A nuclear fuel assembly spect design and self-absorption attenuation correction method

Abstract

This invention relates to the field of imaging and detection technology for nuclear fuel assemblies, and discloses a SPECT design and self-absorption decay correction method for nuclear fuel assemblies, comprising the following steps: constructing a gamma-ray imaging system for nuclear fuel assemblies; based on projection data, performing image reconstruction using convex set projection and an adaptive steepest descent algorithm to obtain an activity distribution image of a specific nuclide; obtaining spatial resolution and sensitivity based on the reconstructed image; using a non-dominated sorting genetic algorithm to perform multi-objective global optimization of the geometric parameters of the imaging system, with the optimization objective being to simultaneously minimize spatial resolution and maximize sensitivity; updating the imaging system according to the optimized geometric parameters to obtain the final reconstructed image; constructing a self-absorption decay correction model based on a deep learning algorithm, and outputting a predicted corrected activity distribution image through the self-absorption decay correction model to eliminate the absorption effect of the nuclear fuel assembly's own material on gamma rays, thereby improving computational efficiency and enhancing correction accuracy.

Description

Technical Field

[0001] This invention relates to the field of imaging and detection technology for nuclear fuel assemblies, and particularly to a method for designing and correcting the self-absorption decay of nuclear fuel assemblies using SPECT. Background Technology

[0002] Nuclear fuel, as the energy source of a nuclear power plant, is located at the center of the reactor core and operates under high temperature, high pressure, and strong radiation conditions. During operation, it is frequently affected by numerous factors such as water flow impact, foreign objects, vibration, corrosion, heat transfer, and radiation, which can easily lead to core swelling, distortion, or even cladding rupture, causing major safety accidents. Therefore, imaging and inspection technologies for nuclear fuel assemblies have received considerable attention from researchers to ensure the safe operation of the reactor.

[0003] Currently, most operating nuclear power plants worldwide utilize light water reactor technology, with pressurized water reactors (PWRs) and boiling water reactors (BWRs) being the mainstream types. The nuclear fuel assemblies produced are also primarily PWR and BWR fuel assemblies. PWR fuel assemblies typically consist of 14 x 14 to 18 x 18 fuel rods assembled at specific intervals. In the reactor, the decay energies of different fission product nuclides release gamma rays of varying energies. Imaging gamma rays of specific energies allows for the spatial distribution of the radioactivity concentration of specific nuclides, thereby providing images of the nuclear fuel assembly's integrity, burnup, and power distribution.

[0004] Gamma-ray emission computed tomography (GECT) is a key tool for the non-destructive testing of nuclear fuel assemblies, and it is now widely used in nuclear fuel assembly imaging systems. The University of Helsinki in Finland uses a passive gamma-ray emission computed tomography (PGET) system to inspect the integrity of spent fuel assemblies and verify for missing fuel rods; this system has been approved by the IAEA for on-site verification. The Idaho National Laboratory in the United States applies the underwater fuel tomography analysis (FIESTA) system to fuel performance research, performing non-destructive imaging of irradiated fuel to study its structural changes and fission product migration. Uppsala University in Sweden uses high-resolution detectors for gamma-ray emission computed tomography to actually measure fuel assemblies within reactors and analyze the distribution of fission products in fuel rods with different burnout levels. To improve the quality of reconstructed images, optimized imaging system design and correction for self-absorption effects in nuclear fuel assemblies are crucial.

[0005] The collimator and detector parameters of a nuclear fuel assembly imaging system affect the accuracy of the reconstructed image. Traditional methods for optimizing imaging system parameters rely on enumeration. Enumeration is a direct and systematic numerical experimental method that defines a collimator aperture that includes all possible collimator apertures for optimizing imaging system parameters. d Collimator length l Distance from the source hThe discrete parameter space is combined with the detector thickness, and then the reconstructed image under each combination condition is simulated one by one to reflect the spatial resolution and sensitivity. Finally, the optimal solution is found by comparison. However, when the number of optimization parameters increases, the parameter combinations grow exponentially, making the calculation difficult and dependent on the choice of the initial range. The enumeration method can also establish an analytical mathematical model of spatial resolution and sensitivity with imaging system parameters through formulas: the formula for spatial resolution is... The formula for sensitivity is ,in t For the collimator wall thickness, w Let be the length of the collimator's intermediate channel. Then, mathematical tools such as differentiation and extrema are used to directly find the theoretical optimal solution or optimal relationship between spatial resolution and sensitivity. However, this relies on model simplification, neglecting complex factors such as scattering, source size, and detector response, which may lead to results deviating from reality and making it difficult to handle too many variables and constraints. Therefore, a multi-objective parameter global optimization method is needed for the optimal design of the imaging system.

[0006] In gamma-ray emission computed tomography (GECT), the radiation intensity measured by the detector is not a direct reflection of the activity of radionuclides, but rather data attenuated by the fuel matrix material. Compared to medical imaging, the high-density fissile material UO2 core and zirconium alloy cladding within nuclear fuel assemblies exhibit a stronger self-absorption effect on the emitted gamma rays. Without correction, the measurement results will systematically be lower than the actual nuclide concentration or total fission products, causing distortion of the energy spectrum shape and changes in characteristic peaks, ultimately leading to a discrepancy between the reconstructed source region intensity spatial distribution and reality. Therefore, self-absorption attenuation correction is a crucial approach to ensuring the accuracy of detection data and improving the quality of reconstructed images.

[0007] Traditional self-absorption attenuation correction methods include CT attenuation correction, transmission source attenuation correction, magnetic resonance (MR) attenuation correction, and the maximum likelihood attenuation and activity joint reconstruction algorithm. CT attenuation correction is commonly used in PET / CT and SPECT / CT equipment. It reconstructs an attenuated image from the projection data obtained from CT scans and corrects the attenuation based on this image. However, it cannot monitor data in real time, is significantly affected by high-density materials, and is prone to errors. Transmission source attenuation correction uses an external radiation source and detector to obtain the photon flux passing through the material and reconstructs an attenuated image. This method has long scan times, high statistical noise, low spatial resolution, and is susceptible to the influence of its own radiation source. MR attenuation correction obtains the attenuated image from MRI images; artifacts from metallic materials can cause severe image distortion. CT, transmission source, and MR attenuation correction all require additional detectors or radiation sources, resulting in high costs. The maximum likelihood decay and activity joint reconstruction algorithm treats the decay coefficient as an unknown variable to be estimated during ML-EM iterative reconstruction, updating it iteratively along with the activity distribution. However, its results are easily affected by initial values, constraints, and data noise, exhibiting poor convergence, computational complexity, and limited accuracy. It can only be used in regions with relatively uniform decay or simple structures. Therefore, a self-absorption decay correction method is needed that can directly learn complex nonlinear mappings from data, adapt to various material properties, and balance computational efficiency with correction accuracy. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention provides a method for SPECT design and self-absorption decay correction of nuclear fuel assemblies, which can improve computational efficiency and enhance correction accuracy.

[0009] The first aspect of this invention provides a method for SPECT design and self-absorption decay correction of nuclear fuel assemblies, comprising the following steps: A gamma-ray imaging system for nuclear fuel assemblies was constructed. The system includes multiple collimators arranged around the nuclear fuel assembly to be tested, as well as a corresponding array of detectors. The projection data of each detector was acquired through the imaging system. Based on the projection data, the image reconstruction is performed using convex set projection plus adaptive steepest descent algorithm to obtain the activity distribution image of a specific nuclide; Spatial resolution and sensitivity are obtained based on reconstructed images; A non-dominated sorting genetic algorithm is used to perform multi-objective global optimization of the following geometric parameters of the imaging system: collimator aperture. d Collimator length l Distance between collimator and radiation source h In addition to the detector thickness, the optimization objective is to simultaneously minimize spatial resolution and maximize sensitivity; The imaging system is updated based on the optimized geometric parameters to obtain the final reconstructed image; A self-absorption decay correction model is constructed based on a deep learning algorithm. The final reconstructed image and its corresponding unaffected activity distribution image are used as the training set to train the self-absorption decay correction model. The corrected activity distribution image is predicted by the output of the self-absorption decay correction model to eliminate the absorption effect of gamma rays by the nuclear fuel assembly material itself.

[0010] Optionally, there are nine groups of collimators, each group containing multiple collimating holes, and the width of each collimator group is greater than the diameter of the area to be measured. The shielding structure of the collimator has a central channel, which enables the detector to receive gamma rays from adjacent collimating holes. The detector array includes a zinc zinc cadmium array detector located behind the multi-hole collimator, and a bismuth germanate array detector located behind the zinc zinc cadmium array detector.

[0011] Optionally, image reconstruction is achieved by alternately performing the following steps: Data fidelity update: By using convex set projection, the predicted projection values ​​of the reconstructed image are made closer to the actual measured values; Image feature constraints: Prior constraints are imposed on the image through total variation minimization to suppress noise and maintain sharp edges; Alternately perform data fidelity updates and image feature constraints until the reconstructed image converges.

[0012] Optionally, spatial resolution is obtained through the following steps: Set the gamma-ray source as a line source and obtain its reconstructed image through Monte Carlo simulation and image reconstruction process; draw intensity profile lines along the radial direction of the reconstructed line source; after normalizing the intensity of the profile lines, find two points with an intensity value of 0.5, and the distance between them is the half-width at half-height (WHM), which is the spatial resolution. Sensitivity is obtained through the following steps: The total count is obtained by summing the activity values ​​of all pixels in the reconstructed image. N Sensitivity is obtained by total count, detection time, and line source activity.

[0013] Optionally, the self-absorption attenuation correction model adopts a generative adversarial network (GAN) framework, which includes a generator network and a discriminator network. The generator network takes the final reconstructed image as input and outputs the predicted corrected image. The discriminator network is used to distinguish the image generated by the generator from the real, attenuated reference image. Through adversarial training between the generator and the discriminator, the generator is forced to output an image that is infinitely close to the real attenuated distribution. The generator network adopts a U-Net structure that introduces attention gating mechanism, residual connections and context-aware convolutional layers.

[0014] Optionally, the activity distribution image is used to extract the integrity distribution, burnup distribution, and power distribution of the nuclear fuel assembly.

[0015] Optionally, the non-dominated sorting genetic algorithm outputs a Pareto optimal solution set through an iterative process of initializing the population, evaluating performance and non-dominated sorting, selection, crossover and mutation, merging the population and retaining elites.

[0016] Optionally, the multi-objective optimization algorithm is the NSGA-II algorithm. The performance evaluation index of the imaging system is obtained by analyzing the simulated reconstructed images. The gamma rays originate from specific-energy gamma rays released by the decay of fission products in nuclear fuel assemblies. The energies of the specific-energy gamma rays include 0.514 MeV, 1.274 MeV, and 1.596 MeV, corresponding to the collimator aperture. d The value range is 0.05cm-0.5cm, and the collimator length is... l The value ranges from 10cm to 100cm, representing the distance between the collimator and the radiation source. h The value ranges from 60cm to 150cm, and the value range for detector thickness is from 1cm to 10cm.

[0017] A second aspect of the present invention provides a nuclear fuel assembly SPECT imaging system for implementing the above-described method, comprising a detection chamber and multiple imaging units, wherein the detection chamber is used to house the nuclear fuel assembly to be tested, and the multiple imaging units are arranged around the detection chamber; each imaging unit includes: A collimator is used to define the incident direction of gamma rays; A zinc-cadmium detector array is positioned behind the collimator to acquire gamma-ray projection data.

[0018] Optionally, the collimator is made of lead, and the shielding structure of the collimator has a central channel. There are nine imaging units, each of which includes 200 collimating holes. The detector array includes a zinc zinc cadmium array detector located behind the multi-hole collimator, and a bismuth germanate array detector located behind the zinc zinc cadmium array detector.

[0019] The technical solution provided by the embodiments of the present invention has the following advantages compared with the prior art: This invention provides a SPECT design and self-absorption decay correction method for nuclear fuel assemblies. It uses specific-energy gamma rays released from the decay of different fission products within the nuclear fuel assembly for imaging, enabling simultaneous measurement of all fuel rods in the assembly. This method eliminates the need for an external excitation source, utilizing gamma rays released from within the source region to achieve multi-parameter tomographic imaging. The measurement results are less affected by external influences, resulting in a high signal-to-noise ratio. Image reconstruction is performed using convex set projection combined with an adaptive steepest descent algorithm to extract the optimal source activity distribution image from the projection data. Based on a non-dominated sorting genetic algorithm, the spatial resolution and sensitivity of the imaging system reflected in the reconstructed image are globally optimized using multiple parameters such as collimator aperture, length, distance from the radiation source, and detector thickness. The obtained parameters represent the optimal solution, thus updating the imaging system to obtain the final reconstructed image. This method balances spatial resolution and sensitivity, improving spatial resolution by 50% and sensitivity by 60% compared to traditional enumeration methods. Training a self-absorption decay correction model using the final reconstructed image eliminates the absorption effect of gamma rays by the nuclear fuel assembly's own materials. By combining global optimization of the imaging system with self-absorption attenuation correction, and using the reconstructed image of the optimized system for self-absorption attenuation correction, the correction accuracy is improved. Attached Figure Description

[0020] Figure 1 A flowchart of a method for SPECT design and self-absorption decay correction of nuclear fuel assemblies provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the overall structure of the nuclear fuel assembly multi-parameter detection energy spectrum imaging system provided in an embodiment of the present invention; Figure 3 A schematic diagram of a collimator provided in an embodiment of the present invention; Figure 4 The flowchart of the non-dominated sorting genetic algorithm (NSGA-II) used in the parameter optimization of the imaging system provided in the embodiments of the present invention is shown below. Figure 5 A flowchart of a self-absorption attenuation correction method based on a deep neural network is provided for an embodiment of the present invention; Figure 6 The diagram shows the results of reconstructing three different energy gamma-ray sources according to an embodiment of the present invention, where a is the shape diagram of the established gamma-ray source, b is the reconstruction result of the 0.514 MeV energy gamma-ray source, c is the reconstruction result of the 1.274 MeV energy gamma-ray source, and d is the reconstruction result of the 1.596 MeV energy gamma-ray source. Figure 7The diagram shows the results of reconstructing three different gamma-ray sources according to an embodiment of the present invention, where a is the source image of the letter E, b is the reconstructed image of the letter E, c is the source image of the letter X, d is the reconstructed image of the letter X, c is the source image of the letter E, and d is the reconstructed image of the letter E.

[0021] Explanation of reference numerals in the attached figures: 1. Nuclear fuel assembly; 2. Collimator; 3. Cadmium zinc telluride array detector; 4. Bismuth germanate array detector. Detailed Implementation

[0022] The following detailed description of a specific embodiment of the present invention is provided in conjunction with the accompanying drawings. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiment.

[0023] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0024] The present invention will be described below through several specific embodiments. To keep the following description of the embodiments clear and concise, detailed descriptions of known functions and components may be omitted. When any component of an embodiment of the present invention appears in more than one drawing, the component may be represented by the same reference numerals in each drawing.

[0025] like Figure 1 As shown, the first part of the embodiments of the present invention provides a method for SPECT design and self-absorption decay correction of nuclear fuel assemblies, including the following steps: A gamma-ray imaging system for nuclear fuel assemblies is constructed. The system includes multiple collimators 2 arranged around the nuclear fuel assembly 1 to be tested, and a corresponding array of detectors. The projection data of each detector is acquired through the imaging system. Based on the projection data, the image reconstruction is performed using convex set projection plus adaptive steepest descent algorithm to obtain the activity distribution image of a specific nuclide; Spatial resolution and sensitivity are obtained based on reconstructed images; A non-dominated sorting genetic algorithm is used to perform multi-objective global optimization of the following geometric parameters of the imaging system: collimator aperture. d Collimator length lDistance between collimator and radiation source h In addition to the detector thickness, the optimization objective is to simultaneously minimize spatial resolution and maximize sensitivity; The imaging system is updated based on the optimized geometric parameters to obtain the final reconstructed image; A self-absorption decay correction model is constructed based on a deep learning algorithm. The final reconstructed image and its corresponding unaffected activity distribution image are used as the training set to train the self-absorption decay correction model. The corrected activity distribution image is predicted by the output of the self-absorption decay correction model to eliminate the absorption effect of gamma rays by the nuclear fuel assembly material itself.

[0026] This invention provides a SPECT design and self-absorption decay correction method for nuclear fuel assemblies. It uses specific-energy gamma rays released from the decay of different fission products within the nuclear fuel assembly for imaging, enabling simultaneous measurement of all fuel rods in the assembly. This method eliminates the need for an external excitation source, utilizing gamma rays released from within the source region to achieve multi-parameter tomographic imaging. The measurement results are less affected by external influences, resulting in a high signal-to-noise ratio. Image reconstruction is performed using convex set projection combined with an adaptive steepest descent algorithm to extract the optimal source activity distribution image from the projection data. Based on a non-dominated sorting genetic algorithm, the spatial resolution and sensitivity of the imaging system reflected in the reconstructed image are globally optimized using multiple parameters such as collimator aperture, length, distance from the radiation source, and detector thickness. The obtained parameters represent the optimal solution, thus updating the imaging system to obtain the final reconstructed image. This method balances spatial resolution and sensitivity, improving spatial resolution by 50% and sensitivity by 60% compared to traditional enumeration methods. Training a self-absorption decay correction model using the final reconstructed image eliminates the absorption effect of gamma rays by the nuclear fuel assembly's own materials. By combining global optimization of the imaging system with self-absorption attenuation correction, and using the reconstructed image of the optimized system for self-absorption attenuation correction, the correction accuracy is improved.

[0027] like Figure 2 and Figure 3 As shown, in this embodiment of the invention, there are nine groups of collimators 2, each group containing multiple collimating holes, and the width of each group of collimators 2 is greater than the diameter of the area to be measured. The shielding structure of the collimator 2 is provided with a central channel, so that the detector can receive gamma rays from adjacent collimating holes. The detector array includes a zinc zinc cadmium array detector 3 disposed behind the multi-hole collimator 2, and a bismuth germanate array detector 4 disposed behind the zinc zinc cadmium array detector 3.

[0028] The nuclear fuel assembly 1 under test is immersed in water and placed at the center of the imaging system. Nine sets of porous collimators 2 are placed around the fuel assembly 1, with the included angle between adjacent collimators 2 being 40 degrees, to limit the incident direction of gamma rays. A cadmium zinc telluride array detector 3 is located behind the collimators 2 and is used to detect the gamma ray energy spectrum. A bismuth germanate array detector 4 is located behind the cadmium zinc telluride array detector 3 and is used to absorb the Compton scattered gamma rays from the cadmium zinc telluride array detector 3, reducing the Compton scattering background and improving the accuracy of the energy spectrum measurement. The nuclear fuel assembly 1 under test in this invention adopts a 17×17 array arrangement, containing a total of 289 fuel rods. Each fuel rod consists of an outer zirconium alloy cladding, an inner uranium dioxide fuel pellet, and a helium gas gap between them. Gamma rays generated by the decay of specific nuclides in the fuel rods are emitted from the inside of the assembly after passing through different material layers.

[0029] The specific steps are as follows: Gamma ray generation and emission: Gamma rays are generated by solid fission products inside the fuel rod or by gaseous fission products released due to cladding failure. These rays penetrate each material layer of the assembly and then exit into the collimation hole of a high-sensitivity porous collimator with added channels in the middle.

[0030] Gamma-ray detection and scattering suppression: Using an improved collimator and detector system with nine directions, gamma-ray projection data around the nuclear fuel assembly from multiple angles are collected to obtain information such as flux and energy spectrum; the Compton scattered gamma rays generated during the detection process are absorbed by the rear bismuth germanate array detector, thereby effectively suppressing background interference.

[0031] Optionally, image reconstruction is achieved by alternately performing the following steps: Data fidelity update: By using convex set projection, the predicted projection values ​​of the reconstructed image are made closer to the actual measured values; Image feature constraints: Prior constraints are imposed on the image through total variation minimization to suppress noise and maintain sharp edges; Alternately perform data fidelity updates and image feature constraints until the reconstructed image converges.

[0032] Specifically, the following steps are included: 3.1 Image Reconstruction Algorithm Principles and Implementation Image reconstruction is performed using convex set projection combined with an adaptive steepest descent algorithm. This algorithm iteratively executes the following two core steps alternately to extract the optimal source activity distribution image from the projected data:

[0033] Step 1: Data Fidelity Update (Convex Set Projection) This step aims to make the predicted projection values ​​of the reconstructed image as close as possible to the actual measured values. The update formula is:

[0034] , in, It is an image x The Middle l The second iteration b The value of each pixel. It is the first a Projected data values ​​of each detector It is the first b The pixel is the first a The contribution coefficient of each detector, It is the relaxation factor, contribution coefficient. That is, gamma rays are produced by the first b The incident pixel is the first a The probability of a detector is calculated using the following formula: ,in, For gamma rays from the first b The incident pixel is the first a The area that a single detector can cover. For the first b From the 1st pixel to the 2nd pixel a The distance between the detectors. Relaxation factor. That is, the parameter that controls the step size update in each iteration process. This invention combines the number of iterations with the relaxation factor. The value is set to 0.01.

[0035] Step 2: Image characteristic constraints (total variation minimization, adaptive steepest descent) This step, while ensuring data consistency, imposes prior constraints on the image to suppress noise, remove artifacts, and maintain sharp object edges. Its goal is to minimize the total variation of the image.

[0036] , in, It is an image x Total variation is the sum of the gradient magnitudes between all adjacent pixels in an image. It is in the image ( i , j The value of the position. Minimizing the total variation tends to produce an image with piecewise constant characteristics, i.e., containing large areas of uniformity and sharp edges, which is very consistent with the structural characteristics of nuclear fuel rod arrays.

[0037] 3.2 Iterative Reconstruction and Results The above steps one and two are performed iteratively, alternating, until the reconstructed image converges. Finally, the value of each pixel in the image is... This represents the gamma-ray source activity of the spatial region corresponding to that pixel. From this, an activity distribution image of a specific nuclide can be obtained.

[0038] 3.3 Status Information Extraction Based on the reconstructed activity distribution images of specific nuclides, the distribution maps of key parameters reflecting the internal state of nuclear fuel assemblies can be further calculated: Integrity distribution: Obtained directly from the distribution image of a specific element. Areas of abnormal activity in the image indicate possible shell damage and leakage of fission products.

[0039] Fuel consumption distribution: Calculated using the following formula: , in, For fuel consumption, For pixel activity, The decay constant is The average energy produced in each fission. The average yield of fission products. The mass of the fuel rod. The decay constant of each element is fixed, and the average fission energy and average yield can be obtained from the database based on the fission process of the specific element.

[0040] Power distribution: Calculated using the following formula: , in, For linear power values, This represents the length of the fuel rod.

[0041] Optionally, spatial resolution is obtained through the following steps: Set the gamma-ray source as a line source and obtain its reconstructed image through Monte Carlo simulation and image reconstruction process; draw intensity profile lines along the radial direction of the reconstructed line source; after normalizing the intensity of the profile lines, find two points with an intensity value of 0.5, and the distance between them is the half-width at half-height (WHM), which is the spatial resolution. Sensitivity is obtained through the following steps: The total count is obtained by summing the activity values ​​of all pixels in the reconstructed image. N Sensitivity is obtained by total count, detection time, and line source activity.

[0042] The specific steps for multi-parameter global optimization of the imaging system are as follows: This step aims to collaboratively optimize the key geometric parameters of the imaging system using a multi-objective optimization algorithm to obtain the optimal system configuration that balances high spatial resolution and high sensitivity. This invention combines deep residual neural networks with Monte Carlo simulation to establish a self-absorption attenuation correction model, which can effectively predict the attenuation-corrected image and reduce reconstruction errors. Compared with traditional CT correction methods, the error in the reconstructed image is reduced by 20%.

[0043] 4.1 Definition and Acquisition Methods of System Performance Evaluation Indicators To quantitatively evaluate the performance of the imaging system under different parameter combinations, the following two core indicators are defined and obtained through simulation and reconstruction processes: Spatial resolution: Represented by the full width at half maximum (FWHM) of the reconstructed line source image. The smaller the value, the stronger the system's ability to resolve details.

[0044] How to obtain: a. Set the gamma-ray source as a line source and obtain its reconstructed image through Monte Carlo simulation and image reconstruction. b. Draw an intensity profile line along the radial direction of the reconstructed line source. c. After normalizing the intensity of the profile line, find two points with an intensity value of 0.5; the distance between them is the half-width at half-maximum (FWHM).

[0045] Sensitivity: expressed as the total number of detector counts generated per unit of radioactive source activity per unit time. The higher the value, the higher the system detection efficiency.

[0046] Method of obtaining the count: a. Sum the activity values ​​of all pixels in the reconstructed image to obtain the total count. N b. Sensitivity S The calculation formula is:

[0047] , in, For the detection time, For line source activity.

[0048] Optionally, the self-absorption attenuation correction model adopts a generative adversarial network (GAN) framework, which includes a generator network and a discriminator network. The generator network takes the final reconstructed image as input and outputs the predicted corrected image. The discriminator network is used to distinguish the image generated by the generator from the real, attenuated reference image. Through adversarial training between the generator and the discriminator, the generator is forced to output an image that is infinitely close to the real attenuated distribution. The generator network adopts a U-Net structure that introduces attention gating mechanism, residual connections and context-aware convolutional layers.

[0049] like Figure 5 As shown, the specific steps for gamma-ray self-absorption attenuation correction are as follows: This step aims to construct a deep learning-based self-absorption attenuation correction method to directly predict a corrected, high-precision activity distribution image from an uncorrected SPECT reconstructed image, thereby eliminating the strong absorption effect of gamma rays by the nuclear fuel assembly's own materials.

[0050] 5.1 Overall Framework of the Correction Method Generative Adversarial Networks (GANs) are used as the basic framework. The core idea of ​​this method is:

[0051] Train a generator network whose input is a reconstructed image without attenuation correction and whose output is a predicted corrected image.

[0052] Train a discriminator network whose task is to distinguish between images generated by the generator and real, non-degradable reference images (i.e., "ground values").

[0053] By adversarial training between the generator and the discriminator, the generator is forced to output an image that is infinitely close to the real attenuation-free distribution, thereby achieving high-precision attenuation correction.

[0054] 5.2 Preparation of Training Data The generator employs an enhanced U-Net architecture specifically designed for accurate image-to-image conversion. Its main design and innovations include: Attention gating mechanism: An attention gate is introduced into the skip connections between the encoder and decoder. This module uses the high-level semantic features of the decoder as the query vector to perform weighted filtering on the low-level detail features of the encoder, enabling the network to focus on key structural regions such as the edges of fuel rods and suppress irrelevant background interference.

[0055] Residual connections: Introducing residual connections between network layers allows the output of the previous layer to be directly added to the output of the current layer. This effectively alleviates the vanishing gradient problem in deep networks, accelerates training convergence, and improves feature transfer efficiency.

[0056] Context-Aware Convolutional (CAC) Layers: In the decoder, context-aware convolutional layers replace standard convolutional layers. CAC layers can dynamically perceive and fuse global contextual information of the image, and adjust the weights of the convolutional kernels by calculating channel and spatial attention maps, enabling the network to have stronger feature representation and adaptability while being computationally efficient.

[0057] Specific parameters: The network input and output dimensions are both 128×128 pixels. The encoder contains 5 layers and performs 4 downsampling operations; the decoder contains 4 layers and performs 4 upsampling operations. The initial number of convolutional channels is 16, which varies layer by layer along the network depth.

[0058] 5.4 Network Training and Optimization Combat training process: The generator receives an uncorrected image and outputs a corrected image.

[0059] The discriminator simultaneously receives the corrected image and the real, non-attenuated image output by the generator and attempts to distinguish between them.

[0060] The training objective is to minimize the adversarial loss of the generator, that is, to make the generator "deceive" the discriminator so that it cannot distinguish between the generated image and the real image.

[0061] Optimize configuration: The Adam optimizer is used to update the network parameters.

[0062] The loss function is based on adversarial loss and can be combined with pixel-level losses such as L1 or L2 distance to stabilize training.

[0063] Set different learning rates for different network layers or training stages to optimize the convergence process.

[0064] Convergence criterion: Iterative training continues until the discriminator can no longer effectively distinguish between the corrected image output by the generator and the real unattenuated image. At this point, the generator is considered to have fully learned the correction mapping relationship of self-absorption attenuation.

[0065] Optionally, the activity distribution image is used to extract the integrity distribution, burnup distribution, and power distribution of the nuclear fuel assembly.

[0066] Optionally, the non-dominated sorting genetic algorithm outputs a Pareto optimal solution set through an iterative process of initializing the population, evaluating performance and non-dominated sorting, selection, crossover and mutation, merging the population and retaining elites.

[0067] like Figure 4 As shown, the specific execution flow of the NSGA-II algorithm is as follows. The implementation of the algorithm specifically includes the following steps: Initialization: Within a preset parameter range, randomly generate a parent population P containing 100 sets of parameter combinations. t .

[0068] Performance evaluation and non-dominated ranking: for P t For each set of parameters, execute the process in Section 4.1 (MCNP modeling simulation → ASD-POCS image reconstruction → calculation of FWHM and sensitivity) to obtain its corresponding spatial resolution and sensitivity values.

[0069] Based on these two target values, all individuals in the population are non-dominatedly ranked. Individuals that are not surpassed by other individuals in both targets are ranked better, forming multiple non-dominated fronts.

[0070] Evolution and Population Regeneration: Selection, crossover, and mutation: For the parent population P t Perform genetic operations (crossover probability 0.8, mutation probability 0.05) to generate a progeny population Q of size 100. t .

[0071] Merging and Elite Preservation: The parent population P t With offspring population Q t The populations merged to form a mixed population R of size 200. t.

[0072] For R t Perform non-dominated sorting and sort individuals within the same frontier according to crowding distance (a measure of the sparsity of individuals in the target space), prioritizing the retention of individuals that can maintain population diversity.

[0073] The top 100 individuals were selected to form the new generation population P. t+1 .

[0074] Iterative convergence: Repeat the above two steps until the preset maximum number of iterations (500 times) is reached. The Pareto optimal solution set output by the algorithm is the set of imaging system parameters that achieves the best trade-off between spatial resolution and sensitivity.

[0075] Optionally, the multi-objective optimization algorithm is the NSGA-II algorithm. The performance evaluation index of the imaging system is obtained by analyzing the simulated reconstructed images. The gamma rays originate from specific-energy gamma rays released by the decay of fission products in nuclear fuel assemblies. The energies of the specific-energy gamma rays include 0.514 MeV, 1.274 MeV, and 1.596 MeV, corresponding to the collimator aperture. d The value range is 0.05cm-0.5cm, and the collimator length is... l The value ranges from 10cm to 100cm, representing the distance between the collimator and the radiation source. h The value ranges from 60cm to 150cm, and the value range for detector thickness is from 1cm to 10cm.

[0076] 4.2 Optimization Objectives, Parameters, and Algorithm Framework Optimization Objective: This optimization is a dual-objective optimization, aiming to simultaneously minimize spatial resolution (FWHM) and sensitivity. S maximize.

[0077] Optimization parameters and ranges: For gamma ray energies of 0.514 MeV, 1.274 MeV, and 1.596 MeV, the following key geometric parameters are optimized, and their physical value ranges are determined as follows: Collimator aperture d The value range is 0.05cm-0.5cm; Collimator length l The value range is 10cm-100cm; Distance between collimator and radiation source h The value range is 60cm-150cm; The thickness of the detector ranges from 1cm to 10cm.

[0078] Optimization Algorithm: The second generation of the Non-Dominated Sorting Genetic Algorithm (NSGA-Ⅱ) is used for global multi-objective optimization. The parameters are substituted into the non-dominated sorting genetic algorithm for optimization to obtain the optimal solution with the minimum spatial resolution and the highest sensitivity.

[0079] Through the above method, the present invention achieves a direct, efficient, and high-precision conversion from heavily degraded SPECT images to true activity distribution images without relying on additional CT scans or transmission source measurements, significantly improving the quantitative accuracy of nuclear fuel assembly imaging.

[0080] The second part of the embodiments of the present invention provides a nuclear fuel assembly SPECT imaging system for implementing the above-described method, including a detection cavity and multiple imaging units. The detection cavity is used to house the nuclear fuel assembly 1 to be tested, and the multiple imaging units are arranged around the detection cavity; each imaging unit includes: Collimator 2 is used to define the incident direction of gamma rays; A zinc-cadmium detector array is positioned behind collimator 2 to acquire gamma-ray projection data.

[0081] Optionally, the collimator 2 is made of lead to enhance shielding efficiency. The shielding structure of the collimator 2 includes a central channel. The imaging unit comprises nine groups, each group of collimators 2 including 200 collimating apertures. The detector array includes a cadmium zinc telluride array detector 3 positioned behind the multi-aperture collimator 2, and a bismuth germanate array detector 4 positioned behind the cadmium zinc telluride array detector 3. This invention adds a channel in the central part of the collimator, enabling each detector to receive two gamma rays from adjacent collimating apertures, thus increasing sensitivity.

[0082] The porous collimators of this invention surround the nuclear fuel assembly 1 in nine groups, each with 200 collimating holes. Each collimator 2 has the same width, larger than the diameter of the area to be measured. Lead is selected as the material for the collimators 2. The cadmium zinc telluride (CZN) array detector 3 is placed behind the collimating holes of the collimators 2 to receive gamma rays whose incident direction is limited by the collimating holes and to acquire their flux and energy spectrum information. The bismuth germanate (BGD) array detector 4 is located behind the CZN array detector 3 and is mainly used to absorb Compton-scattered gamma rays from the CZN array detector 3, thereby reducing background interference and improving the accuracy of energy spectrum measurements. The collimating holes, CZN array detector 3, and BGD array detector 4 have the same width, but their lengths are differentiated according to their functions to optimize the absorption and detection efficiency of the target rays.

[0083] This invention optimizes the collimation and detection system. Addressing the problem of limited projection data and incomplete reconstructed image information caused by traditional nuclear fuel assembly gamma-ray imaging systems that only arrange collimators and detectors in a single direction, this invention uses a nine-directional array of collimators and detectors, increasing the number of collimation apertures and detector units in each direction to 200. This enables the acquisition of projection data from multiple angles surrounding the nuclear fuel assembly 1, improving the accuracy of the reconstructed image. Furthermore, this invention improves the structure of the collimator 2 by adding a central channel to the traditional shielding structure, optimizing the gamma-ray path, increasing the number of gamma rays received by the detector unit, and improving the system's detection sensitivity under the same source conditions.

[0084] The device operates as follows: During reactor operation, fission products inside the fuel rods generate gamma rays. These rays exit from the fuel assembly and enter the collimation aperture of the porous collimator, where they are captured by a cadmium zinc telluride array detector, thus acquiring gamma ray flux and energy spectrum information. The Compton scattered gamma rays generated during this process are absorbed by a subsequent bismuth germanate array detector to reduce background interference. Finally, based on the obtained gamma ray data, an image reconstruction algorithm is used to generate an image reflecting the integrity of the fuel assembly, burnup distribution, and power distribution.

[0085] Nuclear fuel assembly 1 is arranged in a 17×17 configuration, containing a total of 289 fuel rods. Each fuel rod consists of an outer zirconium alloy cladding, an inner uranium dioxide fuel pellet, and a helium gas gap. Water is filled between the fuel rods and in the external area of ​​the assembly to provide the necessary gamma-ray detection environment. The porous collimator 2 is made of lead, with an initial length of 20 cm, an aperture of 0.3 cm, a wall thickness of 0.1 cm, and a distance of 60 cm between the radiation source and the collimator front end. Detector 3 is made of cadmium zinc telluride and is 3 cm long; detector 4 is made of bismuth germanate and is 2 cm long. Reducing the aperture, increasing the collimator aperture length, and decreasing the distance between the radiation source and the collimator front end can optimize the spatial resolution of the imaging system, while increasing the aperture, decreasing the collimator aperture length, and decreasing the aperture wall thickness can improve the sensitivity of the imaging system. Changing the detector thickness can optimize the detection efficiency of the imaging system. Therefore, the imaging system parameters need to be optimized to find the most suitable dimensions.

[0086] This invention will 85 Kr(0.514MeV), 154 Eu (1.274 MeV) and 140 The gamma rays generated by the decay of Ba (1.596 MeV) were used to detect the integrity, burnup distribution, and power distribution of nuclear fuel assemblies. 85 Kr is an inert gas fission product with a long half-life of 10.7 years. Its activity distribution is an important way to detect whether fuel rods are damaged and leaking. 154Eu has a strong correlation between its activity and burnup, a long half-life of 8.6 years, produces high-energy gamma rays with strong penetrating power, making it easy to detect and able to reflect the cumulative burnup throughout the entire operating cycle. 140 Ba is produced directly by fission, and its activity reflects the local power at the time of detection. It has a moderate half-life of 12.8 days, which effectively reflects the power distribution in the immediate period before reactor shutdown. It produces high-energy gamma rays with strong penetrating power, making it easy to detect. Figure 6 As shown, this invention uses the ASD-POCS algorithm to reconstruct the image and establish the shape of the gamma-ray source as follows: Figure 6 As shown in Figure a, the reconstruction results of gamma-ray sources with energies of 0.514 MeV, 1.274 MeV, and 1.596 MeV are as follows: Figure 6 As shown in (b), (c), and (d) in the figure. This invention uses Peak Signal-to-Noise Ratio (PSNR) and Normalized Absolute Average Error (NAAD) as evaluation parameters for reconstructed image quality; the higher the PSNR and the lower the NAAD, the better the image quality. All three gamma-ray sources used in this invention can produce good reconstruction results.

[0087] Establish the shape of the gamma-ray source as follows Figure 7 As shown in a, c, and e, the reconstruction result is as follows: Figure 7 As shown in b, d, and f, this invention can reconstruct clear images with high reconstruction accuracy even with gamma-ray sources of various shapes.

[0088] The above inventions are merely a few specific embodiments of the present invention. However, the embodiments of the present invention are not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for SPECT design and self-absorption decay correction of nuclear fuel assemblies, characterized in that, Includes the following steps: A gamma-ray imaging system for nuclear fuel assemblies is constructed. The system includes multiple collimators arranged around the nuclear fuel assembly to be tested, and a corresponding array of detectors. The projection data of each detector is acquired through the imaging system. Based on the projection data, image reconstruction is performed using convex set projection plus adaptive steepest descent algorithm to obtain the activity distribution image of a specific nuclide; Spatial resolution and sensitivity are obtained based on reconstructed images; A non-dominated sorting genetic algorithm is used to perform multi-objective global optimization of the following geometric parameters of the imaging system: collimator aperture. d Collimator length l Distance between collimator and radiation source h In addition to the detector thickness, the optimization objective is to simultaneously minimize spatial resolution and maximize sensitivity; The imaging system is updated based on the optimized geometric parameters to obtain the final reconstructed image; A self-absorption decay correction model is constructed based on a deep learning algorithm. The final reconstructed image and its corresponding unaffected activity distribution image are used as the training set to train the self-absorption decay correction model. The corrected activity distribution image is predicted by the output of the self-absorption decay correction model to eliminate the absorption effect of gamma rays by the nuclear fuel assembly material itself.

2. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 1, characterized in that, The collimator comprises nine groups, each containing multiple collimating holes, and the width of each collimator group is greater than the diameter of the area to be measured. The shielding structure of the collimator has a central channel, enabling the detector to receive gamma rays from adjacent collimating holes. The detector array includes a zinc zinc cadmium array detector located behind the multi-hole collimator, and a bismuth germanate array detector located behind the zinc zinc cadmium array detector.

3. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 1, characterized in that, The image reconstruction is achieved by alternately performing the following steps: Data fidelity update: By using convex set projection, the predicted projection values ​​of the reconstructed image are made closer to the actual measured values; Image feature constraints: Prior constraints are imposed on the image through total variation minimization to suppress noise and maintain sharp edges; Alternately perform data fidelity updates and image feature constraints until the reconstructed image converges.

4. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 1, characterized in that, The spatial resolution is obtained through the following steps: The gamma-ray source is set as a line source, and its reconstructed image is obtained through Monte Carlo simulation and image reconstruction process; an intensity profile line is drawn along the radial direction of the reconstructed line source; after normalizing the intensity of the profile line, two points with an intensity value of 0.5 are found, and the distance between them is the half-width at half-height, which is the spatial resolution; The sensitivity is obtained through the following steps: The total count is obtained by summing the activity values ​​of all pixels in the reconstructed image. N Sensitivity is obtained by total count, detection time, and line source activity.

5. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 1, characterized in that, The self-absorption attenuation correction model employs a generative adversarial network (GAN) framework, which includes a generator network and a discriminator network. The generator network takes the final reconstructed image as input and outputs the predicted corrected image. The discriminator network distinguishes the image generated by the generator from the real, attenuation-free reference image. Through adversarial training between the generator and the discriminator, the generator is forced to output an image that is infinitely close to the real attenuation-free distribution. The generator network adopts a U-Net structure that incorporates attention gating, residual connections, and context-aware convolutional layers.

6. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 1, characterized in that, The activity distribution image is used to extract the integrity distribution, burnup distribution, and power distribution of nuclear fuel assemblies.

7. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 6, characterized in that, The non-dominated sorting genetic algorithm outputs a Pareto optimal solution set through an iterative process of initializing the population, evaluating performance and non-dominated sorting, selection, crossover and mutation, merging populations and retaining elites.

8. The method for SPECT design and self-absorption decay correction of nuclear fuel assemblies as described in claim 7, characterized in that, The multi-objective optimization algorithm is the NSGA-II algorithm. The performance evaluation index of the imaging system is obtained by analyzing simulated reconstructed images. The gamma rays originate from specific-energy gamma rays released by the decay of fission products in nuclear fuel assemblies. The energies of these specific-energy gamma rays include 0.514 MeV, 1.274 MeV, and 1.596 MeV, corresponding to the collimator aperture. d The value range is 0.05cm-0.5cm, and the collimator length is... l The value ranges from 10cm to 100cm, representing the distance between the collimator and the radiation source. h The value ranges from 60cm to 150cm, and the value range for detector thickness is from 1cm to 10cm.

9. A nuclear fuel assembly SPECT imaging system for implementing the method of any one of claims 1-8, characterized in that, It includes a detection chamber and multiple imaging units. The detection chamber is used to house the nuclear fuel assembly to be tested, and the multiple imaging units are arranged around the detection chamber. Each imaging unit includes: A collimator is used to define the incident direction of gamma rays; A zinc-cadmium detector array is positioned behind the collimator to acquire gamma-ray projection data.

10. The nuclear fuel assembly SPECT imaging system according to claim 9, characterized in that, The collimator is made of lead, and the shielding structure of the collimator has a central channel. The imaging unit has nine groups, and each group of collimators includes 200 collimation holes. The detector array includes a zinc zinc cadmium array detector located behind the multi-hole collimator, and a bismuth germanate array detector located behind the zinc zinc cadmium array detector.

Images