Non-destructive testing method for internal pore volume of graphite components for nuclear reactors

Abstract

This application relates to the field of computer industrial vision measurement technology, specifically to a non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors, comprising the following steps: Step 1: Emitting ultrasonic signals to the graphite component, and constructing a pore distribution model of the graphite component based on the attenuation signal of the ultrasonic signal passing through the graphite component. The pore distribution model includes the probability of the presence of pores in each region of the graphite component; Step 2: Acquiring three-dimensional CT scan images of the graphite component, fusing the three-dimensional CT scan images with the pore distribution model to obtain three-dimensional image information, and slicing the three-dimensional image information into several groups of two-dimensional image information; Step 3: Automatically segmenting the two-dimensional image information based on an image segmentation algorithm to generate the probability level of each pixel in the two-dimensional image information belonging to a pore. This method significantly improves the accuracy of non-destructive testing of the internal pore volume of graphite components by combining the advantages of ultrasonic scanning and CT imaging.

Description

Technical Field

[0001] This application relates to the field of computer industrial vision measurement technology, and more specifically, to a non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors. Background Technology

[0002] The content in this section provides only background information related to this application and may not constitute prior art.

[0003] Graphite components are an important part of nuclear reactors. To ensure the safe operation of nuclear reactors, it is necessary to inspect their internal porosity, including determining the presence of porosity, measuring the pore volume, and analyzing their distribution.

[0004] Currently, ultrasonic technology is mainly used to measure the porosity of graphite components. This method determines the presence of pores and estimates their volume and extent by analyzing the propagation characteristics of ultrasonic waves within the component (such as sound velocity and attenuation). However, this technology suffers from insufficient accuracy in practical applications. During propagation, ultrasonic waves are affected by pores throughout the component, resulting in significant scattering and attenuation. When the number of pores is small and their distribution is concentrated, this method can roughly infer the pore distribution area; however, in more general or complex situations, it is difficult to accurately measure the pore volume or accurately depict the pore distribution pattern.

[0005] In summary, existing ultrasonic-based methods for detecting internal pores in graphite components have significant limitations in quantitative measurement of pore volume and accurate determination of pore distribution patterns, and their accuracy needs to be improved. Summary of the Invention

[0006] The summary section of this application is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this application propose a non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors to address the technical problems mentioned in the background section above.

[0008] As a first aspect of this application, some embodiments of this application provide a non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors, comprising the following steps:

[0009] Step 1: Emit ultrasonic signals to the graphite component. Based on the attenuation of the ultrasonic signals as they pass through the graphite component, construct a porosity distribution model of the graphite component. The porosity distribution model includes the probability of the presence of pores in each region of the graphite component.

[0010] Step 2: Acquire three-dimensional CT scan images of graphite components, fuse the three-dimensional CT scan images with the pore distribution model to obtain three-dimensional image information, and slice the three-dimensional image information into several groups of two-dimensional image information;

[0011] Step 3: Automatically segment the two-dimensional image information based on the image segmentation algorithm to generate the probability level of each pixel in the two-dimensional image information belonging to the aperture, and convert the two-dimensional image information into a probability distribution image according to the probability level of each pixel belonging to the aperture.

[0012] Step 4: Process the probability distribution image using the growth algorithm and extract the pore region from the probability distribution image;

[0013] Step 5: Obtain the pore regions of all probability distribution images, restore all probability distribution images to three-dimensional measurement images based on the slicing method of three-dimensional CT scan images, mark the pore regions in the three-dimensional measurement images, and calculate the pore volume.

[0014] This method significantly improves the accuracy of non-destructive testing of internal pore volume in graphite components by combining the advantages of ultrasonic scanning and CT imaging. First, ultrasonic scanning rapidly obtains the probability distribution information (existence probability) of pores within the component at the physical level. Based on this, high-resolution CT scanning is performed, and the CT image processing is constrained by the ultrasonic model. This strategy overcomes the shortcomings of traditional ultrasonic methods in quantifying pore volume and determining fine distribution, while effectively avoiding misjudgments caused by image noise and distortion (especially for micropores) when relying solely on CT scanning, thus achieving more accurate pore volume measurement and three-dimensional characterization.

[0015] Existing ultrasonic porosity detection methods based on a single direction or sparse measurement points are difficult to accurately measure the porosity of local areas inside graphite components due to the influence of signal attenuation caused by the superposition of multiple pores. This results in insufficient accuracy of the constructed pore distribution model, which cannot provide reliable regional pore constraints for subsequent CT image processing.

[0016] Furthermore, step 1 includes the following steps:

[0017] Step 11: Set N acoustic transducer groups on the front and rear sides of the graphite component;

[0018] N acoustic transducer groups are set on the left and right sides of the graphite component;

[0019] N acoustic transducer groups are set on the upper and lower sides of the graphite component;

[0020] Step 12: Each acoustic transducer group transmits acoustic information to each other, and calculates the attenuation coefficient of the mutually transmitted acoustic signals to obtain attenuation information;

[0021] Step 13: Collect attenuation information for each acoustic transducer group, calculate the porosity distribution probability of the affected area of ​​each acoustic transducer group based on the attenuation information of each group, and arrange the porosity distribution probabilities of all affected areas according to the spatial position of the graphite components to form a porosity distribution model.

[0022] This scheme utilizes the strong correlation between ultrasonic attenuation and porosity by orthogonally deploying multiple sets of acoustic transducers on the six sides of the component and performing cross-measurement. It accurately calculates the porosity of each local area inside the component, forming a high-confidence pore distribution model. This provides key physical constraints for subsequent CT image segmentation and significantly improves the measurement accuracy of pore (especially micropore) volume.

[0023] In existing ultrasonic pore detection methods, the coverage area of ​​a single transducer or sparse measurement point is too large and the measurement frequency is singular, resulting in low spatial resolution of the acquired attenuation information and insufficient sensitivity to pore morphology. It is difficult to accurately correlate acoustic signals with the pore characteristics of specific micro-regions inside the component, which restricts the positioning accuracy of the pore distribution model.

[0024] Furthermore, each acoustic transducer group corresponds to a regular predetermined region on the surface of the graphite component, and the size of the predetermined region is a*a;

[0025] Each acoustic transducer group sequentially emits ultrasonic signals of different frequencies to obtain attenuation coefficients for different acoustic frequencies; the attenuation coefficients of the same acoustic transducer group at different acoustic frequencies are used as the attenuation information of that acoustic transducer group.

[0026] This scheme achieves attenuation information acquisition with high spatial resolution and multi-frequency data fusion by setting a regular detection area of ​​fixed size (a×a) for each transducer group and using multi-frequency ultrasonic scanning. The regular partitioning constructs an accurate spatial mapping relationship, and the multi-frequency measurement enhances the robustness of the pore response. The combination of the two makes the porosity calculation of each small area (a×a) more accurate, providing a reliable basis for accurate positioning and description of pore distribution in subsequent three-dimensional space.

[0027] Existing porosity detection methods based on ultrasonic attenuation struggle to effectively fuse multi-frequency acoustic signals and establish a quantitative mapping relationship between them and the probability of local pores. This results in the inability to generate a high-resolution probability model that describes the likelihood of pores at each spatial location within a component, limiting the accuracy and reliability of subsequent CT image segmentation.

[0028] Furthermore, step 13 includes the following steps:

[0029] Step 131: Input the attenuation coefficient of the same sound wave frequency into the pre-trained exponential regression model to obtain the distribution probability under the corresponding pore. Let represent the probability distribution of the k-th sound wave frequency calculated for the i-th sound wave transducer group, where i∈[1,3N], and i∈[1,N]. The calculated probability distribution for the acoustic transducer groups on the front and rear sides, i∈[N+1,2N] The calculated probability distribution for the acoustic transducer groups on the left and right sides, where i∈[2N+1,3N] The probability distribution of the acoustic transducer arrays on the left and right sides is calculated.

[0030] Step 132: All probability distributions The pore distribution model E is formed by arranging the graphite components according to their spatial positions.

[0031] In the technical solution provided in this application, the pore distribution model is based on the probability distribution of the spatial position of the graphite component. Therefore, the pore distribution model E describes the probability of pores being distributed at each position in the graphite component. Based on this, subsequent 3D CT segmentation will have greater advantages.

[0032] This scheme utilizes a pre-trained exponential regression model to process multi-frequency attenuation coefficients, quantitatively transforming acoustic signals into pore distribution probabilities for each regular detection region (a×a). It then constructs a pore distribution model E strictly according to spatial location, thereby generating a high-resolution, quantitative spatial probability map. This provides crucial and reliable physical constraints for subsequent CT image segmentation, significantly improving the accuracy of pore identification (especially for small or blurry pores).

[0033] When performing pore segmentation by directly processing 3D CT images, the difference in gray values ​​(contrast) between the pore region and the matrix material may be insufficient, especially for micropores or regions affected by noise / distortion. This makes it difficult for traditional segmentation methods to accurately distinguish pore boundaries and is easily affected by image noise and gray-level gradations (intermediate values).

[0034] Step 2 includes the following steps:

[0035] Step 21: Obtain the three-dimensional CT scan image P(x, y, z) of the graphite component, where x, y, z represent the coordinates of a pixel in the three-dimensional CT scan image, and P represents the pixel value at position (x, y, z).

[0036] Step 22: Obtain the pore distribution model E, and fuse the pore distribution model E with the three-dimensional CT scan image P according to their spatial positions to obtain the three-dimensional image information PT(x, y, z);

[0037]

[0038] Where PT(x, y, z) represents the pixel value of the three-dimensional image information PT at the position (x, y, z), k represents the index of the sound wave frequency, and K represents the total number of sound wave frequencies;

[0039] Step 23: Divide the 3D image information PT(x, y, z) into Z 2D image information PQ(x, y). z z represents the two-dimensional image information PQ(x, y). z The index, Z, represents the total number of pixel values ​​in the height direction in the 3D CT scan image P(x, y, z), and PQ(x, y) is... z This represents the pixel value of the z-th two-dimensional image information at position (x, y).

[0040] This scheme spatially fuses the pore distribution probability model E with the original CT image P, significantly enhancing the image features of the pore region: the pixel values ​​of high pore probability regions are effectively suppressed (weighted down), while the pixel values ​​of low pore probability regions are relatively increased, drastically increasing the contrast between pore and non-pore regions. This effectively suppresses the interference of image noise and grayscale gradation (intermediate value), providing more accurate two-dimensional image information for subsequent image segmentation to define pores.

[0041] When directly processing high-dimensional, continuous two-dimensional CT image information (especially the image after fusion and enhancement in step 2) for aperture segmentation, the original data has high dimensionality and high information redundancy, resulting in high computational complexity and low efficiency of subsequent segmentation algorithms, and making it difficult to effectively extract key features.

[0042] Furthermore, step 3 includes the following steps:

[0043] Step 31: Predefine c category labels;

[0044] Step 32: Randomly generate several cluster centers and membership matrices in the two-dimensional image information, and iteratively update the cluster centers and membership matrices based on fuzzy adjustment to generate the category label distribution information of each pixel in the two-dimensional image information;

[0045] Each category label corresponds to the probability level of a pixel belonging to a pore, and the category label distribution information is transformed into a probability distribution image TR.

[0046] This scheme uses a fuzzy clustering algorithm to reduce the dimensionality of high-dimensional image information and map it to a predefined set of c probability levels, which significantly reduces the data dimensionality and computational complexity. The generated probability distribution image TR clearly represents the probability that each pixel belongs to a pore with discrete probability levels (rather than the original continuous pixel values), providing a simpler and easier-to-process feature representation for subsequent growth algorithms and greatly improving segmentation efficiency.

[0047] Traditional image segmentation methods struggle to effectively handle the ambiguity and uncertainty of pixels at the edges of pores (i.e., pixels may belong partly to pores and partly to the matrix) when processing CT images of graphite components. This results in inaccurate pore boundary localization, particularly poor edge segmentation performance for micropores or low-contrast regions.

[0048] Furthermore, step 32 includes the following steps:

[0049] Step 321: Predefine parameters: fuzzy weight m, relative weight of the first spatial function p, relative weight of the second spatial function q, smoothing parameter h, maximum number of iterations T, non-membership parameter λ, neighborhood window size d, termination threshold ε;

[0050] Step 322: Define the initial cluster centers V in the two-dimensional image information. (0) Initial membership matrix U (0) ;

[0051] Step 323: Based on the current cluster center V (t) and membership matrix U (t) Update the temporary membership matrix U (t +1,1) ;

[0052] In the first iteration: t = 0;

[0053] Where i` represents the index of a pixel in the two-dimensional image information, j` and k` represent the indexes of the category label, j`∈[1,c], k`∈[1,c]; This represents the initial membership degree of sample i' to the j'th class label in the (t+1)th iteration; n represents the total number of pixels in the two-dimensional image information, and c represents the total number of class labels;

[0054] d j`i` d represents the Euclidean distance from pixel i' to category label j'. k`i` Let represent the Euclidean distance from sample i` to class label k`;

[0055] Step 324: Based on the temporary membership matrix U (t+1,1) Calculate the intuitive membership matrix U (t+1,2) ;

[0056]

[0057] π j`i` Indicates the degree of hesitation, v j` i ` r represents the membership degree of pixel i' to category label j'; j`i` Indicates the adjustment factor;

[0058] Step 325: Based on the intuitive membership matrix U (t+1,2) And the spatial function matrix H of the domain of the parameter h is calculated;

[0059] h j`i` Represents the element values ​​in the spatial function matrix H;

[0060] This represents a d*d neighborhood window centered at pixel i'. w(i`, z`) represents the membership degree of the neighboring pixel z` to the category label j`; w(i`, z`) represents the weight factor, and z` represents the index of the neighboring pixel;

[0061]

[0062] h represents the smoothing parameter;

[0063]

[0064] Step 326: Based on the intuitive membership matrix U (t+1,2) The updated membership matrix is ​​calculated using the spatial function matrix H, the relative weights p of the first spatial function, and the relative weights q of the second spatial function.

[0065]

[0066] Step 327: Update the cluster centers based on the updated membership matrix;

[0067]

[0068] Step 328: Check the iteration termination condition. If the iteration termination condition is met, generate the category label for pixel i based on the final membership matrix U; i` This represents the category label of the i-th pixel;

[0069]

[0070] Each category label corresponds to the probability level of a pixel belonging to a pore, and the category label distribution information is transformed into a probability distribution image TR.

[0071] The category label distribution information is converted into a probability distribution image TR.

[0072] This scheme introduces intuitionistic fuzzy set theory (including membership, non-membership, and hesitation) and combines it with spatial neighborhood information, enabling the algorithm to more accurately quantify the fuzziness and uncertainty of whether a pixel belongs to a pore (or not), especially for the classification of pixels with blurred boundaries. At the same time, the introduction of spatial functions effectively utilizes the correlation between local pixels, significantly improving the accuracy and robustness of pore edge (especially micropore) segmentation.

[0073] Furthermore, in step 328, the termination condition is:

[0074] Condition 1: ΔU < ε, where ΔU represents the change in the membership matrix. This indicates taking the maximum value;

[0075] Condition 2:t+1>T.

[0076] The technical solution provided in this application can limit the number of iterations, reduce the amount of computation, and minimize the change in the membership matrix by setting two termination conditions, thereby obtaining the optimal membership matrix.

[0077] When directly identifying pores based on probability distribution images, isolated high-probability pixels or local noise may cause pore regions to break (discontinuous) or false pores, making it difficult to accurately extract complete and connected pore structures and affecting the accuracy of subsequent volume calculations.

[0078] Furthermore, step 4 includes the following steps:

[0079] Step 41: Obtain the probability distribution image TR, and perform region growing on the probability distribution image TR to obtain several segmented regions;

[0080] Step 42: For each segmented region, calculate the average pixel value of the segmented region in the original two-dimensional image information. If the average pixel value is lower than the preset selection threshold, the segmented region is regarded as a pore region.

[0081] Step 43: Merge all pore regions. In the original two-dimensional image information, set the pore regions to 1 and the remaining regions to 0 to obtain a binary image.

[0082] The technical solution provided in this application can further segment pixels in a probability distribution image that are close to the pore features by using region growing. The final segmentation result can further reduce the dimensionality of the probability distribution image, thereby transforming the original two-dimensional image information into a background image containing only entities and pores.

[0083] Traditional threshold-based pore segmentation methods struggle to effectively aggregate spatially discrete but similar pixel points, leading to pore region fragmentation or the inclusion of non-pore noise points, thus failing to guarantee the continuity and structural integrity of the extracted pores.

[0084] Further, step 411: Create a label matrix of the same size as the probability distribution image TR, and initialize all pixels to the "unclassified" state (usually represented by 0);

[0085] Initialize an empty queue to store the boundary pixels to be processed;

[0086] Set the starting region label to 1;

[0087]

[0088] Where L represents the label matrix, Q represents the pixel processing queue, and μ label R represents the feature value of the current region, R represents the set of pixels in the current region, and label represents the label of the current region.

[0089] Step 412: Set a seed set S, select a pixel as a seed point, mark it as the current region label, and add it to the region pixel set; use the category label of the seed point as the initial feature of the region, add the seed point to the processing queue, and obtain the updated label matrix L, the queue Q containing the seed point, and the initialized region feature u.

[0090]

[0091] S represents the seed set, s represents the index of the seed in the seed set, and I(s) represents the pixel value of the seed point.

[0092] Step 413: Process the pixels in the queue repeatedly until the queue is empty:

[0093] Retrieve pixel l from the head of the queue;

[0094] Check all neighboring pixels o of pixel l;

[0095] Determine if o is marked; if o is not marked, search for a new neighboring pixel.

[0096] If o is marked, calculate the similarity between o and l. If the similarity is less than the preset value, add o to the queue where l is located, add o to the region where it is located, and update the region features. If the queue becomes an empty set, the growth ends.

[0097] The final output in step 413 is: the updated label matrix L, the empty queue Q, the current region label, and the updated region features;

[0098] Step 414: Check if there are any unprocessed seed points. If so, return to step 412 to process new seeds. If not, check if there are any unlabeled pixels. Mark the unlabeled pixels as background or select new seeds to continue growing. Output the final segmentation result L.

[0099] This scheme effectively aggregates adjacent pixels with similar pore probability characteristics by performing region growing on the probability distribution image and combining it with the gray-level mean of the original image for screening, forming connected candidate regions. Then, it verifies their gray-level characteristics (pore regions should be darker) by setting a preset threshold, and finally generates a binary mask image that accurately represents pores / non-pores, which significantly improves the continuity and completeness of pore structure extraction and lays the foundation for accurate volume calculation.

[0100] This scheme employs a dynamic seed growth and neighborhood similarity determination mechanism: starting from the seed point, it recursively aggregates neighboring pixels that meet the similarity conditions to form a spatially continuous pore region; during the growth process, it updates the region feature values ​​in real time to ensure the consistency of attributes between the expanded region and the seed region; and through a secondary processing mechanism for unclassified pixels (marking the background or new seeds), it completely solves the problem of fragmented pore omission and significantly improves the integrity of the pore structure.

[0101] Step 5 includes the following steps:

[0102] Step 51: Obtain the binary image of each two-dimensional image information, and obtain the volume VT of the pixel in the real world in the binary image;

[0103] Step 52: Replace the two-dimensional image information with the binary image, and restore the three-dimensional measurement image by cutting the two-dimensional image information in the same way;

[0104] Step 53: Mark the pixels with a pixel value of 1 in the 3D measurement image as pore regions, obtain the number G of all pixels marked as pore regions in the 3D measurement image, and calculate the pore volume VH; VH=G×VT.

[0105] The technical solution of this application embodiment has at least the following advantages and beneficial effects:

[0106] This scheme achieves high-precision three-dimensional quantization of the internal pores (especially sub-millimeter-level micropores) of graphite components through multimodal data fusion and adaptive segmentation algorithms:

[0107] Pore ​​distribution modeling: A high-resolution pore probability model is constructed using orthogonal multi-frequency ultrasonic attenuation coefficients to provide physical constraints for CT image processing;

[0108] Image feature enhancement: By fusing the probabilistic model with the original CT at the pixel level, the contrast between pores and matrix is ​​significantly increased, and noise interference is suppressed;

[0109] Intelligent segmentation optimization: Combining intuitive fuzzy clustering dimensionality reduction with dynamic region growing, it accurately identifies fuzzy boundary gaps to ensure connectivity;

[0110] Accuracy of volume calculation: The final generated binary mask completely preserves the pore topology, and the error rate of pore volume measurement is reduced. Attached Figure Description

[0111] Figure 1 A flowchart of a non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors.

[0112] Figure 2 A schematic diagram showing the graphite component divided into rectangular prisms.

[0113] Figure 3 This is a partial view of two-dimensional image information.

[0114] Figure 4 This is a flowchart for step 413.

[0115] Figure label:

[0116] 100. Graphite components; 1. Affected area; 2. Acoustic transducer. Detailed Implementation

[0117] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments. The same reference numerals in the accompanying drawings represent the same components. It should be noted that the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the described embodiments of this application without creative effort are within the scope of protection of this application.

[0118] Compared to the embodiments shown in the accompanying drawings, feasible embodiments within the scope of this application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or components with different connections, etc. Furthermore, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0119] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” and similar terms used in this specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not necessarily indicate a quantity limitation. Terms such as “upper” and “lower” are used only to indicate relative positional relationships, and these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0120] refer to Figure 1 Example 1: A non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors, comprising the following steps:

[0121] Step 1: Emit ultrasonic signals to the graphite component. Based on the attenuation of the ultrasonic signals as they pass through the graphite component, construct a porosity distribution model of the graphite component. The porosity distribution model includes the probability of the presence of pores in each region of the graphite component.

[0122] Step 1 includes the following steps:

[0123] Step 11: Set N acoustic transducer groups on the front and rear sides of the graphite component;

[0124] N acoustic transducer groups are set on the left and right sides of the graphite component;

[0125] N acoustic transducer groups are set on the upper and lower sides of the graphite component;

[0126] Each acoustic transducer group corresponds to a regular predetermined area on the surface of the graphite component, and the size of the predetermined area is a*a.

[0127] like Figure 2 As shown, graphite components are generally regular rectangular bodies. An acoustic transducer matrix is ​​arranged on the surface of the graphite component, with each transducer corresponding to a predetermined region. In effect, these acoustic transducers divide the graphite component into smaller rectangular bodies (affected regions). For each rectangular body, the porosity measured by the acoustic transducer array in three directions can be used as an indicator of the presence of pores within that rectangular body.

[0128] Step 12: Each acoustic transducer group transmits acoustic information to each other, and calculates the attenuation coefficient of the mutually transmitted acoustic signals to obtain attenuation information;

[0129] Each acoustic transducer group sequentially emits ultrasonic signals of different frequencies to obtain attenuation coefficients for different acoustic frequencies; the attenuation coefficients of the same acoustic transducer group at different acoustic frequencies are used as the attenuation information of that acoustic transducer group.

[0130] When ultrasound passes through porous materials, amplitude attenuation is the most sensitive and stable acoustic parameter characterizing porosity. The higher the porosity, the greater the attenuation. Different pore diameters require signals of different wavelengths to ensure that amplitude attenuation remains the most sensitive and stable acoustic parameter characterizing porosity. Therefore, each acoustic transducer group needs to emit ultrasonic signals of different frequencies sequentially.

[0131] For example, 5MHz can be used as the center frequency, with a step size of 1MHz. Each transducer group first emits a 5MHz acoustic signal, then a 4MHz acoustic signal, then a 6MHz acoustic signal, and so on, until the upper and lower limits of the frequency are reached.

[0132]

[0133] in, A represents the attenuation coefficient calculated from the k-th sound wave frequency of the i-th acoustic transducer group, D represents the thickness of the graphite component, and A represents the thickness of the graphite component. r ef represents the peak voltage amplitude of the reference signal. This represents the peak voltage amplitude obtained at the k-th acoustic frequency of the i-th acoustic transducer group.

[0134] Step 13: Collect attenuation information for each acoustic transducer group, calculate the porosity distribution probability of the affected area of ​​each acoustic transducer group based on the attenuation information of each group, and arrange the porosity distribution probabilities of all affected areas according to the spatial position of the graphite components to form a porosity distribution model.

[0135] Step 13 includes the following steps:

[0136] Step 131: Input the attenuation coefficient of the same sound wave frequency into the pre-trained exponential regression model to obtain the distribution probability under the corresponding pore. Let represent the probability distribution of the k-th sound wave frequency calculated for the i-th sound wave transducer group, where i∈[1,3N], and i∈[1,N]. The calculated probability distribution for the acoustic transducer groups on the front and rear sides, i∈[N+1,2N] The calculated probability distribution for the acoustic transducer groups on the left and right sides, where i∈[2N+1,3N] The probability distribution calculated for the acoustic transducer groups on the upper and lower sides;

[0137] Step 132: All probability distributions The pore distribution model E is formed by arranging the graphite components according to their spatial positions.

[0138] The reason why this application adopts the exponential regression model is mainly because the higher the porosity, the greater the signal attenuation, and this growth usually shows a nonlinear trend of rapid increase followed by saturation.

[0139] The exponential regression model is as follows:

[0140]

[0141] in, This represents the probability distribution of the k-th acoustic frequency of the i-th acoustic transducer group calculated by the exponential regression model, where a, b, and c are the parameters that the exponential regression model needs to fit. This represents the attenuation coefficient calculated from the k-th acoustic frequency of the i-th acoustic transducer group.

[0142] In other words, for graphite components, it is necessary to prepare 5-10 samples with the same process and thickness (d_i) as the graphite component to be tested. Then, by applying pressure to the graphite component, more calibration samples with known porosity are obtained in the laboratory. These calibration samples are used to fit the exponential regression model to obtain the specific values ​​of a, b, and c. Then, the exponential regression model is applied to step 131.

[0143] refer to Figure 3 Step 2: Acquire three-dimensional CT scan images of graphite components, fuse the three-dimensional CT scan images with the pore distribution model to obtain three-dimensional image information, and slice the three-dimensional image information into several groups of two-dimensional image information.

[0144] Step 2 includes the following steps:

[0145] Step 21: Obtain the three-dimensional CT scan image P(x, y, z) of the graphite component, where x, y, z represent the coordinates of a pixel in the three-dimensional CT scan image, and P represents the pixel value at position (x, y, z).

[0146] Step 22: Obtain the pore distribution model E, and fuse the pore distribution model E with the three-dimensional CT scan image P according to their spatial positions to obtain the three-dimensional image information PT(x, y, z);

[0147]

[0148] Where PT(x, y, z) represents the pixel value of the 3D image information PT at the position (x, y, z), k represents the index of the sound wave frequency, and K represents the total number of sound wave frequencies. This represents the probability distribution at position (x, y, z);

[0149] Step 23: Divide the 3D image information PT(x, y, z) into Z 2D image information PQ(x, y). zz represents the two-dimensional image information PQ(x, y). z The index, Z, represents the total number of pixel values ​​in the height direction in the 3D CT scan image P(x, y, z), and PQ(x, y) is... z This represents the pixel value of the z-th two-dimensional image information at position (x, y).

[0150] 3D CT scan image slicing involves dividing the voxel in a 3D CT image layer by layer to obtain several layers of 2D image information. Of course, in practice, the voxel resolution of the 3D CT scan image can also be adjusted first.

[0151] Step 3: Automatically segment the two-dimensional image information based on the image segmentation algorithm to generate the probability level of each pixel in the two-dimensional image information belonging to the aperture. Based on the probability level of each pixel belonging to the aperture, the two-dimensional image information is transformed into a probability distribution image.

[0152] Step 3 includes the following steps:

[0153] Step 31: Predefine c category labels; for example, they can be roughly divided into 4 labels: high probability pore label, low probability pore label, low probability entity label, and high probability entity label. Each label corresponds to the probability value of a pixel belonging to a pore.

[0154] Step 32: Randomly generate several cluster centers and membership matrices in the two-dimensional image information. Based on fuzzy adjustment, iteratively update the cluster centers and membership matrices to generate the category label distribution information of each pixel in the two-dimensional image information.

[0155] Step 32 includes the following steps:

[0156] Step 321: Predefine parameters: fuzzy weight m, relative weight of the first spatial function p, relative weight of the second spatial function q, smoothing parameter h, maximum number of iterations T, non-membership parameter λ, neighborhood window size d, termination threshold ε;

[0157] Step 322: Define the initial cluster centers V in the two-dimensional image information. (0) Initial membership matrix U (0) ;

[0158] Step 323: Based on the current cluster center V (t) and membership matrix U (t) Update the temporary membership matrix U (t +1,1) ;

[0159] In the first iteration: t = 0;

[0160] Where i` represents the index of a pixel in the two-dimensional image information, j` and k` represent the indexes of the category label, j`∈[1,c], k`∈[1,c]; This represents the initial membership degree of sample i' to the j'th class label in the (t+1)th iteration; n represents the total number of pixels in the two-dimensional image information, and c represents the total number of class labels;

[0161] d j`i` d represents the Euclidean distance from pixel i' to category label j'. k`i` Let represent the Euclidean distance from sample i` to class label k`;

[0162] Step 324: Based on the temporary membership matrix U (t+1,1) Calculate the intuitive membership matrix U (t+1,2) ;

[0163]

[0164] π j`i` Indicates the degree of hesitation, v j` i ` r represents the membership degree of pixel i' to category label j'; j`i` Indicates the adjustment factor;

[0165] Step 325: Based on the intuitive membership matrix U (t+1,2) And the spatial function matrix H of the domain of the parameter h is calculated;

[0166] h j`i` Represents the element values ​​in the spatial function matrix H;

[0167] This represents a d*d neighborhood window centered at pixel i'. w(i`, z`) represents the membership degree of the neighboring pixel z` to the category label j`; w(i`, z`) represents the weight factor, and z` represents the index of the neighboring pixel;

[0168]

[0169] h represents the smoothing parameter;

[0170]

[0171] Step 326: Based on the intuitive membership matrix U (t+1,2) The updated membership matrix is ​​calculated using the spatial function matrix H, the relative weights p of the first spatial function, and the relative weights q of the second spatial function.

[0172]

[0173] Step 327: Update the cluster centers based on the updated membership matrix;

[0174]

[0175] Step 328: Check the iteration termination condition. If the iteration termination condition is met, generate the category label for pixel i based on the final membership matrix U; i` This represents the category label of the i-th pixel;

[0176]

[0177] Each category label corresponds to the probability level of a pixel belonging to a pore, and the category label distribution information is transformed into a probability distribution image TR.

[0178] The category label distribution information is converted into a probability distribution image TR.

[0179] In step 328, the termination condition is:

[0180] Condition 1: ΔU < ε, where ΔU represents the change in the membership matrix. This indicates taking the maximum value;

[0181] Condition 2:t+1>T.

[0182] Step 3 essentially involves dimensionality reduction of the image, roughly grouping similar image information together using fuzzy segmentation. This removes some noise and smooths the image. Further image segmentation is then required.

[0183] Step 4: Process the probability distribution image using the growth algorithm and extract the pore region from the probability distribution image;

[0184] Step 4 includes the following steps:

[0185] Step 41: Obtain the probability distribution image TR, and perform region growing on the probability distribution image TR to obtain several segmented regions;

[0186] Furthermore, step 41 includes the following steps:

[0187] Step 411: Create a label matrix of the same size as the probability distribution image TR, and initialize all pixels to the "unclassified" state (usually represented by 0);

[0188] Initialize an empty queue to store the boundary pixels to be processed;

[0189] Set the starting region label to 1;

[0190]

[0191] Where L represents the label matrix, Q represents the pixel processing queue, and μ label R represents the feature value of the current region, R represents the set of pixels in the current region, and label represents the label of the current region.

[0192] Step 412: Set a seed set S, select a pixel as a seed point, mark it as the current region label, and add it to the region pixel set; use the category label of the seed point as the initial feature of the region, add the seed point to the processing queue, and obtain the updated label matrix L, the queue Q containing the seed point, and the initialized region feature u.

[0193]

[0194] S represents the seed set, s represents the index of the seed in the seed set, and I(s) represents the pixel value of the seed point.

[0195] refer to Figure 4 Step 413: Process the pixels in the queue repeatedly until the queue is empty.

[0196] Retrieve pixel l from the head of the queue;

[0197] Check all neighboring pixels o of pixel l;

[0198] Determine if o is marked; if o is not marked, search for a new neighboring pixel.

[0199] If o is marked, the similarity between o and l is calculated. If the similarity is less than the preset value, o is added to the queue where l is located, o is added to the region where it is located, and the region features are updated. If the queue becomes an empty set, the growth ends. The similarity is cosine similarity.

[0200] The final output in step 413 is: the updated label matrix L, the empty queue Q, the current region label, and the updated region features;

[0201] Step 414: Check if there are any unprocessed seed points. If so, return to step 412 to process new seeds. If not, check if there are any unlabeled pixels. Mark the unlabeled pixels as background or select new seeds to continue growing. Output the final segmentation result L. The segmentation result L covering the probability distribution image TR is the segmentation result of the probability distribution image TR. In other words, the segmented region of the probability distribution image TR can be obtained from the segmentation result.

[0202] Step 41 involves setting a seed to continuously segment similar continuous regions on the probability distribution image TR, thereby accurately distinguishing pores from solids.

[0203] Step 42: For each segmented region, calculate the average pixel value of the segmented region in the original two-dimensional image information. If the average pixel value is lower than the preset selection threshold, the segmented region is regarded as a pore region.

[0204] Step 5: Obtain the pore regions of all probability distribution images, restore all probability distribution images to three-dimensional measurement images based on the slicing method of three-dimensional CT scan images, mark the pore regions in the three-dimensional measurement images, and calculate the pore volume.

[0205] Step 5 includes the following steps:

[0206] Step 51: Obtain the binary image of each two-dimensional image information, and obtain the volume VT of the pixel in the real world in the binary image;

[0207] Step 52: Replace the two-dimensional image information with the binary image, and restore the two-dimensional image that has replaced the two-dimensional image information to a three-dimensional measurement image by cutting the two-dimensional image information;

[0208] Step 53: Mark the pixels with a pixel value of 1 in the 3D measurement image as pore regions, obtain the number G of all pixels marked as pore regions in the 3D measurement image, and calculate the pore volume VH; VH=G×VT.

[0209] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors, characterized in that, include: Step 1: Emit ultrasonic signals to the graphite component. Based on the attenuation of the ultrasonic signals as they pass through the graphite component, construct a porosity distribution model of the graphite component. The porosity distribution model includes the probability of the presence of pores in each region of the graphite component. Step 2: Acquire three-dimensional CT scan images of graphite components, fuse the three-dimensional CT scan images with the pore distribution model to obtain three-dimensional image information, and slice the three-dimensional image information into several groups of two-dimensional image information; Step 3: Automatically segment the two-dimensional image information based on the image segmentation algorithm to generate the probability level of each pixel in the two-dimensional image information belonging to the aperture, and convert the two-dimensional image information into a probability distribution image according to the probability level of each pixel belonging to the aperture. Step 4: Process the probability distribution image using the growth algorithm and extract the pore region from the probability distribution image; Step 5: Obtain the pore regions of all probability distribution images, reconstruct all probability distribution images into 3D measurement images based on the slicing method of 3D CT scan images, mark the pore regions in the 3D measurement images, and calculate the pore volume; Step 2 includes the following steps: Step 21: Obtain the three-dimensional CT scan image P(x, y, z) of the graphite component, where x, y, z represent the coordinates of a pixel in the three-dimensional CT scan image, and P represents the pixel value at position (x, y, z). Step 22: Obtain the pore distribution model E, and fuse the pore distribution model E with the three-dimensional CT scan image P according to their spatial positions to obtain the three-dimensional image information PT(x, y, z). ； Where PT(x, y, z) represents the pixel value of the 3D image information PT at the position (x, y, z), k represents the index of the sound wave frequency, and K represents the total number of sound wave frequencies. This represents the probability distribution at position (x, y, z); Step 23: Divide the three-dimensional image information PT(x, y, z) into Z two-dimensional image information PQ(x, y). z z represents the two-dimensional image information PQ(x, y). z The index, Z, represents the total number of pixel values ​​in the height direction in the 3D CT scan image P(x, y, z), and PQ(x, y) is... z This represents the pixel value of the z-th two-dimensional image information at position (x, y).

2. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 1, characterized in that, Step 1 includes the following steps: Step 11: Set N acoustic transducer groups on the front and rear sides of the graphite component; N acoustic transducer groups are set on the left and right sides of the graphite component; N acoustic transducer groups are set on the upper and lower sides of the graphite component; Step 12: Each acoustic transducer group transmits acoustic information to each other, and calculates the attenuation coefficient of the mutually transmitted acoustic signals to obtain attenuation information; Step 13: Collect attenuation information for each acoustic transducer group, calculate the porosity distribution probability of the affected area of ​​each acoustic transducer group based on the attenuation information of each group, and arrange the porosity distribution probabilities of all affected areas according to the spatial position of the graphite components to form a porosity distribution model.

3. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 2, characterized in that, Each acoustic transducer group corresponds to a regular predetermined area on the surface of the graphite component, and the size of the predetermined area is a*a; Each acoustic transducer group sequentially emits ultrasonic signals of different frequencies to obtain attenuation coefficients for different acoustic frequencies; the attenuation coefficients of the same acoustic transducer group at different acoustic frequencies are used as the attenuation information of that acoustic transducer group.

4. The non-destructive testing method for the internal pore volume of graphite components used in nuclear reactors according to claim 3, characterized in that, Step 13: Includes the following steps: Step 131: Input the attenuation coefficient of the same sound wave frequency into the pre-trained exponential regression model to obtain the distribution probability under the corresponding pore. , Let represent the probability distribution of the k-th sound wave frequency calculated for the i-th sound wave transducer group, where i∈[1,3N], and i∈[1,N]. The calculated probability distribution for the acoustic transducer groups on the front and rear sides, i∈[N+1,2N] The calculated probability distribution for the acoustic transducer groups on the left and right sides, where i∈[2N+1,3N] The probability distribution calculated for the acoustic transducer groups on the upper and lower sides; Step 132: All probability distributions The pore distribution model E is formed by arranging the graphite components according to their spatial positions.

5. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 1, characterized in that, Furthermore, step 3 includes the following steps: Step 31: Predefine c category labels; Step 32: Randomly generate several cluster centers and membership matrices in the two-dimensional image information, and iteratively update the cluster centers and membership matrices based on fuzzy adjustment to generate the category label distribution information of each pixel in the two-dimensional image information; Each category label corresponds to the probability level of a pixel belonging to a pore, and the category label distribution information is transformed into a probability distribution image TR.

6. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 5, characterized in that, Step 32 includes the following steps: Step 321: Predefine parameters: fuzzy weight m, relative weight of the first spatial function p, relative weight of the second spatial function q, smoothing parameter h, maximum number of iterations T, non-membership parameter λ, neighborhood window size d, termination threshold ε; Step 322: Define the initial cluster centers V in the two-dimensional image information. （0） Initial membership matrix U （0） ; Step 323: Based on the current cluster center V （t） and membership matrix Update the temporary membership matrix ; In the first iteration: t=0; ; Where i` represents the index of a pixel in the two-dimensional image information, j` and k` represent the indexes of the category label, j`∈[1,c], k`∈[1,c]; This represents the initial membership degree of sample i' to the j'th class label in the (t+1)th iteration; n represents the total number of pixels in the two-dimensional image information, and c represents the total number of class labels; This represents the Euclidean distance from pixel i' to category label j'. Let represent the Euclidean distance from sample i` to class label k`; Step 324: Based on the temporary membership matrix Calculate the intuitive membership matrix ; ； ； ； ; Indicates the degree of hesitation. This represents the membership degree of pixel i' to category label j'; Indicates the adjustment factor; Step 325: Based on the intuitive membership matrix And the spatial function matrix H of the domain of the parameter h is calculated; ; Represents the element values ​​in the spatial function matrix H; This represents a d*d neighborhood window centered at pixel i'. This represents the membership degree of the neighboring pixel z' to the category label j'; z represents the weighting factor, and z` represents the index of the pixel in the neighborhood. ； h represents the smoothing parameter; ； Step 326: Based on the intuitive membership matrix The updated membership matrix is ​​calculated using the spatial function matrix H, the relative weights p of the first spatial function, and the relative weights q of the second spatial function. represents the normalization factor, and exp represents the exponential function with the natural constant as the base; ； ； Step 327: Update the cluster centers based on the updated membership matrix; ； ； Step 328: Check the iteration termination condition. If the iteration termination condition is met, generate the category label of pixel i based on the final membership matrix U. This represents the category label of the i-th pixel; ； Each category label corresponds to the probability level of a pixel belonging to a pore, and the category label distribution information is transformed into a probability distribution image TR; The category label distribution information is converted into a probability distribution image TR.

7. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 6, characterized in that, In step 328, the termination condition is: Condition 1: ; <ε, where, This represents the change in the membership matrix. This indicates taking the maximum value; Condition 2:t+1>T.

8. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 1, characterized in that, Step 4 includes the following steps: Step 41: Obtain the probability distribution image TR, and perform region growing on the probability distribution image TR to obtain several segmented regions; Step 42: For each segmented region, calculate the average pixel value of the segmented region in the original two-dimensional image information. If the average pixel value is lower than the preset selection threshold, the segmented region is regarded as a pore region. Step 43: Merge all pore regions. In the original two-dimensional image information, set the pore regions to 1 and the remaining regions to 0 to obtain a binary image.

9. The method for non-destructive testing of the internal pore volume of graphite components for nuclear reactors according to claim 1, characterized in that, Step 5 includes the following steps: Step 51: Obtain the binary image of each two-dimensional image information, and obtain the volume VT of the pixel in the real world in the binary image; Step 52: Replace the two-dimensional image information with the binary image, and restore the two-dimensional image that has replaced the two-dimensional image information to a three-dimensional measurement image by cutting the two-dimensional image information; Step 53: Mark the pixels with a pixel value of 1 in the 3D measurement image as pore regions, obtain the number G of all pixels marked as pore regions in the 3D measurement image, and calculate the pore volume VH; VH=G×VT.

Images