A method and system for monitoring fuel assembly air tightness damage based on core parameters

Abstract

The present application relates to the technical field of core monitoring, and particularly relates to a fuel assembly air tightness damage monitoring method and system based on a core parameter. The method is as follows: collecting core physical and thermal parameters and related 3D images of a nuclear reactor core; carrying out noise removal processing on the obtained data; using a convolutional neural network to deeply analyze and predict the core 3D image data; using a time convolution network to deeply analyze and predict the time sequence data of the parameters; carrying out comprehensive risk assessment and automatically generating a corresponding control strategy; visualizing the data set calculation results of the above steps, and issuing a warning according to the risk assessment results. The system comprises: a data and image acquisition module; a noise removal module; an analysis and prediction module; a visualization module; and a warning module. The present application can predict the change of the local parameter distribution of the core in real time through a 3D model, and the system model automatically gives a control scheme for the core power distortion, which is provided to an operator to assist in the core parameter control scheme, so that the fuel cladding, which is the first barrier of a nuclear power plant, is prevented from being damaged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reactor core monitoring, and particularly relates to a fuel assembly airtightness damage monitoring method and system based on reactor core parameters. BACKGROUND

[0002] The rapid change of reactor power will cause the increase of temperature gradient inside the fuel pellet. The fuel assembly will expand due to heat when it is running in the reactor core, and will produce irradiation growth as the burnup deepens. When the reactor power of VVER unit rises from zero to 100% Nnom, the temperature change inside the pellet is about 1200℃. Therefore, when the reactor power exceeds 75% Nnom, it is required to maintain the power level of 75% to 85% Nnom for at least three hours to release the cumulative elastic deformation of the cladding due to the creep of fuel and cladding, so as to avoid the airtightness damage of the fuel assembly. The number of peak shaving times of a VVER nuclear power single unit in a fuel cycle exceeds 100, and after receiving the peak shaving instruction from the power grid dispatching, the power plant quickly reduces the reactor power, and the local parameters of the reactor core will be distorted, which will affect the airtightness of the fuel assembly.

[0003] The fuel rod cladding is the first barrier of the nuclear power plant, and if it is damaged, the radioactive fission products in the fuel rod will be released into the primary coolant. If the radioactivity in the primary coolant of the VVER unit exceeds the operating limit, the unit needs to be shut down and withdrawn. At present, the main diagnosis methods for fuel assembly damage and leakage include sipping method, fission product radioactivity estimation method, etc. Among them, the sipping method checks all fuel assemblies during refueling, which takes a long time and increases the refueling period, affecting the economic efficiency of the power plant; the result error of the fission product radioactivity estimation method is large, and it cannot accurately locate the specific damaged fuel assembly. SUMMARY

[0004] The technical problem to be solved by the present application is to provide a fuel assembly airtightness damage monitoring method and system based on reactor core parameters, which can predict the change of local parameter distribution of the reactor core in real time through a 3D model, and the system model automatically gives a control scheme for the power distortion of the reactor core, which is provided to the operator to assist in the control of the reactor core parameter control scheme, so as to avoid the damage of the fuel cladding as the first barrier of the nuclear power plant and the release of fission gas in the fuel assembly into the primary coolant.

[0005] The present application provides a fuel assembly airtightness damage monitoring method based on reactor core parameters, which comprises the following steps:

[0006] Step 1: Collecting the reactor core physical and thermal parameters of the nuclear reactor core and related 3D images;

[0007] Step 2: The obtained data is processed by noise removal, and the image data is enhanced by contrast adjustment and edge sharpening image enhancement technology to improve the data quality;

[0008] Step three: deep analysis and prediction of the core 3D image data by using convolutional neural network;

[0009] The time convolution network deeply analyzes and predicts the time series data of the parameters;

[0010] The output results of the convolutional neural network are used to identify the spatial abnormal features at different positions; the time convolution network analyzes the trend of the core parameters over time and predicts the abnormal state that may occur in the future;

[0011] Step four: comprehensive risk assessment is carried out by combining the output results of the convolutional neural network and the time convolution network, and the corresponding control strategy is automatically generated;

[0012] Step five: the data set calculation results of the above steps are visualized, and according to the risk assessment results, a warning is issued.

[0013] In one specific embodiment of the present application, in the step one, the core physical and thermal parameters of the core include the parameters in the nuclear measurement system memory and the parameters collected by various instruments at the key positions of the core;

[0014] The 3D image is collected by various instruments at the key positions of the core.

[0015] In one specific embodiment of the present application, the parameters in the nuclear measurement system memory include the core volume power peak factor, the fuel assembly power distribution unevenness factor, the deviation from the bubble nucleate boiling ratio, and the linear power density of the fuel rod.

[0016] In one specific embodiment of the present application, in the step two,

[0017] The denoising processing adopts a low-pass filter to remove high-frequency noise, and the formula is:

[0018] Y(f)=X(f)H(f) (1)

[0019] Wherein, Y(f) is the signal spectrum after filtering, X(f) is the frequency spectrum of the original signal, and H(f) is the frequency response of the filter;

[0020] The contrast adjustment improves the image contrast through histogram equalization,

[0021] The formula of histogram equalization is:

[0022]

[0023] Wherein, r is the pixel intensity of the original image, a is the pixel intensity of the output image, L is the possible brightness level of the image, and pr(w) is the brightness probability distribution function of the original image;

[0024] The edge sharpening adopts a Laplace operator to enhance image edges,

[0025] The basic formula of Laplace sharpening is represented as:

[0026]

[0027] Wherein, g(x,y) is the sharpened image, f(x,y) is the original image, is the Laplace operator of the image.

[0028] In one specific embodiment of the present application, in the step three, the convolutional neural network automatically extracts features in the 3D image, increases nonlinearity through an activation function, enables the model to learn more complex features, and then identifies the extracted 3D image to extract abnormal features in the core,

[0029] The convolution operation formula is:

[0030]

[0031] Wherein, F(x,y,z) represents the output feature value after the convolution operation at position (x,y,z), I(x+i,y+j,z+k) represents the pixel value of the input 3D image data at position (x+i,y+j,z+k), K(i,j,k) is the weight value of the convolution kernel (filter) at position (i,j,k), and (a,b,c) respectively represents half of the size of the convolution kernel in three dimensions;

[0032] The causal convolution formula of the temporal convolution network is represented as:

[0033]

[0034] Wherein, Y t· represents the predicted output value at time t, W i represents the weight of the convolution kernel at time lag i, X t-i. represents the input value at time t-i, and k represents the size of the convolution kernel, i.e. the longest distance that the model can look back.

[0035] In one specific embodiment of the present application, the step four is specifically:

[0036] The output results of the convolutional neural network and the temporal convolution network are combined to calculate a comprehensive risk indicator R,

[0037] The risk indicator can be represented as:

[0038] R=α·F CNN +β·F TCN (6);

[0039] Wherein, R represents the comprehensive risk indicator, F CNN represents the score of the spatial anomaly feature identified by the module, F TCN represents the score of the time series trend predicted by the module, and alpha and beta are used to adjust F CNN and F TCN contribution to the risk assessment;

[0040] According to the size of the risk indicator R, the corresponding control strategy is automatically generated.

[0041] In one specific embodiment of the present application, the control strategy is dynamically adjusted by the following formula to correspond to the current risk situation:

[0042] Adjust the reactor power: P new = P current - gamma P · (R-R threshold ) (7)

[0043] Modify the cooling water flow: Q new = Q current + gamma Q · (R-R threshold ) (8)

[0044] Condition judgment for starting the emergency cooling system:

[0045]

[0046] Wherein, P new and Q new represent the adjusted reactor power and cooling water flow, P current and Q current represent the current reactor power and cooling water flow, R threshold and R ECS respectively represent the risk indicator threshold for triggering adjustment and starting the emergency cooling system, gamma P and gamma Q respectively represent the adjustment coefficient for adjusting the reactor power and the cooling water flow, and Activate ECS. represents an indication variable indicating whether to start the emergency cooling system.

[0047] In one specific embodiment of the present application, in step five,

[0048] When data and calculation results show that there is a risk, the system sends a warning in multiple forms such as sound, light signal and email.

[0049] The present application also provides a fuel assembly air tightness damage monitoring system based on the core parameters, comprising:

[0050] a data and image acquisition module for acquiring core physical and thermal parameters and related 3D images;

[0051] a denoising module for denoising processing of data;

[0052] an analysis and prediction module for deep processing of data using a convolutional neural network model and a temporal convolution network model, and for evaluating a risk level and forming a control strategy;

[0053] a visualization module for displaying the acquired data and calculation results;

[0054] a warning module for providing a warning when a warning is needed in risk evaluation.

[0055] In one specific embodiment of the present application, the analysis and prediction module includes a convolutional neural network model, a temporal convolution network model, and a risk evaluation model.

[0056] Compared with the prior art, the fuel assembly airtightness damage monitoring method and system based on core parameters of the present application optimize the safety monitoring capability of a nuclear power plant reactor. Through real-time analysis of 3D images and parameter time series data of the core, the method uses the spatial analysis capability of a convolutional neural network (CNN) and the time series prediction capability of a temporal convolution network (TCN) to realize accurate identification of abnormal states of the core and effective prediction of future trends. On the one hand, the spatial analysis of the CNN module optimizes the identification of core structural abnormal features such as cracks and melting areas, improving the accuracy of monitoring. On the other hand, the time series analysis of the TCN module enhances the prediction capability of future changes in core parameters, making early identification of potential risks possible. In combination with the analysis results of the two networks, the comprehensive risk evaluation algorithm of the present application can more accurately evaluate the current risk level and automatically generate a corresponding control strategy according to the risk level. By predicting and preventing potential safety risks, the safety and stability of the operation of the nuclear power plant are improved. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 a flowchart representing a fuel assembly airtightness damage monitoring method based on core parameters. DETAILED DESCRIPTION

[0058] In order to further understand the present application, the embodiments of the present application are described below in conjunction with examples, but it should be understood that these descriptions are only for further illustrating the features and advantages of the present application, and are not limitations of the present application.

[0059] The embodiments of the present application disclose a fuel assembly airtightness damage monitoring method based on core parameters, comprising the following steps:

[0060] Step one: collect the core physical and thermal parameters of the nuclear reactor core and the related 3D image;

[0061] Step two: the obtained data is processed by denoising, and the image data is enhanced by contrast adjustment and edge sharpening image enhancement technology to improve the data quality;

[0062] Step three: the 3D image data of the core is analyzed and predicted in depth by using the convolutional neural network;

[0063] The time convolution network analyzes and predicts the time series data of the parameters;

[0064] The output result of the convolutional neural network is used to identify the spatial anomaly characteristics of different positions; the time convolution network analyzes the change trend of the core parameters with time and predicts the abnormal state that may occur in the future;

[0065] Step four: combine the output results of the convolutional neural network and the time convolution network, and perform comprehensive risk assessment, and automatically generate the corresponding control strategy;

[0066] Step five: the data set calculation results of the above steps are visualized, and according to the risk assessment result, an early warning is issued.

[0067] The embodiment of the application also discloses a fuel assembly air tightness damage monitoring system based on the core parameters, comprising:

[0068] A data and image acquisition module is used to collect the core physical and thermal parameters and the related 3D image;

[0069] A denoising module is used for denoising processing of data;

[0070] An analysis and prediction module uses a convolutional neural network model and a time convolution network model to deeply process data, and evaluates the risk level to form a control strategy;

[0071] A visualization module is used to display the collected data and calculation results;

[0072] A warning module is used to provide a warning when a warning is needed in risk assessment.

[0073] The patent application proposes a nuclear reactor core parameter and fuel assembly airtightness real-time monitoring and control method combining convolutional neural network (CNN) and temporal convolution network (TCN), aiming to improve the safety monitoring level of nuclear power plants. By analyzing the 3D image and parameter time series data of the core in real time, this method can accurately predict the change of the core state, timely identify and respond to potential risks, and automatically generate control strategies to prevent fuel assembly airtightness damage and fissile gas leakage. This invention provides scientific risk assessment and control strategies for nuclear power plants, reduces human operation errors, optimizes reactor operation efficiency, and enhances emergency response capability, thereby significantly improving the safety operation efficiency and response capability of nuclear power plants, and has important significance for public and environmental safety.

[0074] In order to further understand the present application, the following embodiments will be described in detail based on the fuel assembly airtightness damage monitoring method and system based on core parameters provided by the present application, and the protection scope of the present application is not limited by the following embodiments.

[0075] Example 1

[0076] As Figure 1 shown, a core parameter and fuel assembly airtightness monitoring method based on a deep learning model includes the following steps:

[0077] Step 1: Collecting core physical and thermal parameters and related 3D images of the nuclear reactor core. The data collection step covers the collection of key physical and thermal parameters and corresponding 3D image data from the VVER unit nuclear reactor core.

[0078] Among them, high-precision sensors are deployed at key positions of the nuclear reactor core, including KNIT2T, KNIT3T and KNITU neutron temperature measurement channels, as well as thermocouples and self-powered detectors. These devices collectively collect physical and thermal parameter data and 3D images of the core for monitoring key information including reactor power field distribution, local core parameters such as deviation from nucleate boiling ratio DNBR and linear power density of fuel elements.

[0079] The KNIT2T type measurement channel is provided with 46, characterized by 7 self-powered detectors distributed equidistantly along the core axis for monitoring neutron flux density (i.e. power density). Two thermocouples are arranged at the outlet of the fuel assembly, and one is arranged at the inlet, which is specially used for monitoring the coolant temperature.

[0080] The KNI T3T type measurement channel has a total of 4, which is similar to the configuration of the KNI T2T type, but additionally arranges one thermocouple under the lower part of the reactor head to provide more comprehensive monitoring data for the coolant temperature.

[0081] The KINITU type measuring channel is also 4, which is characterized by arranging 1 thermocouple thermometer at the outlet of the fuel assembly and the lower part of the reactor head, and additionally arranging 3 thermocouples to monitor the coolant level, and monitoring the coolant state in all directions.

[0082] The self-sufficient detector is distributed with 7 SPNDs along the core height, and after neutron irradiation, the SPND generates a current signal, which is mainly used for monitoring the neutron flux density. In order to eliminate the current delay effect caused by decay, two filter combinations are used for signal processing to ensure that the delay of the processed signal does not exceed 0.05 seconds.

[0083] Through the refinement of the steps, it is ensured that the collected core state data has high quality and high precision, which provides a solid data foundation for subsequent deep learning analysis and real-time monitoring.

[0084] Step 2: The data denoising processing and image enhancement technology aims to provide clear and accurate input data for deep learning analysis. The denoising processing refers to using a low-pass filter algorithm, and the image enhancement technology includes contrast adjustment and edge sharpening technology.

[0085] The collected physical and thermal parameter data often accompanies background noise, which may come from the electronic noise of the sensor itself, environmental interference or vibration of the equipment working, etc. In order to improve the signal-to-noise ratio of the data, a low-pass filter algorithm is used for denoising processing, which mainly limits the high-frequency components in the data to effectively remove the interference of non-signal components. The mathematical formula of denoising processing can be expressed as:

[0086] Y(f) = X(f)H(f) (1)

[0087] Where Y(f) is the filtered signal spectrum, X(f) is the original signal spectrum, and H(f) is the frequency response of the filter.

[0088] The contrast adjustment of the image is realized through the histogram equalization technology, which adjusts the brightness distribution of the image, so that the contrast of the image is enhanced, and the visualization effect of the core structure is more obvious. The mathematical formula of contrast adjustment is:

[0089]

[0090] Where r is the pixel intensity of the original image, s is the pixel intensity of the output image, L is the possible brightness level of the image, and p r (w) is the brightness probability distribution function of the original image.

[0091] The edge sharpening uses Laplace operator to enhance the image edge, so that the structural details of the core are clearer. The basic formula of Laplace sharpening can be expressed as:

[0092]

[0093] where g(x, y) is the sharpened image, f(x, y) is the original image, is the Laplacian operator of the image.

[0094] Through the above refined data acquisition and preprocessing steps, it is ensured that the collected core state data and 3D image data have high quality and high precision.

[0095] Step 3: In order to realize the accurate monitoring and risk assessment of the core state of the nuclear reactor, based on the data obtained in steps 1 and 2, the advanced deep learning technology of convolutional neural network (CNN) and time convolution network (TCN) is used to deeply analyze and predict the 3D image data of the core and the time series data of its parameters.

[0096] For the 3D image data of the core, the CNN architecture is used for processing. This architecture is composed of multiple convolutional layers, pooling layers and fully connected layers, which can effectively identify abnormal features in the core, such as cracks, melting areas, etc. First, the convolutional layer is used to automatically extract the features in the image, and each convolutional layer increases the nonlinearity through an activation function such as ReLU, so that the model can learn more complex features. The convolution operation formula is:

[0097]

[0098] where F(x, y, z) represents the output feature value after convolution operation at position (x, y, z), I(x+i, y+j, z+k) represents the pixel value of the input 3D image data at position (x+i, y+j, z+k), K(i, j, k) is the weight value of the convolution kernel (filter) at position (i, j, k), and (a, b, c) represents half of the size of the convolution kernel in three dimensions, respectively.

[0099] For the time series data of the core parameters, the TCN model is used for in-depth analysis and prediction. TCN captures long-term dependencies in time series data through extended convolution operations. In particular, the causal convolution structure of TCN ensures that the model only uses data up to the current time when making predictions, avoiding the leakage of future information. The main formula can be expressed in the form of time series prediction, where the causal convolution ensures the time sequence property of the model prediction. The causal convolution formula of TCN is represented as:

[0100]

[0101] where Y t· represents the predicted output value at time t, W i represents the weight of the convolution kernel at time lag i. Xt-i. Represents the input value at time ti, and k represents the size of the convolution kernel, that is, the longest distance the model can see into the past.

[0102] Step 4: The CNN output is used to identify spatial anomaly features at different locations. The CNN is focused on identifying spatial anomaly features such as cracks, melted areas, or other critical structural changes from 3D image data collected from the nuclear reactor core.

[0103] The TCN focuses on analyzing the temporal trends of core parameters such as temperature and pressure, and uses this information to predict possible future abnormal conditions. By deeply analyzing historical and current core parameter data, the TCN can identify long-term and short-term patterns and predict future risk trends. This strategy includes actions such as adjusting reactor power, modifying cooling water flow, or activating the emergency cooling system, while also providing specific operational recommendations.

[0104] First, by combining the core spatial anomaly characteristics identified by the CNN model and the core parameter time series data trends predicted by the TCN model, a comprehensive risk index R is calculated. This index comprehensively considers the spatial anomaly characteristics of the core state and the predicted trend of the time series data. The risk index can be expressed as:

[0105] R=α·F CNN +β·F TCN (6)

[0106] Among them, R represents the comprehensive risk index, F CNN represents the score of the spatial anomaly features identified by the module, F TCN Indicates the score of the time series trend predicted by the module, α and β are used to adjust F CNN and F TCN Contribution to risk assessment.

[0107] Next, based on the size of the risk indicator R and the risk level, the corresponding control strategy is automatically generated. These control strategies can be dynamically adjusted to respond to the current risk situation using the following formula:

[0108] 1. Adjust the reactor power:

[0109] P new =P current -γ P ·(RR threshold ) (7)

[0110] 2. Modify the cooling water flow:

[0111] Q new =Q current +γ Q ·(RRthreshold ) (8)

[0112] 3. Condition judgment of starting emergency cooling system:

[0113]

[0114] wherein, P new. and Q new represent the adjusted reactor power and cooling water flow, P current and Q current represent the current reactor power and cooling water flow, R threshold and R ECS represent the risk index threshold values for triggering adjustment and starting emergency cooling system, γ P· and γ Q represent the adjustment coefficients for adjusting reactor power and cooling water flow, Activat,e -ECS. represents the indication variable of whether to start the emergency cooling system.

[0115] Step 5: Through the data and calculation results of the above steps, the core state is visually displayed through heat map and dynamic chart through data visualization technology, and the decision support ability of the operator is enhanced. At the same time, according to the risk assessment result, the system sends early warning through sound, light signal and email and other ways, to ensure that the potential risk can be identified and responded in time.

[0116] The above description of the embodiments is only used to help understand the method of the present application and its core idea. It should be pointed out that for those skilled in the art, without departing from the principles of the present application, some improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0117] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring fuel assembly airtightness damage based on core parameters, characterized in that: The following steps are involved: Step 1: Collect core physical and thermal parameters of the nuclear reactor core and related 3D images; The core physical and thermal parameters of the core include parameters stored in the nuclear measurement system and parameters collected from key parts of the core; The 3D images are collected from key parts of the core; The parameters stored in the nuclear measurement system include core volume power peak factor, fuel assembly power distribution unevenness factor, deviation from nucleate boiling ratio, and fuel rod linear power density; Step 2: The acquired data is subjected to denoising, and the image data is enhanced using contrast adjustment and edge sharpening techniques to improve data quality; Step 3: Use convolutional neural networks to conduct in-depth analysis and prediction of the core 3D image data; Temporal convolutional networks conduct in-depth analysis and prediction of parameter time series data; The output of the convolutional neural network is used to identify spatial anomaly features at different locations; the temporal convolutional network analyzes the changing trends of core parameters over time and predicts possible abnormal conditions in the future; Step 4: Combine the output results of the convolutional neural network and the temporal convolutional network to calculate the comprehensive risk index R. The risk indicator is expressed as: R=α·F CNN +β·F TCN (6); Among them, R represents the comprehensive risk index, F CNN represents the score of the spatial anomaly features identified by the module, F TCN Indicates the score of the time series trend predicted by the module, α and β are used to adjust F CNN and F TCN Contribution to risk assessment; According to the size of the risk index R, the corresponding control strategy is automatically generated; The control strategy is dynamically adjusted to the current risk situation using the following formula: Adjust reactor power: P new =P current -γ P ·(RR threshold )(7); Modify cooling water flow: Q new =Q current +γ Q ·(RR threshold )(8); Conditions for starting the emergency cooling system: Among them, P new and Q new represents the adjusted reactor power and cooling water flow rate, P current and Q current Represents the current reactor power and cooling water flow, R threshold and R ECS represent the risk indicator thresholds for triggering adjustment and starting the emergency cooling system, γ P and γ Q Respectively represent the adjustment coefficients for adjusting the reactor power and cooling water flow, Activate ECS. Indicator variable indicating whether the emergency cooling system is activated; Step 5: Visualize the calculation results of the data set in the above steps and issue an early warning based on the risk assessment results.

2. The method for monitoring fuel assembly airtightness damage based on core parameters according to claim 1, characterized in that: In the step 2, The denoising process uses a low-pass filter to remove high-frequency noise, and the formula is: Y(f)=X(f)H(f) (1) Where Y(f) is the spectrum of the filtered signal, X(f) is the spectrum of the original signal, and H(f) is the frequency response of the filter. The contrast adjustment improves the image contrast by histogram equalization. The formula for histogram equalization is: Where r is the pixel intensity of the original image, a is the pixel intensity of the output image, L is the possible brightness level of the image, and pr(w) is the brightness probability distribution function of the original image; The edge sharpening uses the Laplacian operator to enhance the image edge. The basic formula of Laplace sharpening is expressed as: Among them, g(x,y) is the sharpened image, f(x,y) is the original image, is the Laplacian of the image.

3. The method for monitoring fuel assembly airtightness damage based on core parameters according to claim 1, characterized in that: In step 3, the convolutional neural network automatically extracts features from the 3D image, increases nonlinearity through activation functions, and enables the model to learn more complex features. The extracted 3D image is then identified to extract abnormal features in the core. The convolution operation formula is: Where F(x,y,z) represents the output feature value after the convolution operation at position (x,y,z), I(x+i,y+j,z+k) represents the pixel value of the input 3D image data at position (x+i,y+j,z+k), K(i,j,k) is the weight value of the convolution kernel at position (i,j,k), and (a,b,c) represents half of the size of the convolution kernel in the three dimensions respectively; The causal convolution formula of the temporal convolutional network is expressed as: Among them, Y t. Represents the predicted output value at time t, W i represents the weight of the convolution kernel at time lag i, X t-i. Represents the input value at time ti, and k represents the size of the convolution kernel, that is, the longest distance the model can see into the past.

4. The method for monitoring fuel assembly airtightness damage based on core parameters according to claim 1, characterized in that: In the step five, When data and calculation results indicate a risk, the system issues an early warning via sound, light signals, and email.

5. A monitoring system using the fuel assembly airtightness damage monitoring method based on core parameters according to any one of claims 1 to 4, characterized in that: include: Data and image acquisition module, used to collect core physical and thermal parameters and related 3D images; Denoising module, used for data noise reduction; The analysis and prediction module uses convolutional neural network models and temporal convolutional network models to deeply process data, assess risk levels, and form control strategies; Visualization module, used to display collected data and calculation results; The early warning module is used to provide early warning when a warning is needed during risk assessment.

6. The monitoring system according to claim 5, characterized in that: The analysis and prediction module includes a convolutional neural network model, a temporal convolutional network model and a risk assessment model.

Images