A real-time defect detection method for nuclear fuel elements

Abstract

The present invention discloses a real-time defect detection method for nuclear fuel elements, which relates to the technical field of nuclear fuel element detection and includes: inputting the image of the nuclear fuel element to be detected into a pre-trained defect detection model based on the NCSN network, calculating the NCSN multi-level loss, where the NCSN multi-level loss is the score loss of the nuclear fuel element image under different levels of noise in the NCSN network; using a defect screening network based on the K-means algorithm to screen out defect images from the images of the nuclear fuel elements to be detected according to the NCSN multi-level loss; using the NCSN network to reconstruct the defect images, obtaining normal images with defects removed based on the defect images, calculating the residual images of the defect images and the normal images, and using Otsu threshold segmentation to locate the defect positions. The present invention can provide higher efficiency, stability and accuracy in the real-time defect detection of nuclear fuel elements, and promotes the intelligent and efficient development of industrial detection technologies.

Description

Technical Field

[0001] The present invention relates to the technical field of nuclear fuel element detection, and more particularly, to a real-time defect detection method for nuclear fuel elements. Background Art

[0002] Nuclear fuel elements are the most core and critical components of nuclear reactors, and their structural integrity is one of the key factors directly affecting the technical state of nuclear reactors. The structural integrity problems of fuel elements have obvious effects on the corrosion resistance, impact resistance, and heat transfer performance of nuclear fuels, and also cause problems such as defect migration, breakage, and leakage of fuel elements, leading to the leakage of radioactive substances. Therefore, the structural integrity detection of fuel elements is an essential step to ensure the safe operation of reactors.

[0003] For the defect detection of fuel elements, the goal is to identify defect information through image features. And the defect information has the following characteristics: unknownness: the defect information has unknown sudden behaviors, data structures, and distributions; heterogeneity: defects are usually irregular; class imbalance: normal data will account for the vast majority of the data, while the data containing defects is relatively small.

[0004] Currently, conventional manual methods are mainly used for defect identification, and a large amount of detection data will be generated in batch detection, resulting in low efficiency, easy missed detection, and the influence of subjective factors. With the development of computer technology, methods for automatically detecting fuel elements using intelligent algorithms have emerged. For example, many supervised learning methods have been applied to anomaly detection and have achieved good results. However, it is difficult to obtain a large number of labeled data sets in the defect detection of fuel elements, which will affect the detection results. Summary of the Invention

[0005] The present invention aims to provide a real-time defect detection method for nuclear fuel elements, which can solve the above problems.

[0006] To solve the above problems, the technical solution adopted by the present invention is as follows:

[0007] The present invention provides a real-time defect detection method for nuclear fuel elements, including the following steps:

[0008] S1. Input the image of the nuclear fuel element to be detected into a pre-trained defect detection model based on the NCSN network, and calculate the NCSN multi-level loss. The NCSN multi-level loss is the score loss of the nuclear fuel element image under different levels of noise of the NCSN network;

[0009] S2. Use a defect screening network based on the K-means algorithm to screen out defect images from the images of the nuclear fuel elements to be detected according to the NCSN multi-level loss;

[0010] S3. Use the NCSN network to reconstruct the defective image, obtain the reconstructed image with defects removed from the defective image, calculate the residual map between the defective image and the reconstructed image, and use Otsu threshold segmentation to locate the defect position.

[0011] As a further description of the above technical solution: By calculating the fractional loss of the nuclear fuel element image under different levels of noise, train a defect detection model based on the NCSN network; the loss function used is:

[0012]

[0013] Among them, the first term is the estimated score function output by the model sθ(x), the second term is the true score function, and the loss function is the squared Euclidean distance between the score output by the model and the true score.

[0014] As a further description of the above technical solution: The network noise level of the defect detection model based on the NCSN network is a total of 10 levels.

[0015] As a further description of the above technical solution: In step S2, select the losses with noise levels of 5, 6, 7, and 8 from the NCSN multi-level losses as the input of the defect screening network.

[0016] As a further description of the above technical solution: In the process of the K-means algorithm in step S2, set the number of clustering categories to 2, which represent the normal image set and the defective image set respectively, and set the number of nuclear fuel element images to be detected for each clustering to 1.

[0017] As a further description of the above technical solution: In the process of the K-means algorithm in step S2, when the number of a certain category in the two categories is 0, perform centroid initialization C(c1, c2), where c1 is the centroid of the normal image set and c2 is the centroid of the abnormal image set; the centroid initialization process includes inputting the loss value Loss of the nuclear fuel element image to be clustered. If c1≥Loss, then let c1 = Loss, otherwise do not adjust. If c2≤Loss, then let c2 = Loss, otherwise do not adjust.

[0018] As a further description of the above technical solution: The K-means algorithm process in step S2 also includes calculating the distance between the centroid and the input image. When the number of a certain category in the two categories is 0, perform inter-class distance judgment. When the inter-class distance is greater than the threshold λ, the input image will be classified into a new category; when the number of both categories is >0, compare the distances between the input image and the two centroids, and divide the image into the category with a smaller distance.

[0019] As a further description of the above technical solution: The K-means algorithm process in step S2 further includes updating the centroid coordinates after classifying the new image.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] 1) A defect image recognition method based on the loss of different noise levels of the NCSN network is proposed. Through the training of multi-level noise loss, the recognition accuracy and efficiency of the unsupervised learning method for defect images can be effectively improved, providing a more reliable basis for subsequent real-time defect detection;

[0022] 2) Combining the K-means clustering algorithm, through real-time clustering of the loss of different noise levels of the NCSN network, combining class spacing control and historical clustering data memory, edge detection and clustering are effectively realized, thus greatly improving the efficiency of defect detection and the robustness of the model, avoiding the computational consumption of each image reconstruction, and improving the real-time detection ability.

[0023] 3) For defect images, the annealing Langevin dynamics sampling method is used to remove defects, and Otsu threshold segmentation is performed based on the residual between the image after removing defects and the original image, so as to accurately locate the defect position, further improving the accuracy and positioning accuracy of defect detection.

[0024] 4) The present invention overcomes the problems of low efficiency and instability of manual detection methods, realizes intelligent detection through deep learning, avoids the dependence on a large amount of labeled data in traditional supervised learning methods, and uses the unsupervised learning NCSN generation network method for anomaly detection, improving the automation level of detection and adapting to the challenge of scarce data in industrial environments.

[0025] 5) The real-time detection method combining K-means clustering proposed by the present invention can effectively improve the detection speed without increasing the additional computational burden, ensure the real-time performance of the anomaly detection of the generation network, and optimize the defect detection efficiency in production practice.

[0026] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specifically gives embodiments of the present invention and, in conjunction with the accompanying drawings, makes a detailed description as follows. Description of the Drawings

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0028] Figure 1 It is the overall framework diagram of the model in the embodiment;

[0029] Figure 2 It is the framework diagram of the K-means clustering algorithm in the embodiment;

[0030] Figure 3 It is the AUROC curve diagram of different noise levels in the embodiment;

[0031] Figure 4 It is the defective image and positioning result in the embodiment. Specific implementation manner

[0032] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0033] Figure 1 It is the main flow chart of the technical solution of the embodiment of the present invention. A real-time defect detection method for nuclear fuel elements proposed in the embodiment of the present invention includes the following steps:

[0034] Step (1): Construct a dataset covering 1460 X-ray images of fuel element substitutes. These images are cross-sectional images of CT images of fuel element substitutes and include normal image samples and defective image samples. The dataset includes a training set and a test set. Among them, the training set is 1260 normal image samples to simulate the scarcity of defective samples in the natural industrial environment. The test set includes 80 normal images and 120 defective images for a more comprehensive evaluation.

[0035] Step (2): Build a defect detection model based on the NCSN network and perform model training. Input the training set data into the network, and set the network noise level to 10. Train by calculating the score loss of the images under different levels of noise. The final loss function is as follows:

[0036]

[0037] Where represents the estimated score function output by model s θ (x), s θ is the defect detection model with the parameters of the NCSN network model being θ; The image obtained by adding different noises to the input image is defined as where x network is the input image, σ i is the noise level, and ∈ represents the standard Gaussian distribution. Approximately, p(x) represents the probability density function of the input image x, then Represents the true score function, that is, the score function of the input image. To sum up, the loss function is the squared Euclidean distance between the score output by the model and the true score.

[0038] Step (3): Build a defect screening network based on the K-means algorithm. Based on the network trained in step (2), the test mode of the network can output the loss Loss of different images at different noise levels. The AUROC curves of the test set images at different noise levels are as Figure 3 shown. To improve the detection efficiency and ensure the robustness of the network, the output losses at noise levels 5, 6, 7, and 8 are selected as the input of the K-means clustering network.

[0039] As Figure 2 shown, the process of the K-means clustering algorithm is as follows:

[0040] 1) The number of clusters in the clustering algorithm is 2, representing the normal image set and the defect image set respectively. A is the number of images in the normal image set, and B is the number of images in the defect image set;

[0041] 2) Input an image F of a nuclear fuel element to be detected. To achieve synchronous detection and discrimination, the number of images for each clustering is 1;

[0042] 3) When the number of a certain category in the two categories is 0, the centroid initialization C(c1, c2) will be performed, where c1 is the centroid of the normal image set and c2 is the centroid of the abnormal image set;

[0043] 4) Centroid initialization process: Input the Loss of the image F to be clustered. Since the Loss of the defect image is greater than that of the normal image, if c1≥Loss, then let c1 = Loss, otherwise no adjustment is made; if c2≤Loss, then let c2 = Loss, otherwise no adjustment is made;

[0044] 5) Calculate the distance between the centroid and the input image. When the number of a certain category in the two categories is 0, the inter-class distance judgment will be performed. Only when the inter-class distance is greater than the threshold λ, the image will be classified into a new category; when the number of both categories > 0, compare the distances of the input image to the two centroids, and divide the image into the category with the smaller distance;

[0045] 6) After dividing the image F into the corresponding category, update the centroid coordinates as follows (taking c1 as an example), where is the centroid cluster of the k + 1 generation, is the centroid cluster of the k generation, n k is the number of samples in the c1 cluster of the k generation, n k = n k-1 + m k-1, α is a forgetting factor representing the memory degree of the historical centroid cluster, and its value range is [0, 1]. represents the centroid of the new image in the current generation, m k is the number of samples added to cluster c1 in the new image h.

[0046]

[0047] 7) Repeat steps 5)-6) until the distance between the centroid and the image is less than the threshold θ;

[0048] 8) Repeat steps 2)-7) until all image clustering is completed.

[0049] Step (4): For the defective images screened out in step (3), use the NCSN network to reconstruct the images, obtain the reconstructed images with defects removed based on the defective images, calculate the residual map of the two images, and use Otsu threshold segmentation to locate the defect positions, as Figure 4 shown.

[0050] In the embodiments of the present invention, the area under the receiver operating characteristic curve (ROC) (AUC) is selected as the quantitative evaluation index of the model performance. The ROC curve is an image plotted with the false positive rate (FPR) as the abscissa and the true positive rate (TPR) as the ordinate. The ROC curve depicts in detail the performance of the classification model under various threshold conditions. Specifically, the definitions of FPR and TPR are as follows:

[0051]

[0052] where TP, TN, FN, and FP represent true positive, true negative, false negative, and false positive, respectively.

[0053] The AUROC curve obtained in step (3) is as Figure 3 shown. The AUC values of different curves are shown in Table 1. After comparison and selection, the output losses with noise levels of 5, 6, 7, and 8 are selected as the inputs of the K-means clustering network. Using the test set for testing, the defect recognition accuracy in the final clustering result is 100.00%, and the error is 0. For the screened defective images, use the NCSN network to reconstruct the images, obtain the normal images with defects removed based on the defective images, calculate the residual map of the two images, and use Otsu threshold segmentation to locate the defect positions, as Figure 4 shown.

[0054] Table 1 AUC of different noise levels

[0055]

[0056] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A real-time defect detection method for nuclear fuel elements, characterized in that It includes the following steps: S1. Input the image of the nuclear fuel element to be detected into the pre-trained defect detection model based on the NCSN network, and calculate the NCSN multi-level loss. The NCSN multi-level loss is the score loss of the nuclear fuel element image under different levels of noise in the NCSN network; S2. Use the defect screening network based on the K-means algorithm to screen out the defect images from the images of the nuclear fuel elements to be detected according to the NCSN multi-level loss; S3. Use the NCSN network to reconstruct the defect images, obtain the reconstructed images with defects removed based on the defect images, calculate the residual images of the defect images and the reconstructed images, and use Otsu threshold segmentation to locate the defect positions; Among them, the defect detection model based on the NCSN network is trained by calculating the score loss of the nuclear fuel element images under different levels of noise; the loss function used is: Among them, the first item is the estimated score function output by the model s θ (x), the second item is the true score function, and the loss function is the squared Euclidean distance between the score output by the model and the true score; In the process of the K-means algorithm in step S2, the number of clustering categories is set to 2, representing the normal image set and the defect image set respectively, and the number of nuclear fuel element images to be detected for each clustering is set to 1; when the number of a certain category in the two categories is 0, the centroid initialization C(c1, c2) is performed, where c1 is the centroid of the normal image set and c2 is the centroid of the abnormal image set; the centroid initialization process includes inputting the loss value Loss of the nuclear fuel element image to be clustered. If c1≥Loss, then let c1 = Loss, otherwise no adjustment is made. If c2≤Loss, then let c2 = Loss, otherwise no adjustment is made; The K-means algorithm process also includes calculating the distance between the centroid and the input image. When the number of a certain category in the two categories is 0, the inter-class distance is judged. When the inter-class distance is greater than the threshold λ, the input image will be classified into a new category; when the number of both categories is >0, compare the distances of the input image from the two centroids, and divide the image into the category with a smaller distance; The K-means algorithm process also includes updating the centroid coordinates after classifying the new image; 2. The real-time defect detection method for nuclear fuel elements according to claim 1, characterized in that, The network noise level of the defect detection model based on the NCSN network is a total of 10 levels; 3. The real-time defect detection method for nuclear fuel elements according to claim 2, characterized in that, In step S2, the losses with noise levels of 5, 6, 7, and 8 are selected from the NCSN multi-level loss as the input of the defect screening network.

Images