraction, a potential crack feature mask is formed; the visible light band reflection image is subjected to local contrast enhancement processing, such as using the CLAHE (Contrast-Limited Adaptive Histogram Equalization) algorithm, and after enhancement, local reflectivity analysis is performed to extract the area with significantly low reflectivity. The area with a value lower than the neighborhood mean is preliminarily determined to be a low reflectivity area related to the abnormal material texture. The potential crack features extracted from the ultraviolet band are spatially convolved with the low reflectivity areas identified in the visible light band, and a Gaussian kernel is used for smooth fusion to generate an initial defect probability distribution. The initial defect probability distribution reflects the consistency of multi-source features in spatial position and is a preliminary indicator for judging the possibility of defects. The near-infrared band image is further used to analyze the spectral absorptivity characteristics of the material. According to the specific absorption characteristics of copper and its oxides in the near-infrared region, the difference between the reflectivity and the standard absorption model is calculated, and the areas in the initial defect probability map are screened to eliminate pseudo-defect areas that do not meet the preset absorption characteristics, thereby generating a surface defect probability distribution.

[0029] Furthermore, the subsurface defect phase shift distribution related to the subsurface defect of the internal thread of the copper tube is obtained, including: Apply alternating electromagnetic field excitation to the abnormal area, and the alternating electromagnetic field excitation frequency is dynamically adjusted according to the copper tube wall thickness and the preset detection depth; Collect eddy current induction signals in abnormal areas and obtain real-time phase response data through orthogonal phase-locked amplification and demodulation; Perform spatial domain normalization on the real-time phase response data to eliminate phase distortion caused by edge effects and generate sub-surface defect phase offset distribution; Based on the experimental data of phase response of standard defects at different depths, the phase offset in the subsurface defect phase offset distribution is converted into a defect depth level parameter.

[0030] As a preferred embodiment of the above, in the internal thread processing area of the copper tube, an alternating electromagnetic field excitation is first applied to the area suspected of having sub-surface defects. In order to adapt to the different wall thicknesses of the copper tube and the preset detection depth requirements, the excitation frequency will be dynamically adjusted according to the actual wall thickness and detection depth. By adjusting the frequency, specific excitation can be performed for sub-surface defects of different depths to ensure the sensitivity and accuracy of the detection; after applying the alternating electromagnetic field excitation, the eddy current sensor array is used to collect the eddy current induction signal reflected by the inner surface of the copper tube, and the signal is demodulated by an orthogonal lock-in amplifier (LIA) to extract the amplitude and phase response data of the eddy current signal. The phase response is very sensitive to identifying the defect location and depth, so the focus is on the change of the phase signal; since the copper tube has a circular shape, the eddy current sensor array is used to collect the eddy current induction signal reflected by the inner surface of the copper tube. For curved surfaces, the response of eddy current signals in different areas may be affected by edge effects, resulting in phase distortion. Therefore, after obtaining the phase response data, spatial domain normalization processing is required, including signal smoothing, edge area correction and local mean normalization to eliminate the influence of edge effects; by conducting phase response experiments on standard defect samples (with defects of different depths) in advance, the relationship between defect depth and phase offset is obtained. During the detection process, the actual phase offset data collected will be compared with the preset standard data, and the phase offset value will be converted into the depth level parameter of the defect through a mapping relationship. The depth level parameter can be used to determine the location and severity of sub-surface defects, help process optimization or determine subsequent processing steps.

[0031] Furthermore, the surface defect probability distribution and the subsurface defect phase offset distribution are weightedly fused, including: The peak intensity of the surface defect probability distribution is used as the optical characteristic component, and the gradient change rate of the sub-surface defect phase shift distribution is extracted as the electromagnetic characteristic component. Perform dynamic weighted calculation on the optical characteristic components and the electromagnetic characteristic components to generate a comprehensive judgment value; When the temperature of the detection area exceeds the preset threshold, the weight coefficient of the electromagnetic characteristic component is increased; when the surface roughness index exceeds the critical value, the weight coefficient of the optical characteristic component is reduced; Compare the comprehensive judgment value with the preset defect classification standard. If the comprehensive judgment value is in the first interval, output the surface crack defect type. If the comprehensive judgment value is in the second interval, the sub-surface inclusion defect type is output; If the comprehensive judgment value spans multiple intervals at the same time, the manual review mechanism is triggered.

[0032] As a preferred embodiment of the above, the peak intensity of the surface defect probability distribution is extracted from the surface defect detection module as the optical feature component. The optical feature component can directly reflect the significance of the surface texture abnormality. At the same time, the gradient change rate in the phase offset distribution is extracted from the sub-surface defect detection module as the electromagnetic feature component reflecting the change of the defect contour. The optical feature component and the electromagnetic feature component represent the abnormal characteristics of the surface and internal structure respectively, and are complementary. The optical feature component and the electromagnetic feature component are input into the fusion module, and the optical feature component and the electromagnetic feature component are dynamically weighted and calculated according to the current detection environment to obtain a comprehensive judgment value. The weight coefficient is adjusted. If the temperature of the detection area exceeds the preset safety threshold, the electromagnetic feature will be enhanced to avoid thermal noise interfering with the optical feature recognition. The weight of the component; if the surface roughness index of the detection area exceeds the preset critical value and the optical imaging quality decreases, the weight of the optical feature component will be reduced, and the reliance on sub-surface data will be increased. The dynamic weighting mechanism can automatically adjust according to the temperature data and roughness index fed back by the sensor in real time without manual intervention; the obtained comprehensive judgment value will be compared with the defect classification standard to complete the automatic judgment of the defect type. When the comprehensive judgment value is in the first interval (for example, the optical feature is dominant), it is judged as a surface crack defect; when the comprehensive judgment value is in the second interval (for example, the electromagnetic feature is dominant), it is judged as a sub-surface inclusion defect; when the comprehensive judgment value falls into multiple critical intervals at the same time, the manual review mechanism will be automatically triggered, prompting the operator to review the image or scan again to ensure the reliability of the recognition result.

[0033] Further, if Figure 3 As shown in the figure, the coupling relationship is realized through a preset weight distribution matrix to realize the fusion judgment of multiple physical quantities, including: The weight distribution matrix is composed of optical weight coefficient, electromagnetic weight coefficient and deformation compensation weight coefficient. The initial value of each weight coefficient is based on the properties of the copper material and the sum is 1. The optical eigenvector, electromagnetic eigenvector and deformation eigenvector are multiplied by the corresponding weight coefficients in the weight distribution matrix respectively and then superimposed to generate a fused eigenvector; Match the fused feature vector with the preset defect feature space and output the defect type and depth level; Among them, the defect feature space is constructed by the joint distribution of optical, electromagnetic and deformation characteristics of historical defect samples.

[0034] As a preference of the above embodiment, a weight distribution matrix consisting of an optical weight coefficient, an electromagnetic weight coefficient and a deformation compensation weight coefficient is constructed, and the distribution of initial weights is preset based on the material properties of the copper material being tested, such as conductivity, light reflection characteristics, and thermal deformation sensitivity. For example, for copper materials with strong conductivity and complex surface texture, the electromagnetic weight can be appropriately increased and the optical weight can be reduced. The sum of all weight coefficients is always kept at 1 to maintain the balance of the overall judgment energy; in the feature vector extraction and fusion operation, the optical feature vector output by the surface detection module is extracted to characterize surface cracks, scratches, etc.; the electromagnetic feature vector output by the sub-surface detection module is used to characterize phase offset, eddy current response, etc.; the deformation feature vector output by the geometric measurement module or the visual reconstruction module is used to identify processing deformation or offset; the optical feature vector, The electromagnetic eigenvector and the deformation eigenvector are weighted with the corresponding weight coefficients in the weight distribution matrix respectively, and then the vectors are superimposed to generate a fused eigenvector. The fused eigenvector retains the multi-dimensional information of surface, internal and geometric changes at the same time, and has a more comprehensive characterization capability. A defect feature space is constructed based on historical defect samples. The defect feature space is jointly constructed by the optical, electromagnetic and deformation information of a large amount of real defect data, and is organized through multi-dimensional clustering, dimensionality reduction mapping, etc. During the detection process, the fused eigenvector is matched with the defect feature space for similarity, and the corresponding defect type (such as surface cracks, inclusions, delamination, etc.) and depth level (such as level I shallow defects, level II medium defects, level III deep defects) are output based on the matching results. The matching can be measured based on Euclidean distance, Mahalanobis distance or machine learning model.

[0035] Furthermore, the fused feature vector is matched with the preset defect feature space, including: Based on the weight coefficients in the weight distribution matrix, the weighted Euclidean distance between the fusion feature vector and the reference vector of each defect category in the defect feature space is calculated; According to the sample distribution density of each defect category in the defect feature space, density attenuation compensation is performed on the weighted Euclidean distance; The corrected weighted Euclidean distance is mapped to the probability distribution of defect categories through the probability conversion model, and the defect type corresponding to the maximum probability is selected as the preliminary judgment result; When the maximum probability is lower than the confidence threshold, the manual review process is triggered and the defect feature space is updated with the current fused feature vector.

[0036] As a preferred embodiment of the above, based on the fused feature vector composed of optical, electromagnetic and deformation features, the distance between it and various defect reference vectors is calculated respectively in the preset defect feature space, and the distance calculation process is based on the weighted Euclidean distance method, wherein the weight of each feature dimension is provided by the weight coefficient in the weight allocation matrix to reflect the importance of various physical information to the identification of different types of defects; in order to avoid misjudgment caused by uneven sample distribution, density attenuation compensation is introduced, specifically, the distribution density of each defect category sample in the statistical defect feature space is calculated. If the number of defect samples of a certain type is small or they are discretely distributed in the feature space, the weight of the influence of their distance on the final judgment result is reduced accordingly; on the contrary, if a certain type of defect is densely distributed in the feature space, the distance weight of this category is enhanced. Density attenuation compensation helps to improve the recognition ability of small sample categories and avoid The main type of defects excessively dominates the judgment result; based on the distance calculation after completing the density attenuation compensation, the weighted Euclidean distance between each type of defect and the target fusion feature vector is input into the preset probability conversion model. The probability conversion model can use Gaussian function, Softmax function or other probability distribution mapping methods to convert the weighted distance value into the probability value distribution of each type of defect. Finally, the defect type corresponding to the maximum probability is selected as the preliminary judgment result of the current detection target; if the maximum defect probability value obtained is lower than the set confidence threshold (for example, lower than a certain percentage), it is judged that the current judgment result does not have sufficient confidence, and the manual re-inspection process is automatically triggered, prompting the inspector to review the relevant images or data. At the same time, the current fusion feature vector and the judgment label are used as new samples and included in the defect feature space for subsequent model training and matching optimization.

[0037] Further, if Figure 4 As shown, the deformation error correction of thread surface feature data is performed in combination with dynamic compensation, including: Collect the axial temperature distribution data of the copper tube in real time and calculate the maximum temperature difference and temperature change rate; Obtain the radial deformation compensation and axial expansion compensation of the copper tube based on the maximum temperature difference and the temperature change rate; The optically measured thread tooth height is compensated based on the radial deformation compensation amount. The compensation value is determined by the product of the thermal expansion coefficient of the copper material and the radial deformation compensation amount. The optically measured thread lead angle is compensated and corrected based on the axial telescopic compensation amount, and the compensation value is determined by the linear relationship between the deformation angle conversion coefficient and the axial telescopic compensation amount.

[0038] As a preferred embodiment of the above, a temperature sensor array is used to collect axial temperature distribution data at different positions on the surface of the copper tube in real time. By calculating the maximum temperature difference and temperature change rate in the axial temperature distribution data, the temperature fluctuation of the copper tube during the processing can be obtained; according to the maximum temperature difference and temperature change rate, the deformation compensation amount of the copper tube in the radial and axial directions is calculated by using the known thermal expansion characteristics of the copper material. The radial deformation compensation amount reflects the change in the diameter of the copper tube, while the axial expansion compensation amount represents the change in the axial length of the copper tube; based on the radial deformation compensation amount, the thread tooth height obtained by optical measurement is corrected. The thermal expansion of the copper tube will cause the change in the thread tooth height, so by compensating for the radial deformation caused by thermal expansion, the thread tooth height is corrected. Deformation can accurately correct the thread tooth height measurement results, thereby ensuring that the true size of the thread is accurately reflected; according to the axial expansion compensation, the thread lead angle obtained by optical measurement is corrected. The axial expansion compensation characterizes the change in axial length of the copper tube caused by temperature change, which will affect the lead angle of the thread. By correcting this change, it can be ensured that the lead angle of the thread is consistent with the actual processing shape; after radial and axial deformation compensation, the obtained thread tooth height and thread lead angle will be more accurate. These corrected feature data will be used as input to the defect detection system, and further used for defect identification and classification. By eliminating the deformation error caused by temperature change, more accurate defect judgment results can be provided.

[0039] Embodiment 2: Based on the same inventive concept as the method for detecting secondary processing defects of copper tube internal threads in the aforementioned embodiment, the present invention further provides a system for detecting secondary processing defects of copper tube internal threads, the system comprising: A data acquisition module, which acquires thread surface feature data based on multimodal optical parameter measurement, where the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; The error correction module synchronously obtains the temperature gradient parameters during the processing and combines dynamic compensation to correct the deformation error of the thread surface feature data; The defect detection module performs hierarchical defect detection based on the pre-trained multi-dimensional defect judgment model based on the corrected thread surface feature data; The result output module outputs the defect measurement results based on the coupling relationship between multi-modal optical parameters and electromagnetic parameters. The coupling relationship is realized through a preset weight distribution matrix to realize the fusion judgment of multiple physical quantities.

[0040] The above-mentioned adjustment system in the present invention can be effectively implemented, and the technical effects that can be achieved are as described in the above-mentioned embodiments, which will not be repeated here.

[0041] More specifically, the defect detection module includes: Correlation analysis unit, which obtains reflection intensity distribution based on multimodal optical parameter measurement and generates surface defect probability distribution based on multi-band feature correlation analysis; a threshold judgment unit, which sets a first judgment threshold and marks an abnormal area exceeding the first judgment threshold according to a surface defect probability distribution map; The offset measurement unit performs electromagnetic field phase response measurement in the abnormal area to obtain the sub-surface defect phase offset distribution related to the sub-surface defect of the internal thread of the copper tube; The fusion judgment unit performs weighted fusion on the surface defect probability distribution and the sub-surface defect phase offset distribution, and outputs the judgment results of the defect type and depth level.

[0042] Similarly, the above-mentioned optimization schemes for the system can also respectively achieve the corresponding optimization effects of the method in Example 1, which will not be repeated here.

[0043] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and drawings are merely illustrative of the present application as defined herein and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, the present application is intended to include such modifications and variations as fall within the scope of the present application and its equivalents.

Claims

1. A method for detecting secondary processing defects of internal threads of copper tubes, characterized in that: The method comprises: Acquiring thread surface feature data based on multimodal optical parameter measurement, wherein the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; Synchronously acquiring temperature gradient parameters during the machining process, and performing deformation error correction on the thread surface feature data in combination with dynamic compensation; Pre-training a multi-dimensional defect determination model based on the corrected thread surface feature data and performing hierarchical defect detection based on electromagnetic field phase response measurement; The defect measurement result is output by integrating the coupling relationship between the multimodal optical parameters and the electromagnetic parameters, wherein the coupling relationship realizes multi-physical quantity fusion judgment through a preset weight distribution matrix.

2. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 1, characterized in that: Perform hierarchical defect detection, including: Obtaining a reflection intensity distribution based on the multimodal optical parameter measurement, and generating a surface defect probability distribution based on a multi-band feature correlation analysis; Setting a first determination threshold, and marking abnormal areas exceeding the first determination threshold according to the surface defect probability distribution map; Performing the electromagnetic field phase response measurement in the abnormal area to obtain a sub-surface defect phase offset distribution related to the sub-surface defect of the internal thread of the copper tube; The surface defect probability distribution and the sub-surface defect phase shift distribution are weightedly fused to output the determination result of the defect type and depth level.

3. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 2, characterized in that: Generate surface defect probability distribution based on multi-band feature correlation analysis, including: performing gradient amplitude calculation on the reflection intensity distribution in the ultraviolet band, and extracting high gradient change areas as potential crack features; performing local contrast enhancement on the reflection intensity distribution in the visible light band to identify low reflectivity regions associated with material texture anomalies; Performing a spatial convolution operation on the potential crack feature and the low reflectivity area to generate an initial defect probability distribution; Based on the difference in spectral absorptivity in the near-infrared band, effective defect areas that meet preset absorption characteristics are screened in the initial defect probability distribution to generate the surface defect probability distribution.

4. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 2, characterized in that: Obtain the subsurface defect phase shift distribution associated with the copper tube internal thread subsurface defects, including: Applying an alternating electromagnetic field excitation to the abnormal area, wherein the alternating electromagnetic field excitation frequency is dynamically adjusted according to the copper tube wall thickness and the preset detection depth; Collecting eddy current induction signals from the abnormal area, and obtaining real-time phase response data through orthogonal phase-locked amplification and demodulation; Performing spatial domain normalization processing on the real-time phase response data to eliminate phase distortion caused by edge effects and generate the sub-surface defect phase offset distribution; The phase shift amount in the subsurface defect phase shift distribution is converted into a defect depth level parameter based on the phase response experimental data of standard sample defects at different depths.

5. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 2, characterized in that: The surface defect probability distribution and the sub-surface defect phase offset distribution are weightedly fused, comprising: Taking the peak intensity of the surface defect probability distribution as the optical characteristic component, and extracting the gradient change rate of the sub-surface defect phase offset distribution as the electromagnetic characteristic component; Performing dynamic weighted calculation on the optical characteristic component and the electromagnetic characteristic component to generate a comprehensive judgment value; When the temperature of the detection area exceeds a preset threshold, the weight coefficient of the electromagnetic characteristic component is increased; when the surface roughness index exceeds a critical value, the weight coefficient of the optical characteristic component is reduced; Comparing the comprehensive judgment value with a preset defect classification standard, if the comprehensive judgment value is in the first interval, outputting the surface crack defect type; If the comprehensive judgment value is in the second interval, the sub-surface inclusion defect type is output; If the comprehensive judgment value spans multiple intervals at the same time, the manual review mechanism is triggered.

6. The method for detecting secondary processing defects of internal threads of copper tubes according to claim 1, characterized in that: The coupling relationship is implemented through a preset weight distribution matrix to achieve multi-physical quantity fusion judgment, including: The weight distribution matrix is composed of an optical weight coefficient, an electromagnetic weight coefficient, and a deformation compensation weight coefficient, wherein the initial value of each weight coefficient is allocated based on the properties of the copper material and the sum is 1; Multiplying the optical eigenvector, the electromagnetic eigenvector, and the deformation eigenvector by the corresponding weight coefficients in the weight distribution matrix respectively and then superimposing them to generate a fused eigenvector; Matching the fused feature vector with a preset defect feature space, and outputting the defect type and depth level; The defect feature space is constructed by jointly distributing the optical, electromagnetic and deformation features of historical defect samples.

7. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 6, characterized in that: Matching the fused feature vector with a preset defect feature space includes: Calculating, based on each of the weight coefficients in the weight distribution matrix, a weighted Euclidean distance between the fused feature vector and each defect category reference vector in the defect feature space; Performing density attenuation compensation on the weighted Euclidean distance according to the sample distribution density of each defect category in the defect feature space; The modified weighted Euclidean distance is mapped into a probability distribution of defect categories through a probability conversion model, and the defect type corresponding to the maximum probability is selected as the preliminary judgment result; When the maximum probability value is lower than the confidence threshold, a manual review process is triggered and the defect feature space is updated with the current fused feature vector.

8. The method for detecting secondary processing defects of internal threads of copper pipes according to claim 1, characterized in that: Combining dynamic compensation to correct deformation errors of the thread surface feature data includes: Collect the axial temperature distribution data of the copper tube in real time and calculate the maximum temperature difference and temperature change rate; Obtaining a radial deformation compensation amount and an axial expansion compensation amount of the copper tube based on the maximum temperature difference and the temperature change rate; Performing compensation correction on the optically measured thread tooth height based on the radial deformation compensation amount, wherein the compensation value is determined by the product relationship between the thermal expansion coefficient of the copper material and the radial deformation compensation amount; The optically measured thread lead angle is compensated and corrected based on the axial telescopic compensation amount, and the compensation value is determined by the linear relationship between the deformation angle conversion coefficient and the axial telescopic compensation amount.

9. A copper tube internal thread secondary processing defect detection system, characterized in that: The system comprises: A data acquisition module, which acquires thread surface feature data based on multimodal optical parameter measurement, where the multimodal optical parameters at least include polarized light suppression reflection parameters and multi-band spectral reflection parameters; The error correction module synchronously obtains the temperature gradient parameters during the processing and combines dynamic compensation to correct the deformation error of the thread surface feature data; The defect detection module performs hierarchical defect detection based on the pre-trained multi-dimensional defect judgment model based on the corrected thread surface feature data; The result output module outputs the defect measurement results based on the coupling relationship between multi-modal optical parameters and electromagnetic parameters. The coupling relationship is realized through a preset weight distribution matrix to realize the fusion judgment of multiple physical quantities.

10. The copper tube internal thread secondary processing defect detection system according to claim 9, characterized in that: The defect detection module includes: Correlation analysis unit, which obtains reflection intensity distribution based on multimodal optical parameter measurement and generates surface defect probability distribution based on multi-band feature correlation analysis; a threshold judgment unit, which sets a first judgment threshold and marks an abnormal area exceeding the first judgment threshold according to a surface defect probability distribution map; The offset measurement unit performs electromagnetic field phase response measurement in the abnormal area to obtain the sub-surface defect phase offset distribution related to the sub-surface defect of the internal thread of the copper tube; The fusion judgment unit performs weighted fusion on the surface defect probability distribution and the sub-surface defect phase offset distribution, and outputs the judgment results of the defect type and depth level.

Images