Molding defect detection method for silicon carbide thin-walled tube preparation process

Abstract

The invention relates to the field of image processing, in particular to a forming defect detection method for a silicon carbide thin-walled tube preparation process. Comprising the steps of obtaining a surface image, and recording a working condition identifier and spatial position information; dividing the surface image into at least two local sub-regions, screening abnormal sub-regions, obtaining a preliminary abnormal quantity, and clustering the abnormal sub-regions into defect candidate regions; obtaining an edge average gradient value, and calculating development fluctuation complexity; calculating development fluctuation working condition response dispersion; obtaining the number of edge lines of the defect candidate area under different working conditions, and calculating the consistent attenuation of development composition; obtaining a texture gradient main direction and development response intensity of the defect candidate area under different working conditions, and calculating a development structure response consistency index; and generating a constraint factor. The method can improve the detection precision in the preparation process of the silicon carbide thin-walled tube.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more specifically to a method for detecting forming defects in the fabrication process of silicon carbide thin-walled tubes. Background Technology

[0002] Silicon carbide is a typical third-generation wide-bandgap ceramic material, possessing properties such as high temperature resistance, corrosion resistance, high thermal conductivity, high mechanical strength, and excellent chemical stability. Therefore, it has wide applications in high-temperature heat exchange tubes, semiconductor process piping, nuclear energy equipment, aerospace propulsion systems, and high-end chemical reaction equipment. As high-end equipment continues to develop towards lightweight, miniaturization, and high reliability, silicon carbide thin-walled tubes, while ensuring structural strength, can significantly reduce material usage and improve heat transfer efficiency, thus gradually becoming key functional components in the aforementioned fields.

[0003] However, in the process of detecting molding defects in silicon carbide thin-walled tubes, commonly used visual inspection methods generally face the problem of unstable defect characterization. Specifically, when there are real molding defects in the thin-walled tube, due to the combined effects of uneven micro-density distribution on the tube surface, differences in molding flow orientation, and local thickness variations, these defects generated during the molding stage are often accompanied by changes in the internal stress gradient and surface microstructure of the material, resulting in them not exhibiting fixed grayscale, texture, or edge features under visible light imaging conditions.

[0004] Most existing detection methods are based on static feature assumptions or single-scale image representation, failing to establish the intrinsic relationship between the true physical state of the forming defect and the evolution of its image features under different conditions. Therefore, the same defect often exhibits significantly different image features in different imaging locations or different forming batches, leading to misjudgments, missed judgments, or fluctuations in detection results due to process conditions. Ultimately, this results in a lack of reliability in terms of spatial distribution and batch consistency. Summary of the Invention

[0005] This invention provides a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes, in order to solve existing problems.

[0006] The present invention provides a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes, which adopts the following technical solution: One embodiment of the present invention provides a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes, the method comprising the following steps: Acquire surface images of silicon carbide thin-walled tubes under different operating conditions, and record the operating condition identifier and spatial location information corresponding to each image; Based on spatial location information, the surface image is divided into at least two local sub-regions, abnormal sub-regions are filtered out, and the preliminary abnormality of each abnormal sub-region is obtained. The abnormal sub-regions are then clustered into at least one defect candidate region. Edge detection is performed on each defect candidate region to obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and the imaging fluctuation complexity of the defect candidate region is calculated based on the average gradient value of the edges. Obtain the imaging fluctuation complexity of any candidate defect region under different operating conditions, and calculate the imaging fluctuation operating condition response dispersion of the candidate defect region based on the difference in imaging fluctuation complexity under different operating conditions. The number of edge lines in the candidate defect region under different working conditions is obtained. Based on the dispersion of the imaging fluctuation response and the change of the number of edge lines with the working conditions, the imaging composition uniform attenuation of the candidate defect region is calculated. Obtain the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different working conditions, and calculate the imaging structure response consistency index of the candidate defect region. Constraint factors are generated based on the consistent attenuation of the imaging structure and the consistency index of the imaging structure response. These constraint factors are then used to modulate the initial anomaly of the defect candidate region, and the true defect detection value is output.

[0007] Optionally, based on spatial location information, the surface image is divided into at least two local sub-regions, abnormal sub-regions are filtered out, and a preliminary anomalous quantity of each abnormal sub-region is obtained, specifically including: Based on spatial location information, the surface image is divided into grids in the axial and circumferential directions to obtain at least two local sub-regions. The spatial location information includes the axial coordinates and circumferential angle information of the thin-walled tube. For each local sub-region, its image features are extracted, wherein the image features include at least one of the following: average gray value, gray distribution variance, and texture roughness; The image features of each local sub-region are compared with a preset normal feature threshold. If the comparison result exceeds the preset normal feature threshold, the local sub-region is determined to be an abnormal sub-region. For each abnormal sub-region, the preliminary abnormality amount of the abnormal sub-region is calculated based on the degree of deviation between its image features and the preset normal feature threshold.

[0008] Optionally, the abnormal sub-regions are clustered into at least one defect candidate region, specifically including: Establish spatial location indexes for all abnormal sub-regions; Based on the preset spatial proximity determination rules and spatial location index, it is determined whether the abnormal sub-regions are adjacent. If the axial distance and circumferential angle difference between two abnormal sub-regions are both less than the preset threshold, then the two abnormal sub-regions are determined to be spatially adjacent. Spatially adjacent anomalous sub-regions are merged into the same set to form a defect candidate region; A single anomalous sub-region that has no spatial adjacency relationship is identified as a defect candidate region.

[0009] Optionally, edge detection is performed on each defect candidate region to obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and the imaging fluctuation complexity of the defect candidate region is calculated based on the average gradient value of the edges, specifically including: Edge detection is performed on each abnormal sub-region within the defect candidate region to obtain edge pixels; Calculate the average gradient magnitude of all edge pixels in each abnormal sub-region, and use it as the average edge gradient value of that abnormal sub-region. Arrange all abnormal sub-regions within the defect candidate region according to their spatial order to obtain the sorting result; Calculate the absolute value of the difference between the marginal average gradient values ​​of adjacent abnormal sub-regions in each pair of sorting results within the defect candidate region, and take the first average value of the absolute values ​​of the differences of all adjacent pairs. Calculate the absolute value of the difference between the average edge gradient values ​​of all pairs of anomalous sub-regions within the defect candidate region, and take the second average value of all the absolute values ​​of the differences. Multiplying the first average value by the second average value yields the imaging fluctuation complexity of the defect candidate region.

[0010] Optionally, the imaging fluctuation complexity of any candidate defect region under different operating conditions is obtained. Based on the difference in imaging fluctuation complexity under different operating conditions, the imaging fluctuation operating condition response dispersion of the candidate defect region is calculated, specifically including: Obtain the imaging fluctuation complexity of the same defect candidate region under at least three different working conditions; Calculate the absolute value of the difference in imaging fluctuation complexity between the two consecutive operating conditions; The absolute values ​​of the differences in imaging fluctuation complexity for all continuous operating conditions are averaged, and this average value is used as the imaging fluctuation response dispersion of the defect candidate region.

[0011] Optionally, the number of edge lines in the candidate defect region under different operating conditions is obtained. Based on the variation of the imaging fluctuation response dispersion and the number of edge lines with the operating conditions, the imaging composition uniform attenuation of the candidate defect region is calculated, specifically including: Obtain the number of edge lines for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference in the number of edge lines corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the number of edge lines for all continuous working conditions is used to obtain the average value of the change in the number of edge lines. Multiplying the dispersion of the imaging fluctuation response of the defect candidate region by the mean of the change in the number of edge lines yields the uniform attenuation of the imaging structure of the defect candidate region.

[0012] Optionally, the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different operating conditions are obtained, and the imaging structure response consistency index of the candidate defect region is calculated, specifically including: Obtain the principal direction angle values ​​of the texture gradient for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the principal direction angle values ​​of the texture gradient corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the principal direction angle of the texture gradient for all continuous operating conditions is used to obtain the average value of the texture direction change. Obtain the imaging response intensity values ​​of the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the image response intensity values ​​corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in image response intensity for all continuous operating conditions is taken to obtain the average value of the image response change. The reciprocal of the mean variation of texture direction is multiplied by the reciprocal of the mean variation of image response to obtain the image structure response consistency index of the defect candidate region.

[0013] Optionally, a constraint factor is generated based on the image structure uniformity attenuation and the image structure response uniformity index. This constraint factor is then used to modulate the initial anomaly value of the defect candidate region, and the true defect detection value is output. Specifically, this includes: Based on the image composition uniform attenuation and the image structure response uniformity index, a constraint factor is calculated. The value of the constraint factor is positively correlated with the image structure response uniformity index and negatively correlated with the image composition uniform attenuation. The modulated anomaly is obtained by multiplying the initial anomaly value of the defect candidate region by the constraint factor. The modulated anomaly is output as the true defect detection value of the defect candidate region.

[0014] Optionally, the method further includes: Based on the spatial location information of each defect candidate region, defect candidate regions that are spatially adjacent and whose actual defect detection values ​​all meet the preset conditions are aggregated into the same formed defect region. Based on the axial position range and circumferential angle range covered by the molding defect area after polymerization, the spatial position, coverage area and contour boundary of the molding defect on the surface of the thin-walled tube are determined. Based on the distribution of the actual defect detection values ​​of each defect candidate region within the aggregated forming defect region, the imaging response intensity and spatial stability of the forming defect are evaluated. Based on the assessment results, the severity level of the molding defect is classified or risk is identified, and a structured defect report is generated. The structured defect report includes at least one of the following: spatial location, outline boundary, and level information.

[0015] This invention proposes a molding defect detection system for the silicon carbide thin-walled tube manufacturing process, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of a molding defect detection method for the silicon carbide thin-walled tube manufacturing process.

[0016] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, a complete analysis chain, from local anomaly screening and spatial imaging fluctuation analysis to multi-condition response behavior evaluation, is constructed to achieve dynamic characterization of the true physical state of defects. This method not only quantifies the spatial imaging fluctuation complexity of defects under a single condition, but also establishes a correlation model between defect image features and the evolution of their internal physical state by analyzing the response dispersion of imaging fluctuations to changes in conditions, the attenuation behavior of imaging composition with changing conditions, and the synchronous consistency between imaging and structural changes. Finally, by fusing the above multi-dimensional indicators to generate adaptive constraint factors, the preliminary anomaly detection results are intelligently modulated, thereby significantly improving the ability to identify and detect true defects, effectively suppressing misjudgments and missed judgments caused by process fluctuations and surface micro-inhomogeneities, and providing a reliable technical basis for the molding quality control, process optimization, and reliability assessment of silicon carbide thin-walled tubes. Attached Figure Description

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

[0018] Figure 1 This is a flowchart of a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes, provided in one embodiment of the present invention. Figure 2This is a structural diagram of a molding defect detection system for the fabrication process of silicon carbide thin-walled tubes, provided in one embodiment of the present invention. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a molding defect detection method for silicon carbide thin-walled tube manufacturing process proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following describes in detail, with reference to the accompanying drawings, a specific scheme for a molding defect detection method in the silicon carbide thin-walled tube manufacturing process provided by the present invention.

[0022] This invention provides a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a flowchart of a method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to an embodiment of the present invention. The method includes the following steps: S101. Obtain surface images of silicon carbide thin-walled tubes under different operating conditions, and record the operating condition identifier and spatial location information corresponding to each image.

[0023] For example, after the silicon carbide thin-walled tube completes the forming process and enters the defect detection station, the forming condition identification information corresponding to the tube is first obtained, including the forming batch, process parameter combination, etc., and the identification information is uniquely associated with this thin-walled tube to distinguish different production batches and process states.

[0024] Subsequently, as the thin-walled tube moves forward at a constant speed along its axial direction while simultaneously rotating around its axis (i.e., moving circumferentially), one or more industrial cameras positioned in the inspection area continuously acquire images of its outer surface. The camera sampling frequency is matched to the tube's speed to ensure that the acquired image sequence completely covers every axial position from the tube end to the tube tail, as well as every circumferential angle around the tube, thus obtaining raw image data of the entire outer surface of the thin-walled tube.

[0025] Simultaneously with image acquisition, the system uses an integrated position encoder to acquire and record the precise spatial position information of the thin-walled tube at the moment of each image acquisition, including axial displacement coordinates and circumferential rotation angle. This spatial position information is timestamped and associated with the original surface image data acquired at the corresponding moment, forming an imaging dataset where each image has a clear spatial coordinate identifier.

[0026] The above acquisition and association process is repeated on the same thin-walled tube until its entire surface area is imaged, thereby obtaining multiple sets of image data reflecting the surface morphology and texture of the tube in a single molding state at different spatial locations.

[0027] For silicon carbide thin-walled tubes produced in different batches, the same imaging layout, motion control and acquisition timing process are strictly followed to obtain the corresponding full-surface image data, and the image data is always stored in a structured association with the batch identification information to which it belongs.

[0028] S102. Based on spatial location information, the surface image is divided into at least two local sub-regions, abnormal sub-regions are filtered out, and the preliminary abnormality of each abnormal sub-region is obtained. The abnormal sub-regions are then clustered into at least one defect candidate region.

[0029] In this embodiment, based on spatial location information, the surface image is divided into at least two local sub-regions, abnormal sub-regions are filtered out, and a preliminary abnormality value is obtained for each abnormal sub-region. Specifically, this includes: Based on spatial location information, the surface image is divided into grids in the axial and circumferential directions to obtain at least two local sub-regions. The spatial location information includes the axial coordinates and circumferential angle information of the thin-walled tube. For each local sub-region, its image features are extracted, wherein the image features include at least one of the following: average gray value, gray distribution variance, and texture roughness; The image features of each local sub-region are compared with a preset normal feature threshold. If the comparison result exceeds the preset normal feature threshold, the local sub-region is determined to be an abnormal sub-region. For each abnormal sub-region, the preliminary abnormality amount of the abnormal sub-region is calculated based on the degree of deviation between its image features and the preset normal feature threshold.

[0030] Clustering abnormal sub-regions into at least one defect candidate region specifically includes: Establish spatial location indexes for all abnormal sub-regions; Based on the preset spatial proximity determination rules and spatial location index, it is determined whether the abnormal sub-regions are adjacent. If the axial distance and circumferential angle difference between two abnormal sub-regions are both less than the preset threshold, then the two abnormal sub-regions are determined to be spatially adjacent. Spatially adjacent anomalous sub-regions are merged into the same set to form a defect candidate region; A single anomalous sub-region that has no spatial adjacency relationship is identified as a defect candidate region.

[0031] For example, after obtaining the original surface image with spatial location markers, the image is first processed by an image segmentation algorithm (such as threshold-based segmentation or edge detection) to accurately extract the area occupied by the silicon carbide thin-walled tube in the image, remove irrelevant areas such as the device background and support structure, and retain only the effective imaging area corresponding to the outer surface of the thin-walled tube.

[0032] Within the effective imaging area, based on pre-set axial segment lengths (e.g., every 5 millimeters per segment) and circumferential angular intervals (e.g., every 10 degrees per sector), the continuous thin-walled tube outer surface image is divided into multiple local sub-regions of uniform size and fixed position on a two-dimensional unfolded plane. Each local sub-region physically corresponds to a specific axial interval and circumferential angular range on the tube surface.

[0033] Optionally, during the meshing of the thin-walled tube surface based on the axial segment length and circumferential angle interval, the mesh size of local sub-regions can be adaptively set according to actual inspection needs, the physical scale characteristics of defects, and the system's processing capabilities. Typically, the axial segment length can be adjusted from 1 mm to 20 mm, and the circumferential angle interval can be selected from 5 degrees to 30 degrees. For example, in high-precision inspection scenarios requiring the identification of micro-cracks, pores, and other minute defects, a smaller mesh size (e.g., 1 to 2 mm axially, 5 to 10 degrees circumferentially) can be used to capture more detailed local features; while in rapid screening or initial production line inspection of macroscopic defects such as scratches and dents, the mesh size can be appropriately increased (e.g., 10 to 20 mm axially, 20 to 30 degrees circumferentially) to improve image processing and analysis efficiency. The system supports presetting and fixing the mesh parameters according to specific process requirements and quality objectives during the initialization phase of the inspection task, and maintains consistent meshing rules throughout the entire batch inspection process, thereby ensuring the comparability of data and the consistency of analysis results between different tubes and different imaging positions.

[0034] For each segmented local sub-region, its image feature indices are calculated, such as average gray level, gray level variance, and Local Binary Pattern (LBP) texture descriptor. By comparing the image features of this sub-region with the feature mean of its neighboring sub-regions, or with a preset normal feature threshold, sub-regions with significant deviations in brightness distribution, structural continuity, or texture pattern are identified. This identification process can be implemented using known techniques such as statistical significance tests (e.g., Z-score tests) or machine learning classifiers (e.g., single-class support vector machines).

[0035] Optionally, a preset normal feature threshold is established based on a reference standard derived from statistical analysis of numerous defect-free silicon carbide thin-walled tube surface images. The threshold setting process includes: acquiring surface images of multiple batches of qualified thin-walled tubes under stable production conditions, and extracting image features (such as average grayscale, grayscale distribution variance, texture roughness, etc.) for each local sub-region using the same grid division method; subsequently, statistically analyzing the distribution of each feature value and calculating its mean and standard deviation; finally, setting the normal feature threshold to a range of "mean ± N times standard deviation," where N can be set according to the stringency of quality control, typically between 2 and 3, to cover approximately 95% to 99.7% of normal feature fluctuations. In actual testing, this threshold can also be fine-tuned based on specific material batches, surface treatment processes, or imaging conditions to ensure effective differentiation between normal surfaces and abnormal areas under different production scenarios.

[0036] Each local sub-region identified as having an imaging anomaly is marked as an "abnormal sub-region". Based on the degree of deviation between the image features of the abnormal sub-region and its corresponding normal reference features, a quantified "preliminary anomaly quantity" is calculated. This value is used to preliminarily characterize the probability of a defect in the region.

[0037] Subsequently, based on the spatial location information (axial coordinates and circumferential angles) of all abnormal sub-regions, and according to preset spatial proximity determination rules (e.g., setting axial distance thresholds and circumferential angle difference thresholds), multiple abnormal sub-regions that are spatially close to each other are merged into the same set, forming a spatially continuous "defect candidate region". For a single abnormal sub-region that is spatially isolated and does not meet the condition of being adjacent to any other abnormal sub-region, it is determined as a defect candidate region on its own.

[0038] S103. Perform edge detection on each defect candidate region, obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and calculate the imaging fluctuation complexity of the defect candidate region based on the average gradient value of the edges.

[0039] In this embodiment, edge detection is performed on each defect candidate region to obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region. The imaging fluctuation complexity of the defect candidate region is then calculated based on the average gradient value, specifically including: Edge detection is performed on each abnormal sub-region within the defect candidate region to obtain edge pixels; Calculate the average gradient magnitude of all edge pixels in each abnormal sub-region, and use it as the average edge gradient value of that abnormal sub-region. Arrange all abnormal sub-regions within the defect candidate region according to their spatial order to obtain the sorting result; Calculate the absolute value of the difference between the marginal average gradient values ​​of adjacent abnormal sub-regions in each pair of sorting results within the defect candidate region, and take the first average value of the absolute values ​​of the differences of all adjacent pairs. Calculate the absolute value of the difference between the average edge gradient values ​​of all pairs of anomalous sub-regions within the defect candidate region, and take the second average value of all the absolute values ​​of the differences. Multiplying the first average value by the second average value yields the imaging fluctuation complexity of the defect candidate region.

[0040] For example, after completing the screening of abnormal sub-regions and obtaining their preliminary abnormality values, for each defect candidate region, it is necessary to further analyze its internal spatial imaging fluctuation characteristics.

[0041] First, edge detection is performed on each anomalous sub-region within the defect candidate region to extract the edge pixels of that region. The gradient magnitudes of all edge pixels within that region are calculated, and the arithmetic mean of these gradient magnitudes is taken. This value is denoted as the average edge gradient value of the anomalous sub-region, T. This value quantifies the intensity of surface structure undulations in that local region.

[0042] Next, the imaging fluctuation complexity of the candidate defect region is calculated. This metric quantifies the spatial inconsistency and dispersion of a candidate region, and its calculation is based on the average edge gradient value T of all anomalous sub-regions within the region. The specific calculation formula is as follows: ; in, and These represent the average gradient values ​​of the i-th and (i+1)-th edges in the sorting results, respectively, where n represents the number of outlier sub-regions. and Let e ​​and o represent the marginal average gradient values ​​of any two anomalous subregions e and o, respectively. It is a very small positive constant used to ensure that the calculation is numerically stable.

[0043] In the formula, The first average value reflects the degree of drastic change in the intensity of the imaging structure within the defect candidate region between adjacent spatial locations. The larger the first average value, the more unstable and volatile the imaging in that region is spatially. The second average value reflects the overall dispersion of the imaging structure intensity among all abnormal sub-regions within the region. The larger the second average value, the greater the difference in imaging performance at different locations within the region, and the worse the consistency.

[0044] Modeling by a product means that the imaging variability complexity f only increases significantly when the defect candidate region simultaneously exhibits high local spatial variability (large first average value) and high overall imaging inconsistency (second average value). This design enhances the selectivity for identifying real defects that exhibit spatially “jumpy and scattered” imaging characteristics, which may be caused by complex physical states.

[0045] S104. Obtain the imaging fluctuation complexity of any candidate defect region under different operating conditions. Based on the difference in imaging fluctuation complexity under different operating conditions, calculate the imaging fluctuation operating condition response dispersion of the candidate defect region.

[0046] In this embodiment, the imaging fluctuation complexity of any candidate defect region under different operating conditions is obtained. Based on the difference in imaging fluctuation complexity under different operating conditions, the imaging fluctuation operating condition response dispersion of the candidate defect region is calculated, specifically including: Obtain the imaging fluctuation complexity of the same defect candidate region under at least three different working conditions; Calculate the absolute value of the difference in imaging fluctuation complexity between the two consecutive operating conditions; The absolute values ​​of the differences in imaging fluctuation complexity for all continuous operating conditions are averaged, and this average value is used as the imaging fluctuation response dispersion of the defect candidate region.

[0047] For example, in order to effectively distinguish between imaging changes caused by real defects and imaging differences caused by normal fluctuations in the production process, it is necessary to compare and analyze the behavior of the same defect candidate area under different molding conditions, that is, to calculate its imaging fluctuation response dispersion.

[0048] The forming conditions refer to the combination of all controllable process parameters that affect the final internal structure and surface state of silicon carbide thin-walled tubes during the manufacturing process, such as sintering temperature, pressure, holding time, and raw material ratio. Each defect candidate region can be uniquely traced and located in images of different batches of thin-walled tubes through its associated spatial location information (axial coordinates and circumferential angles).

[0049] The calculation of the dispersion of the imaging fluctuation response is based on the imaging fluctuation complexity f obtained by calculating the candidate defect region under various forming conditions. The specific steps are as follows: For the same spatial location, the imaging fluctuation complexity of the candidate defect region was calculated using the aforementioned method under at least three different molding conditions, resulting in a set of values: , where m is greater than 3.

[0050] Calculate the absolute value of the difference in imaging fluctuation complexity between each pair of consecutive operating conditions in the sequence, i.e., calculate... ,in, This difference reflects the amount of change in the imaging instability of the defect candidate region when the production process conditions are adjusted from one state to the next adjacent state. The larger the value, the more sensitive the imaging features of the region are to subtle changes in process conditions, and the more prone they are to drift.

[0051] Calculate the arithmetic mean of the absolute values ​​of the complexity differences between all consecutive operating conditions. Where m represents the number of molding processes. It is a very small positive constant used to ensure that the calculation is numerically stable.

[0052] Image fluctuation response dispersion This characterizes the overall sensitivity of the imaging instability (measured by f) of the candidate defect region to changes in molding conditions. The higher the dispersion value of the imaging fluctuation response, the more easily the imaging characteristics of this region can undergo systematic changes with adjustments to production process parameters, and the worse its consistency across batches.

[0053] S105. Obtain the number of edge lines in the candidate defect region under different working conditions. Based on the variation of the imaging fluctuation response dispersion and the number of edge lines with the working conditions, calculate the consistent attenuation of the imaging composition of the candidate defect region.

[0054] In this embodiment, the number of edge lines in the candidate defect region under different operating conditions is obtained. Based on the dispersion of the imaging fluctuation response and the change of the number of edge lines with the operating conditions, the uniform attenuation of the imaging structure of the candidate defect region is calculated, specifically including: Obtain the number of edge lines for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference in the number of edge lines corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the number of edge lines for all continuous working conditions is used to obtain the average value of the change in the number of edge lines. Multiplying the dispersion of the imaging fluctuation response of the defect candidate region by the mean of the change in the number of edge lines yields the uniform attenuation of the imaging structure of the defect candidate region.

[0055] For example, to more accurately identify the pattern of molding defect characterization changing with operating conditions, it is necessary to further analyze the evolution of the imaging composition consistency of defect candidate regions with process conditions, i.e., to calculate its imaging composition consistency attenuation. The calculation of this index integrates the aforementioned information on the dispersion of imaging fluctuation response under operating conditions and the changes in the internal edge structure organization of the region with operating conditions. The specific process is as follows: For the same defect candidate region, under m different forming conditions corresponding to the calculation of imaging fluctuation complexity, the number of edge lines in all abnormal sub-regions within that region is counted. The number of edge lines is measured by the number of connected contours in the binarized edge image of that region, denoted as g. Thus, a set of edge line count values ​​corresponding to the forming condition sequence can be obtained: , where m is greater than 3.

[0056] Calculate the absolute value of the difference between the corresponding values ​​of each pair of consecutive operating conditions in the sequence of edge line quantities, i.e. ,in, This difference reflects the degree to which the microstructure organization of the imaging area (manifested as the continuity, dispersion, or aggregation of edges) changes when operating conditions change. The smaller the value, the more stable the edge structure, and the better the consistency of the imaging composition under different working conditions; conversely, the larger the value, the more significant the change in the imaging composition, such as the edge structure changing from continuous to broken, or from dispersed to aggregated.

[0057] Calculate the arithmetic mean of the absolute values ​​of the differences in the number of edge lines between all consecutive operating conditions, i.e., calculate... This average value quantifies the overall activity of the edge structure of the defect candidate region as process conditions change.

[0058] The dispersion of the response under imaging fluctuation conditions Multiplying this by the mean of the edge composition changes yields the uniform attenuation of the image composition in that region. ,Right now Where m represents the number of molding processes, It is a very small positive constant used to ensure that the calculation is numerically stable.

[0059] Image formation consistent attenuation It is a coupling index, and its physical meaning is as follows: when When the value is high (i.e., the area imaging is sensitive to changes in operating conditions) and the mean value of edge composition changes is also large, The value will increase significantly. This indicates that not only is the overall imaging instability of this region prone to drift with process fluctuations, but the internal edge structure organization also undergoes significant and systematic changes simultaneously. This synergistic change pattern strongly suggests that the observed imaging differences are not accidental imaging noise or local perturbations, but more likely a systematic migration or degradation of the defect characterization itself induced by changes in molding conditions.

[0060] therefore, The higher the value, the more severe the decay of the imaging consistency of the defect candidate region with process conditions, and the greater the possibility that it is an "unstable characterization" that is significantly affected by process fluctuations.

[0061] S106. Obtain the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different working conditions, and calculate the imaging structure response consistency index of the candidate defect region.

[0062] In this embodiment, the principal direction of the texture gradient and the imaging response intensity of the defect candidate region under different operating conditions are obtained, and the imaging structure response consistency index of the defect candidate region is calculated, specifically including: Obtain the principal direction angle values ​​of the texture gradient for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the principal direction angle values ​​of the texture gradient corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the principal direction angle of the texture gradient for all continuous operating conditions is used to obtain the average value of the texture direction change. Obtain the imaging response intensity values ​​of the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the image response intensity values ​​corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in image response intensity for all continuous operating conditions is taken to obtain the average value of the image response change. The reciprocal of the mean variation of texture direction is multiplied by the reciprocal of the mean variation of image response to obtain the image structure response consistency index of the defect candidate region.

[0063] For example, to assess whether the imaging changes in a defect candidate region maintain physical consistency with the state changes of its internal material structure, it is necessary to calculate the imaging structure response consistency index. The specific steps are as follows: For the same candidate defect region, under m different molding conditions, directional analysis (e.g., using gradient orientation histogram or structure tensor method) is performed on the image of the region to extract its principal texture gradient direction. The angle value of this principal direction is normalized to a fixed range (e.g., 0°-180°) and denoted as θ. This yields the sequence of principal texture directions: , where m is greater than 3.

[0064] Under the same set of operating conditions, the average gray value of the image in that region is calculated as the display response intensity of that region under the current operating conditions, representing its overall brightness or contrast level. A display response intensity sequence is obtained.

[0065] For each pair of continuous operating conditions, calculate the rate of change of the operating conditions themselves (e.g., the normalized scalar of the changes in process parameters such as temperature change rate and pressure change rate), denoted as b.

[0066] Correspondingly, the rate of change of the imaging response intensity of the candidate defect region between these two consecutive operating conditions is calculated and denoted as u.

[0067] Calculate the absolute value of the difference between the two. The smaller this value, the more synchronized the changes in the imaging intensity of the region are with the changes in external process conditions.

[0068] Calculations for all continuous operating condition changes And calculate their average value, denoted as . This value comprehensively reflects the overall level of synchronization between the regional imaging response and the changing operating conditions throughout all process adjustments.

[0069] Calculate the absolute value of the difference between corresponding values ​​of two consecutive working conditions in the main direction sequence of the texture, i.e. Calculate the absolute value of the difference in principal texture directions across all consecutive operating conditions. And calculate their average value to obtain... This value reflects the stability of the dominant orientation of the texture as it changes with operating conditions. The larger the value, the more consistent the dominant orientation of the texture structure is under different operating conditions, and the more continuous the change is.

[0070] Multiplying the average value reflecting the synchronicity of the imaging response by the average value reflecting the stability of the texture direction yields the imaging structure response consistency index of the defect candidate region. ,Right now: ; Image structure response consistency index The physical meaning is as follows: The higher the index value, the more it indicates that the region meets two conditions simultaneously: first, the overall brightness or contrast of its image can follow the rhythm of changes in external process conditions well (high synchronicity); second, the main arrangement direction of its internal texture structure remains stable or changes continuously under different process conditions (stable structural orientation).

[0071] This pattern of "synchronous response" and "structural stability" suggests that the imaging changes in this region are not random noise or disordered fluctuations, but may be intrinsically related to changes in the physical structural state of its internal materials, exhibiting a structure-driven and regular response behavior.

[0072] therefore, The higher the value, the more it supports the idea that the changes in the candidate defect region are driven by its true physical structure state, thus enhancing the credibility of classifying it as a real forming defect rather than an accidental artifact.

[0073] S107. Generate a constraint factor based on the consistent attenuation of the imaging structure and the consistent index of the imaging structure response. Use this constraint factor to modulate the preliminary anomaly of the defect candidate region and output the true defect detection value.

[0074] In this embodiment, a constraint factor is generated based on the image structure uniform attenuation and the image structure response uniformity index. This constraint factor is then used to modulate the initial anomaly value of the defect candidate region, and the true defect detection value is output. Specifically, this includes: Based on the image composition uniform attenuation and the image structure response uniformity index, a constraint factor is calculated. The value of the constraint factor is positively correlated with the image structure response uniformity index and negatively correlated with the image composition uniform attenuation. The modulated anomaly is obtained by multiplying the initial anomaly value of the defect candidate region by the constraint factor. The modulated anomaly is output as the true defect detection value of the defect candidate region.

[0075] The method also includes: Based on the spatial location information of each defect candidate region, defect candidate regions that are spatially adjacent and whose actual defect detection values ​​all meet the preset conditions are aggregated into the same formed defect region. Based on the axial position range and circumferential angle range covered by the molding defect area after polymerization, the spatial position, coverage area and contour boundary of the molding defect on the surface of the thin-walled tube are determined. Based on the distribution of the actual defect detection values ​​of each defect candidate region within the aggregated forming defect region, the imaging response intensity and spatial stability of the forming defect are evaluated. Based on the assessment results, the severity level of the molding defect is classified or risk is identified, and a structured defect report is generated. The structured defect report includes at least one of the following: spatial location, outline boundary, and level information.

[0076] For example, in this embodiment, it is mainly used to generate constraint factors. It is used to intelligently modulate the preliminary detection results, thereby strengthening the real defect signals with physical consistency in the final judgment and suppressing the false abnormal signals caused by process fluctuations or accidental interference.

[0077] constraint factors The calculation incorporates the aforementioned imaging composition and uniform attenuation. and the consistency index of imaging structure response The specific calculation method can be as follows: ; in, This characterizes the degree to which the imaging composition of the candidate defect region systematically degrades with process conditions. The larger the value, the more easily the imaging characteristics of the area are disturbed or even destroyed by fluctuations in the production process, and the worse its own stability. It characterizes the degree to which the imaging changes in this region maintain physical consistency with the changes in the internal structure of the material. The larger the value, the more likely the observed imaging changes are to originate from real changes in the physical state of the material rather than random noise, and therefore the higher the reliability of the information it reflects. This represents the normalization function, which maps the calculation results to a standardized range (e.g., [0,1]) to facilitate subsequent uniform weighting.

[0078] constraint factors The calculation logic is as follows: constraint factor The value and Positive correlation with Negative correlation. This means that only when the defect candidate region simultaneously satisfies the condition that the imaging changes have physical consistency (high) Furthermore, the imaging structure is not easily damaged by process fluctuations (low) A higher constraint factor value will only be assigned when both of these conditions are met. Conversely, if a region is overly sensitive to process fluctuations (high... ) or its changes lack physical laws (low) If ), then its constraint factor value will be reduced.

[0079] After obtaining the constraint factor, the preliminary detection results are modulated: The initial anomaly value U obtained for each defect candidate region in the initial screening step, and its corresponding constraint factor are used to... Multiply by the product to obtain the true defect detection value D.

[0080] Through this modulation: for real defect signals with high physical consistency and stability (high... ,Low Therefore, high The initial abnormality U will be significantly enhanced or preserved.

[0081] For unstable artifacts mainly caused by process fluctuations (high) ) or random noise lacking physical laws (low) Therefore, low ), and its initial abnormality U will be greatly suppressed.

[0082] Ultimately, the output is the actual defect detection value D after physical and operational robustness calibration. This value more accurately reflects the probability and severity of actual forming defects at each location.

[0083] After obtaining the actual defect detection values ​​D for all defect candidate regions, the following structured processing is performed: Defect region aggregation: Based on the spatial location information of each region, multiple spatially adjacent candidate defect regions whose D values ​​all exceed a set threshold are merged into a complete formed defect region.

[0084] Defect feature extraction: Based on the boundary of the aggregation region, determine the precise spatial location, coverage area (axial length and circumferential width) and contour shape of the defect on the surface of the thin-walled tube.

[0085] Comprehensive Defect Assessment: Based on the D-value distribution of all points within the aggregation area, calculate the average intensity, uniformity, and other indicators of the defect, and evaluate its imaging response intensity and spatial stability.

[0086] Defect Classification and Reporting: Based on the assessment results, defects are classified into severity levels (e.g., minor, moderate, severe) or risk labels. A structured defect report containing information such as defect location, size, shape, and severity level is generated to guide quality assessment, process optimization, or historical tracing.

[0087] In summary, this invention achieves robust detection of defects in silicon carbide thin-walled tube forming by constructing a complete analysis chain from multi-condition data acquisition and spatial local anomaly detection to cross-condition behavior pattern analysis. The method first locates potential defect regions through gridding and initial anomaly screening. Then, it calculates multi-dimensional quantitative indicators such as imaging fluctuation complexity, imaging fluctuation response dispersion, imaging composition consistency attenuation, and imaging structure response consistency index to dynamically characterize the spatial instability and physical response behavior of defects under different process conditions. Finally, by fusing the above indicators to generate an adaptive constraint factor, the preliminary anomaly detection results are subjected to physical consistency weighted modulation, thereby significantly distinguishing between real defects and process noise and suppressing false alarms and missed alarms caused by operating condition fluctuations. This method effectively improves the cross-batch consistency and spatial reliability of detection results, providing a stable and reliable data foundation for forming process optimization, accurate quality assessment, and product reliability analysis.

[0088] This invention also proposes a molding defect detection system for the fabrication process of silicon carbide thin-walled tubes. Please refer to [link to relevant documentation]. Figure 2 The diagram shows a structural diagram of a molding defect detection system for the silicon carbide thin-walled tube manufacturing process provided by an embodiment of the present invention. The system includes: a data acquisition module 101, a data processing module 102, and a defect detection module 103.

[0089] The data acquisition module 101 is used to acquire surface images of silicon carbide thin-walled tubes under different working conditions and record the working condition identifier and spatial location information corresponding to each image. The data processing module 102 is used to divide the surface image into at least two local sub-regions based on spatial location information, filter out abnormal sub-regions, obtain the preliminary abnormality of each abnormal sub-region, and cluster the abnormal sub-regions into at least one defect candidate region. Edge detection is performed on each defect candidate region to obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and the imaging fluctuation complexity of the defect candidate region is calculated based on the average gradient value of the edges. Obtain the imaging fluctuation complexity of any candidate defect region under different operating conditions, and calculate the imaging fluctuation operating condition response dispersion of the candidate defect region based on the difference in imaging fluctuation complexity under different operating conditions. The number of edge lines in the candidate defect region under different working conditions is obtained. Based on the dispersion of the imaging fluctuation response and the change of the number of edge lines with the working conditions, the imaging composition uniform attenuation of the candidate defect region is calculated. Obtain the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different working conditions, and calculate the imaging structure response consistency index of the candidate defect region. The defect detection module 103 is used to generate a constraint factor based on the image structure consistency attenuation and the image structure response consistency index, and to use the constraint factor to modulate the preliminary anomaly of the defect candidate region and output the true defect detection value.

[0090] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the molding defect detection system for the silicon carbide thin-walled tube manufacturing process and the molding defect detection method for the silicon carbide thin-walled tube manufacturing process provided in the above embodiments belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0091] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0092] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes, characterized in that, include: Acquire surface images of silicon carbide thin-walled tubes under different operating conditions, and record the operating condition identifier and spatial location information corresponding to each image; Based on spatial location information, the surface image is divided into at least two local sub-regions, abnormal sub-regions are filtered out, and the preliminary abnormality of each abnormal sub-region is obtained. The abnormal sub-regions are then clustered into at least one defect candidate region. Edge detection is performed on each defect candidate region to obtain the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and the imaging fluctuation complexity of the defect candidate region is calculated based on the average gradient value of the edges. Obtain the imaging fluctuation complexity of any candidate defect region under different operating conditions, and calculate the imaging fluctuation operating condition response dispersion of the candidate defect region based on the difference in imaging fluctuation complexity under different operating conditions. The number of edge lines in the candidate defect region under different working conditions is obtained. Based on the dispersion of the imaging fluctuation response and the change of the number of edge lines with the working conditions, the imaging composition uniform attenuation of the candidate defect region is calculated. Obtain the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different working conditions, and calculate the imaging structure response consistency index of the candidate defect region. Constraint factors are generated based on the consistent attenuation of the imaging structure and the consistency index of the imaging structure response. These constraint factors are then used to modulate the initial anomaly of the defect candidate region, and the true defect detection value is output.

2. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The process of dividing the surface image into at least two local sub-regions based on spatial location information, filtering out abnormal sub-regions, and obtaining a preliminary anomaly quantity for each abnormal sub-region specifically includes: Based on spatial location information, the surface image is divided into grids in the axial and circumferential directions to obtain at least two local sub-regions. The spatial location information includes the axial coordinates and circumferential angle information of the thin-walled tube. For each local sub-region, its image features are extracted, wherein the image features include at least one of the following: average gray value, gray distribution variance, and texture roughness; The image features of each local sub-region are compared with a preset normal feature threshold. If the comparison result exceeds the preset normal feature threshold, the local sub-region is determined to be an abnormal sub-region. For each abnormal sub-region, the preliminary abnormality amount of the abnormal sub-region is calculated based on the degree of deviation between its image features and the preset normal feature threshold.

3. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The step of clustering abnormal sub-regions into at least one defect candidate region specifically includes: Establish spatial location indexes for all abnormal sub-regions; Based on the preset spatial proximity determination rules and spatial location index, it is determined whether the abnormal sub-regions are adjacent. If the axial distance and circumferential angle difference between two abnormal sub-regions are both less than the preset threshold, then the two abnormal sub-regions are determined to be spatially adjacent. Spatially adjacent anomalous sub-regions are merged into the same set to form a defect candidate region; A single anomalous sub-region that has no spatial adjacency relationship is identified as a defect candidate region.

4. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The step of performing edge detection on each defect candidate region, obtaining the average gradient value of the edges of all abnormal sub-regions in each defect candidate region, and calculating the imaging fluctuation complexity of the defect candidate region based on the average gradient value specifically includes: Edge detection is performed on each abnormal sub-region within the defect candidate region to obtain edge pixels; Calculate the average gradient magnitude of all edge pixels in each abnormal sub-region, and use it as the average edge gradient value of that abnormal sub-region. Arrange all abnormal sub-regions within the defect candidate region according to their spatial order to obtain the sorting result; Calculate the absolute value of the difference between the marginal average gradient values ​​of adjacent abnormal sub-regions in each pair of sorting results within the defect candidate region, and take the first average value of the absolute values ​​of the differences of all adjacent pairs. Calculate the absolute value of the difference between the average edge gradient values ​​of all pairs of anomalous sub-regions within the defect candidate region, and take the second average value of all the absolute values ​​of the differences. Multiplying the first average value by the second average value yields the imaging fluctuation complexity of the defect candidate region.

5. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The step of obtaining the imaging fluctuation complexity of any candidate defect region under different operating conditions, and calculating the imaging fluctuation operating condition response dispersion of the candidate defect region based on the difference in imaging fluctuation complexity under different operating conditions, specifically includes: Obtain the imaging fluctuation complexity of the same defect candidate region under at least three different working conditions; Calculate the absolute value of the difference in imaging fluctuation complexity between the two consecutive operating conditions; The absolute values ​​of the differences in imaging fluctuation complexity for all continuous operating conditions are averaged, and this average value is used as the imaging fluctuation response dispersion of the defect candidate region.

6. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The process of obtaining the number of edge lines in the candidate defect region under different operating conditions, and calculating the consistent attenuation of the imaging composition of the candidate defect region based on the dispersion of the imaging fluctuation response and the variation of the number of edge lines with the operating conditions, specifically includes: Obtain the number of edge lines for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference in the number of edge lines corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the number of edge lines for all continuous working conditions is used to obtain the average value of the change in the number of edge lines. Multiplying the dispersion of the imaging fluctuation response of the defect candidate region by the mean of the change in the number of edge lines yields the uniform attenuation of the imaging structure of the defect candidate region.

7. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The process of obtaining the principal direction of the texture gradient and the imaging response intensity of the candidate defect region under different operating conditions, and calculating the imaging structure response consistency index of the candidate defect region, specifically includes: Obtain the principal direction angle values ​​of the texture gradient for the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the principal direction angle values ​​of the texture gradient corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in the principal direction angle of the texture gradient for all continuous operating conditions is used to obtain the average value of the texture direction change. Obtain the imaging response intensity values ​​of the same defect candidate region under at least three different molding conditions; Calculate the absolute value of the difference between the image response intensity values ​​corresponding to each of the two consecutive operating conditions; The average value of the absolute value of the difference in image response intensity for all continuous operating conditions is taken to obtain the average value of the image response change. The reciprocal of the mean variation of texture direction is multiplied by the reciprocal of the mean variation of image response to obtain the image structure response consistency index of the defect candidate region.

8. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The process of generating a constraint factor based on the consistency attenuation of the imaging structure and the consistency index of the imaging structure response, and using this constraint factor to modulate the preliminary anomaly amount of the defect candidate region to output the true defect detection value, specifically includes: Based on the image composition uniform attenuation and the image structure response uniformity index, a constraint factor is calculated. The value of the constraint factor is positively correlated with the image structure response uniformity index and negatively correlated with the image composition uniform attenuation. The modulated anomaly is obtained by multiplying the initial anomaly value of the defect candidate region by the constraint factor. The modulated anomaly is output as the true defect detection value of the defect candidate region.

9. The method for detecting molding defects in the fabrication process of silicon carbide thin-walled tubes according to claim 1, characterized in that, The method further includes: Based on the spatial location information of each defect candidate region, defect candidate regions that are spatially adjacent and whose actual defect detection values ​​all meet the preset conditions are aggregated into the same formed defect region. Based on the axial position range and circumferential angle range covered by the molding defect area after polymerization, the spatial position, coverage area and contour boundary of the molding defect on the surface of the thin-walled tube are determined. Based on the distribution of the actual defect detection values ​​of each defect candidate region within the aggregated forming defect region, the imaging response intensity and spatial stability of the forming defect are evaluated. Based on the assessment results, the severity level of the molding defect is classified or risk is identified, and a structured defect report is generated. The structured defect report includes at least one of the following: spatial location, outline boundary, and level information.

10. A molding defect detection system for the fabrication process of silicon carbide thin-walled tubes, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by the processor, it implements the steps of a method for detecting molding defects in the silicon carbide thin-walled tube manufacturing process as described in any one of claims 1-9.

Images