Fuel assembly end plug abnormity intelligent detection method and system

Abstract

The invention belongs to the technical field of fuel assembly end plug abnormity detection, and discloses a fuel assembly end plug abnormity intelligent detection method and system. The method comprises the steps of predicting a feature position in an image, triggering an end plug abnormal detection state, judging whether the center of a tube socket angle feature frame is within the range of a tube socket positioning frame or not, screening the tube socket angle feature frame, fitting a reference straight line according to the center point of the screened tube socket angle feature frame, and solving the slope and intercept of the reference straight line. Judging whether the slope of the reference straight line is within a set range according to the slope of the reference straight line, calculating the pixel distance from the top of each end plug to the reference straight line, and judging whether the whole end plugs move or not. The system comprises a fuel assembly feature positioning module, an upper end plug anomaly detection module and a lower end plug anomaly detection module. According to the invention, fuel assembly end plug abnormity identification can be completed in a high-accuracy and high-speed manner, the working intensity of workers is reduced, and the working efficiency of appearance inspection of four surfaces of the fuel assembly is improved.

Description

Technical Field

[0001] This application belongs to the field of fuel assembly end plug anomaly detection technology, and particularly relates to an intelligent detection method and system for fuel assembly end plug anomalies. Background Technology

[0002] Nuclear fuel assemblies are among the most important structural units in a nuclear reactor, and nuclear fuel rods are the core of these assemblies. End plugs, as structural materials for the nuclear fuel cladding rods, are responsible for sealing the uranium dioxide (UO2) pellets. The quality of the end plug welds directly affects the safe operation of the nuclear reactor; any defects can lead to nuclear leakage, threatening normal human life and the further development of the nuclear industry. Through precise end plug anomaly detection, defective end plugs can be detected, repaired, or replaced in a timely manner, thereby improving the reliability and service life of the nuclear fuel assembly.

[0003] Fuel assembly end plug anomaly detection plays a crucial role in ensuring the safe operation of nuclear reactors, improving detection efficiency, meeting regulatory requirements, preventing nuclear contamination, enhancing the reliability of nuclear fuel assemblies, and supporting process improvements.

[0004] Currently, fuel assembly end plug anomaly detection is typically performed manually, which involves a significant degree of subjectivity. Furthermore, due to limited worker attention, anomalies in the fuel assembly end plugs may not be detected promptly. Using deep learning for intelligent end plug detection can quantify end plug information to a certain extent, allowing for a more objective assessment of the fuel assembly end plug's condition.

[0005] Taking the upper plug of the fuel assembly as an example, in a frontal view from one side, there are typically 17 upper plugs on the outermost side. Anomalies in these 17 upper plugs include the following:

[0006] First, the relative distance between this set of upper plugs as a whole and the upper tube seat does not meet the specified threshold. For example, the overall relative distance does not reach the specified threshold (i.e., the upper plug is too close to the upper tube seat).

[0007] Second, the relative distance between one or more of the upper plugs in this group and the upper tube seat does not meet the specified threshold.

[0008] Third, under normal circumstances, there are 17 end plugs on one side of the fuel assembly. If the number of end plugs detected is insufficient, it indicates that there are missing end plugs.

[0009] Similarly, the lower plug also has similar problems, and the abnormality type detected by the lower plug is the same as that of the upper plug. Summary of the Invention

[0010] The purpose of this application is to overcome the shortcomings of the prior art and provide a method and system for intelligent detection of fuel assembly end plug anomalies, which can complete the identification of fuel assembly end plug anomalies with high accuracy and high speed, reduce the workload of workers and improve the efficiency of four-sided appearance inspection of fuel assemblies.

[0011] To achieve the above objectives, this application provides the following technical solution:

[0012] In a first aspect, this application provides a method for intelligent detection of fuel assembly end plug anomalies, including:

[0013] S10: Receive video of the appearance of the fuel assembly from all four sides and extract images at set intervals;

[0014] S20: Preprocess the extracted image;

[0015] S30: Input the preprocessed image and predict the location of features in the image;

[0016] S40: When the predicted feature is a pipe seat, determine whether the y coordinate of the lower right corner of the pipe seat's positioning frame reaches the threshold. If it does, trigger the end plug anomaly detection state.

[0017] S50: Input the preprocessed image, locate the corner features of the tube socket and multiple end plugs;

[0018] S60: Based on the range of the pipe seat positioning frame, determine whether the center of the pipe seat corner feature box is within the range of the pipe seat positioning frame, and filter the pipe seat corner feature box.

[0019] S70: Based on the center point of the selected tube seat corner feature box, fit a reference line and calculate the slope and intercept of the reference line.

[0020] S80: Based on the slope of the reference line, determine whether the slope of the reference line is within the set range. If it is not within the range, continue to wait for the next detection image; if it is within the range, flip the end plug anomaly detection state and execute the subsequent process.

[0021] S90: Based on the number of end plug positioning frames and the actual number of end plugs, determine whether an end plug is missing and output the detection result of the end plug;

[0022] S100: Based on the end plug positioning frame and the reference line, calculate the pixel distance from the top of each end plug to the reference line, and determine whether the end plug as a whole has moved;

[0023] S110: Display and save the results of the end plug anomaly detection.

[0024] According to one embodiment of this application, one frame of image is extracted every N frames.

[0025] According to one embodiment of this application, the preprocessing includes:

[0026] S21: Input an extracted image and delineate the regions that do not need to be detected;

[0027] S22: Automatically extract the coordinates of areas that do not need to be detected using existing algorithms;

[0028] S23: In subsequent video images, image regions that do not need to be detected are automatically cropped using existing algorithms.

[0029] According to one embodiment of this application, a fuel assembly feature localization model is used to predict feature locations in an image.

[0030] According to one embodiment of this application, training a fuel assembly feature localization model includes:

[0031] S31: Using videos from multiple nuclear power plant overhauls, extract a number of images, label the upper tube seat, the outer strip of the grid, and the lower tube seat, and construct a fuel assembly feature dataset;

[0032] S32: Train and test a fuel component feature localization model using a deep learning framework and a fuel component feature dataset.

[0033] According to one embodiment of this application, the fuel assembly features include an upper tube seat, an outer grid strip, and a lower tube seat.

[0034] According to one embodiment of this application, the tube seat is an upper tube seat or a lower tube seat.

[0035] According to one embodiment of this application, when the predicted feature is the upper tube seat, it is determined whether the y coordinate of the lower right corner of the positioning frame of the upper tube seat reaches the threshold. If it does, the upper end plug abnormality detection state is triggered.

[0036] According to one embodiment of this application, when the predicted feature is the lower tube seat, it is determined whether the y coordinate of the lower right corner of the positioning frame of the lower tube seat reaches a threshold. If it does, the lower end plug abnormality detection state is triggered.

[0037] According to one embodiment of this application, the threshold is 0.8 times the height of the preprocessed image.

[0038] According to one embodiment of this application, an angle plug positioning model is used to locate the angle features of the upper tube seat and a plurality of upper end plugs, or to locate the angle features of the lower tube seat and a plurality of lower end plugs.

[0039] According to one embodiment of this application, training the corner plug positioning model includes:

[0040] S51: Using videos from multiple nuclear power plant overhauls, extract a number of images covering pipe sockets and end plugs, and label the corner features of the pipe sockets and the end plugs to construct a corner plug localization dataset;

[0041] S52: Train and test the corner location model using a deep learning framework and a corner location dataset.

[0042] According to one embodiment of this application, based on the range of the upper tube seat positioning frame, it is determined whether the upper tube seat corner feature frame is in the lower half of the upper tube seat positioning frame, and the upper tube seat corner feature frame is filtered.

[0043] According to one embodiment of this application, it is determined whether the corner feature box of the lower tube seat is in the upper half of the positioning frame of the lower tube seat, and the corner feature box of the lower tube seat is filtered.

[0044] According to one embodiment of this application, the range is set such that the absolute value of the slope is not greater than 1.

[0045] According to one embodiment of this application, based on the upper plug positioning frame and the reference line, the pixel distance from the top of each upper plug to the reference line is calculated to determine whether the upper plug as a whole has shifted upward.

[0046] According to one embodiment of this application, the method for calculating the pixel distance of the upper plug includes:

[0047] Calculate the center point on the horizontal axis of the top of each upper plug, and calculate the distance L1 from the center point to the reference line;

[0048] Obtain the distance S1 between the two corners of the upper tube socket below it, where the pixel distance is equal to L1 / S1;

[0049] Calculate the pixel distance from each top edge to the reference line in sequence.

[0050] According to one embodiment of this application, based on the lower end plug positioning frame and the reference line, the pixel distance from the top of each lower end plug to the reference line is calculated to determine whether the lower end plug as a whole has shifted downward.

[0051] According to one embodiment of this application, the method for calculating the pixel distance of the lower plug includes:

[0052] Calculate the center point on the horizontal axis at the bottom of each lower plug, and calculate the distance L3 from the center point to the reference line;

[0053] Obtain the distance S2 between the two corners of the lower tube socket, where the pixel distance is equal to L3 / S2;

[0054] Calculate the pixel distance from each lower end to the reference line in sequence.

[0055] According to one embodiment of this application, S100 includes:

[0056] S101: Calculate the shoulder moment coefficient of each end plug based on the pixel distance from the top of the end plug to the reference line and the pixel distance from the center point of the tube seat corner feature box;

[0057] S102: Determine whether there is a drop bar on each end plug based on the shoulder moment coefficient of each end plug and the set threshold.

[0058] S103: Determine whether the end plug as a whole has moved based on the shoulder moment coefficient of each end plug and the set threshold.

[0059] Secondly, this application provides an intelligent detection system for fuel assembly end plug anomalies, comprising:

[0060] The fuel assembly feature localization module is used to extract images of the four sides of the fuel assembly from video, preprocess the images, and predict the feature locations in the images using the fuel assembly feature localization model.

[0061] The upper plug anomaly detection module is used to calculate the relative distance between the upper plug and the upper tube seat to determine whether there is an anomaly in the upper plug;

[0062] The lower end plug anomaly detection module is used to calculate the relative distance between the lower end plug and the lower tube seat to determine whether there is an anomaly in the lower end plug.

[0063] According to one embodiment of this application, the fuel assembly feature localization module includes:

[0064] The video input unit is used to input video of the appearance of the fuel assembly from all four sides.

[0065] The image extraction unit is used to extract the video of the appearance of the fuel assembly from the four sides input by the video input unit;

[0066] The image preprocessing unit is used to preprocess the images extracted by the image extraction unit;

[0067] The fuel assembly feature localization unit is used to predict the feature locations in the image after the image preprocessing unit has processed it.

[0068] According to one embodiment of this application, the fuel assembly feature localization unit extracts a certain number of images from multiple nuclear power plant overhaul videos, labels the upper tube seat, the outer strip of the grid, and the lower tube seat, constructs a fuel assembly feature dataset, and uses a deep learning framework and the fuel assembly feature dataset to train and test the fuel assembly feature localization model.

[0069] According to one embodiment of this application, the upper plug anomaly detection module includes:

[0070] The upper plug anomaly detection unit is used to determine whether the y-coordinate of the lower right corner of the positioning frame of the upper tube seat has reached the set threshold. Once the threshold is reached, the upper plug anomaly detection state is triggered.

[0071] Upper tube seat positioning unit, used to position the corner features of the upper tube seat and multiple upper end plugs;

[0072] The upper tube seat corner feature box judgment unit is used to determine whether the upper tube seat corner feature box is in the lower half of the upper tube seat positioning box and to filter the upper tube seat corner feature box.

[0073] The upper plug reference line fitting unit is used to fit a reference line based on the center point of the selected upper tube seat corner feature box, and to calculate the slope and intercept of the reference line.

[0074] The reference line judgment unit is used to determine whether the slope of the reference line is within the set range based on the slope of the reference line.

[0075] The upper plug anomaly detection unit is used to determine whether there is an anomaly in the upper plug;

[0076] The upper plug pixel distance calculation unit is used to calculate the pixel distance from the top of each upper plug to the reference line;

[0077] The upper plug detection and storage unit is used to display and save the detection results based on the results of the upper plug anomaly detection.

[0078] According to one embodiment of this application, the lower end plug anomaly detection module includes:

[0079] The lower end plug abnormality detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the lower tube seat has reached the set threshold. If it has, the lower end plug abnormality detection state is triggered.

[0080] The lower tube seat positioning unit is used to position the corner features of the lower tube seat and multiple lower end plugs;

[0081] The lower tube seat corner feature box judgment unit is used to determine whether the lower tube seat corner feature box is in the upper half of the lower tube seat positioning box and to filter the lower tube seat corner feature box.

[0082] The lower end plug reference line fitting unit is used to fit a reference line based on the center point of the selected lower tube seat corner feature box, and to calculate the slope and intercept of the reference line.

[0083] The reference line judgment unit is used to determine whether the slope of the reference line is within the set range based on the slope of the reference line.

[0084] The lower plug abnormality judgment unit is used to determine whether there is an abnormality in the lower plug;

[0085] The lower plug pixel distance calculation unit is used to calculate the pixel distance from the top of each lower plug to the reference line;

[0086] The lower end plug detection and storage unit is used to display and save the detection results based on the results of the lower end plug anomaly detection.

[0087] According to one embodiment of this application, the threshold is set to 0.8 times the height of the preprocessed image.

[0088] According to one embodiment of this application, the range is set such that the absolute value of the slope does not exceed 1.

[0089] Thirdly, this application provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the above-mentioned intelligent detection method for fuel assembly end plug abnormalities when executing the computer-readable instructions.

[0090] Fourthly, this application provides a computer-readable storage medium storing computer-readable instructions, which, when executed, implement the above-mentioned intelligent detection method for fuel assembly end plug anomalies.

[0091] Compared with the prior art, the intelligent detection method and system for fuel assembly end plug anomalies provided in this application have the following advantages:

[0092] This application can simultaneously support four video feeds for fuel assembly end plug anomaly detection, significantly improving detection efficiency. This application can complete fuel assembly end plug anomaly identification with high accuracy and speed, reducing the workload of workers and improving the efficiency of four-sided visual inspection of fuel assemblies.

[0093] This application presents a deep learning-based end plug anomaly detection method, which can more accurately and efficiently detect fuel assembly end plug anomalies. This application uses deep learning for intelligent end plug detection, which can quantify end plug information to a certain extent and more objectively determine the state of the fuel assembly end plugs.

[0094] This application uses an artificial intelligence model to achieve intelligent detection of fuel assembly end plug anomalies. Attached Figure Description

[0095] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the technical description will be briefly introduced below.

[0096] Figure 1 A flowchart of the intelligent detection method for fuel assembly end plug anomalies provided in this application;

[0097] Figure 2 A structural diagram of the intelligent detection method for fuel assembly end plug anomalies provided in this application;

[0098] Figure 3 This is a diagram illustrating the detection effect of the upper plug anomaly provided in this application.

[0099] Figure 4 This is a diagram illustrating the detection effect of the lower end plug abnormality provided in this application;

[0100] Figure 5 This is a schematic diagram of the intelligent detection system for fuel assembly end plug anomalies provided in this application. Detailed Implementation

[0101] The following detailed description provides further details on specific implementation methods.

[0102] like Figures 1 to 4 As shown, this application provides a method for intelligent detection of fuel assembly end plug anomalies, including steps S10 to S230, specifically including the following:

[0103] S10: Receive video footage of the fuel assembly's exterior from all four sides. Extract one frame from every N frames. The value of N can be determined based on the actual situation; N is a positive integer.

[0104] S20: Preprocess the extracted image and crop out the undetected areas in the image.

[0105] S21: Input an extracted image and manually draw out the areas that do not need to be detected using specialized software or tools;

[0106] S22: Automatically extract the coordinates of areas that do not need to be detected using existing algorithms;

[0107] S23: In subsequent video images, image regions that do not need to be detected are automatically cropped using existing algorithms.

[0108] S30: Using the fuel assembly feature localization model, input the preprocessed image and predict the location of features in the image. Fuel assembly features include the upper tube seat, the outer strip of the grid, and the lower tube seat.

[0109] In one embodiment, training the fuel component feature localization model includes:

[0110] S31: Using videos from multiple nuclear power plant overhauls, extract an appropriate number (e.g., 5000 images), label the upper tube seat, outer strip of the grid, and lower tube seat, and construct a fuel assembly feature dataset;

[0111] S32: Train and test a fuel component feature localization model using a deep learning framework and a fuel component feature dataset.

[0112] S40: When the predicted feature is the upper pipe seat, determine whether the y-coordinate of the lower right corner of the upper pipe seat's positioning box reaches a certain threshold. If it does, trigger the upper end plug anomaly detection state. It should be noted that this threshold can be set according to the actual situation. For example, the threshold can be 0.8 times the height of the preprocessed image.

[0113] S50: Using the corner plug positioning model, input the preprocessed image to locate the corner features of the upper tube seat and multiple upper plugs.

[0114] In one embodiment, training the corner pin localization model includes:

[0115] S51: Using videos from multiple nuclear power plant overhauls, extract an appropriate number (e.g., 5000 images) covering the upper / lower pipe seats and upper / lower end plugs, and label the corner features of the upper / lower pipe seats and the upper / lower end plugs to construct a corner plug localization dataset;

[0116] S52: Train and test the corner location model using a deep learning framework and a corner location dataset.

[0117] S60: Based on the range of the upper tube seat positioning frame, determine whether the upper tube seat corner feature box is in the lower half of the upper tube seat positioning frame, and select a suitable upper tube seat corner feature box.

[0118] S70: Based on the center point of the selected upper tube seat corner feature box, fit a reference straight line and calculate the slope and intercept of the reference straight line, such as... Figure 3 As shown.

[0119] S80: Based on the slope of the reference line, determine whether the slope of the reference line is within a suitable range (e.g., the absolute value of the slope is not greater than 1). If it is within an unsuitable range, continue waiting for the next detection image; if it is within a suitable range, flip the upper plug anomaly detection state and execute the subsequent process.

[0120] S90: Based on the number of upper plug positioning frames and the actual number of upper plugs, determine whether the upper plug is missing. If it is missing, output the detection result of the upper plug being missing; if it is not missing, output the detection result of the upper plug not being missing.

[0121] S100: Based on the upper plug positioning frame and the reference line, calculate the pixel distance from the top of each upper plug to the reference line, and determine whether the upper plug as a whole has shifted upward.

[0122] The pixel distance is calculated as follows:

[0123] like Figure 3 As shown, the pixel distance used in this application is a relative distance, that is, the pixel distance is calculated as follows: first, calculate the pixel distance for each upper plug ( Figure 3 The center point on the horizontal axis (x-axis) at the top of the purple box (the center point above the purple box) is used to calculate the distance from this center point to the reference line in S70. Figure 3 The distance L1 from the yellow line in the middle is used to obtain the distance S1 between the two corners of the upper tube seat below the tube seat. Figure 3The distance between the two yellow boxes is calculated as L1 / S1, and the pixel distance or relative distance is equal to L1 / S1. Then, the pixel distance from each upper plug to the reference line is calculated sequentially. By calculating the relative distance, the positional / distance relationship between each upper plug and the upper tube seat can be obtained more accurately, and the inaccuracy of anomaly detection will not be caused by errors in image extraction and processing.

[0124] Furthermore, when it is necessary to calculate the relative distance of the entire upper plug group relative to the upper tube seat, it is only necessary to calculate the average value L2 of each calculated distance L1, and the relative distance of the entire upper plug group is equal to L2 / S1.

[0125] In one embodiment, S100 specifically includes:

[0126] S101: Calculate the shoulder moment coefficient of each upper plug based on the pixel distance from the top of the upper plug to the reference line and the pixel distance from the center point of the upper tube seat corner feature box;

[0127] S102: Determine whether there is a falling rod for each upper plug based on the shoulder moment coefficient of each upper plug and the set threshold.

[0128] S103: Based on the shoulder moment coefficient of each upper plug and the set threshold, determine whether the upper plug as a whole has moved upward.

[0129] S110: Display and save the results of the upper plug anomaly detection.

[0130] S120: Set the corresponding trigger and termination conditions (for the lower plug). Refer to the upper plug anomaly detection process in S10-S110 to execute the lower plug anomaly detection process. It should be noted that re-entering the video is unnecessary because the video order is from upper plug to lower plug. The lower plug anomaly detection process includes:

[0131] S130: Receives video of the appearance of the fuel assembly from all four sides, extracting one frame every N frames.

[0132] S140: Preprocess the extracted image by cropping out undetected regions. The preprocessing method is the same as in S20 and will not be described in detail here.

[0133] S150: Using the fuel assembly feature localization model, input the preprocessed image and predict the feature locations in the image.

[0134] S160: When the predicted feature is the lower pipe seat, determine whether the y coordinate of the lower right corner of the positioning frame of the lower pipe seat reaches a certain threshold. If it does, trigger the lower end plug abnormality detection state.

[0135] S170: Using the corner plug positioning model, input the preprocessed image to locate the corner features of the lower tube seat and multiple lower plugs.

[0136] S180: Based on the range of the lower tube seat positioning frame, determine whether the lower tube seat corner feature box is in the upper half of the lower tube seat positioning frame, and select a suitable lower tube seat corner feature box.

[0137] S190: Based on the center point of the selected lower tube seat corner feature box, fit a reference line and calculate the slope and intercept of the reference line.

[0138] S200: Based on the slope of the reference line, determine whether the slope of the reference line is within a suitable range (the absolute value of the slope does not exceed 1). If it is in an unsuitable range, continue waiting for the next detection image; if it is within a suitable range, flip the lower end of the plug anomaly detection state and execute the subsequent process.

[0139] S210: Based on the number of lower end plug positioning frames and the actual number of lower end plugs, determine whether the lower end plug is missing. If it is missing, output the detection result of the lower end plug being missing; if it is not missing, output the detection result of the lower end plug not being missing.

[0140] S220: Based on the lower end plug positioning frame and the reference line, calculate the pixel distance from the top of each lower end plug to the reference line, and determine whether the lower end plug as a whole has shifted downward.

[0141] The pixel distance is calculated as follows:

[0142] like Figure 4 As shown, first calculate each lower plug ( Figure 4 The center point on the horizontal axis (x-axis) at the bottom of the purple box (the center point below the purple box) is calculated to be the reference line in S190. Figure 4 The distance L3 (from the yellow line in the middle) is used to obtain the distance S2 (from the two corners of the lower tube seat above the tube seat). Figure 4 The distance between the two yellow boxes is calculated as L3 / S2, and the pixel distance or relative distance is equal to L3 / S2. Then, the pixel distance from each lower plug to the reference line is calculated sequentially. By calculating the relative distance, the positional / distance relationship between each lower plug and the lower tube seat can be obtained more accurately, and the inaccuracy of anomaly detection caused by errors in image extraction and processing can be avoided.

[0143] Furthermore, when it is necessary to calculate the relative distance of this set of lower end plugs relative to the lower tube seat, it is only necessary to calculate the average value L4 of each calculated distance L3, and the relative distance of the lower end plugs is equal to L4 / S2.

[0144] S230: Display and save the results of the lower plug anomaly detection.

[0145] Furthermore, based on the aforementioned intelligent detection method for fuel assembly end plug anomalies, this application also provides an intelligent detection system for fuel assembly end plug anomalies, such as... Figure 5 As shown, it includes:

[0146] The fuel assembly feature localization module is used to extract images of the four sides of the fuel assembly from video, preprocess the images, and predict the feature locations in the images using the fuel assembly feature localization model.

[0147] The upper plug anomaly detection module is used to calculate the relative distance between the upper plug and the upper tube seat to determine whether there is an anomaly in the upper plug; the abnormal states include missing upper plug, upper plug moving upward, and upper plug moving downward;

[0148] The lower plug anomaly detection module is used to calculate the relative distance between the lower plug and the lower tube seat to determine whether there is an anomaly in the lower plug. Anomalies include missing lower plug, lower plug moving upward, and lower plug moving downward.

[0149] In one embodiment, the fuel assembly feature localization module includes:

[0150] The video input unit is used to input video of the appearance of the fuel assembly from all four sides.

[0151] The image extraction unit is used to extract the video of the appearance of the fuel assembly on all four sides input by the video input unit. One frame of image is extracted every N frames.

[0152] The image preprocessing unit is used to preprocess the image extracted by the image extraction unit and crop out the undetected areas in the image;

[0153] The fuel assembly feature localization unit is used to predict the location of features in the image after the image preprocessing unit. Based on multiple nuclear power plant overhaul videos, the fuel assembly feature localization unit extracts a certain number of images, labels the upper tube socket, the outer strip of the grid, and the lower tube socket, constructs a fuel assembly feature dataset, and uses a deep learning framework and the fuel assembly feature dataset to train and test the fuel assembly feature localization model.

[0154] In one embodiment, the upper plug anomaly detection module includes:

[0155] The upper plug anomaly detection unit is used to determine whether the y-coordinate of the lower right corner of the positioning frame of the upper tube seat has reached the set threshold. Once the threshold is reached, the upper plug anomaly detection state is triggered.

[0156] Upper tube seat positioning unit, used to position the corner features of the upper tube seat and multiple upper end plugs;

[0157] The upper tube seat corner feature box judgment unit is used to determine whether the upper tube seat corner feature box is in the lower half of the upper tube seat positioning box and to filter the appropriate upper tube seat corner feature box.

[0158] The upper plug reference line fitting unit is used to fit a reference line based on the center point of the selected upper tube seat corner feature box, and to calculate the slope and intercept of the reference line.

[0159] The reference line judgment unit is used to determine whether the slope of the reference line is within a suitable range (the absolute value of the slope does not exceed 1).

[0160] The upper plug anomaly detection unit is used to determine whether there is an anomaly in the upper plug;

[0161] The upper plug pixel distance calculation unit is used to calculate the pixel distance from the top of each upper plug to the reference line;

[0162] The upper plug detection and storage unit is used to display and save the detection results based on the results of the upper plug anomaly detection.

[0163] In one embodiment, the lower plug anomaly detection module includes:

[0164] The lower end plug anomaly detection unit is used to determine whether the y coordinate of the lower right corner of the positioning frame of the lower tube seat reaches the set threshold (e.g., 0.8 times the height of the preprocessed image). If it does, the lower end plug anomaly detection state is triggered.

[0165] The lower tube seat positioning unit is used to position the corner features of the lower tube seat and multiple lower end plugs;

[0166] The lower tube seat corner feature box judgment unit is used to determine whether the lower tube seat corner feature box is in the upper half of the lower tube seat positioning box and to filter the appropriate lower tube seat corner feature box.

[0167] The lower end plug reference line fitting unit is used to fit a reference line based on the center point of the selected lower tube seat corner feature box, and to calculate the slope and intercept of the reference line.

[0168] The reference line judgment unit is used to determine whether the slope of the reference line is within a suitable range (the absolute value of the slope does not exceed 1).

[0169] The lower plug abnormality judgment unit is used to determine whether there is an abnormality in the lower plug;

[0170] The lower plug pixel distance calculation unit is used to calculate the pixel distance from the top of each lower plug to the reference line;

[0171] The lower end plug detection and storage unit is used to display and save the detection results based on the results of the lower end plug anomaly detection.

[0172] This application achieves efficient detection of fuel assembly end plug anomalies by inputting video of the four-sided appearance inspection of the fuel assembly and using techniques such as image preprocessing, image enhancement, deep learning, target optimization, and end plug anomaly judgment mechanism.

[0173] This application can complete the identification of fuel assembly end plug anomalies with high accuracy and high speed, reduce the workload of workers and improve the efficiency of the four-sided appearance inspection of fuel assemblies.

[0174] This application enables more accurate and efficient detection of fuel assembly end plug anomalies. It utilizes deep learning for intelligent end plug detection, which can quantify end plug information to a certain extent and more objectively determine the state of the fuel assembly end plugs.

[0175] In addition, this application also provides an electronic device, including a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the above-mentioned intelligent detection method for fuel assembly end plug abnormalities.

[0176] In addition, this application also provides a computer-readable storage medium storing computer-readable instructions, which, when executed, implement the above-described intelligent detection method for fuel assembly end plug anomalies.

[0177] Computer-readable media, including both permanent and non-permanent, removable and non-removable media, can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0178] The above description is only a specific embodiment of this application, but the protection scope of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application.

Claims

1. A method for intelligent detection of fuel assembly end plug anomaly, characterized in that, The method comprises the following steps: S10: receiving a fuel assembly four-face appearance video, and extracting an image at a set interval; S20: preprocessing the extracted image; S30: inputting the preprocessed image, and predicting a feature position in the image; S40: when the predicted feature is a tube seat, judging whether a y coordinate of a right lower corner point of a positioning frame of the tube seat reaches a threshold value, and triggering an end plug abnormality detection state after the threshold value is reached; S50: inputting the preprocessed image, and locating an angle feature and a plurality of end plugs of the tube seat; S60: judging whether a center of the angle feature frame is within a range of the tube seat positioning frame according to the range of the tube seat positioning frame, and screening the angle feature frame of the tube seat; S70: fitting a reference straight line according to a center point of the screened angle feature frame of the tube seat, and calculating a slope and an intercept of the reference straight line; S80: judging whether the slope of the reference straight line is within a set range according to the slope of the reference straight line, if not, continuing to wait for a next detection image, and if yes, reversing the end plug abnormality detection state and executing a subsequent process; S90: judging whether an end plug is missing according to a number of end plug positioning frames and an actual number of end plugs, and outputting a detection result of the end plug; S100: calculating a pixel distance from a top of each end plug to the reference straight line according to the end plug positioning frame and the reference straight line, and judging whether the end plug as a whole moves; S110: displaying and saving according to a result of the end plug abnormality detection.

2. The method of claim 1, wherein the method further comprises: In S10, an image is extracted every N frames of images.

3. The method of claim 1, wherein the method further comprises: In S20, the preprocessing comprises the following steps: S21: inputting an extracted image, and drawing an area that does not need to be detected; S22: automatically extracting coordinates of the area that does not need to be detected through an algorithm; S23: automatically cropping the image area that does not need to be detected in subsequent video images through the algorithm.

4. The method of claim 1, wherein, In S30, a fuel assembly feature positioning model is used to predict a feature position in the image.

5. The method of claim 4, wherein the method further comprises: Training the fuel assembly feature positioning model comprises the following steps: S31: using a plurality of nuclear power overhauls of videos, extracting a plurality of images, labeling a tube seat, a grid outer strip and a lower tube seat, and constructing a fuel assembly feature data set; S32: using a deep learning framework and the fuel assembly feature data set, training and testing the fuel assembly feature positioning model.

6. The method of claim 4 or 5, wherein the method further comprises: The fuel assembly feature comprises the tube seat, the grid outer strip and the lower tube seat.

7. The method of claim 1, wherein, The tube seat is an upper tube seat or a lower tube seat.

8. The method of claim 1, wherein, In S40, when the predicted feature is the upper tube seat, a y coordinate of a right lower corner point of a positioning frame of the upper tube seat is judged whether to reach a threshold value, and an upper end plug abnormality detection state is triggered after the threshold value is reached.

9. The method of claim 1, wherein, In S40, when the predicted feature is the lower tube seat, a y coordinate of a right lower corner point of a positioning frame of the lower tube seat is judged whether to reach a threshold value, and a lower end plug abnormality detection state is triggered after the threshold value is reached.

10. The method of claim 1, 8, or 9, wherein, The threshold value is 0.8 times of a height of the preprocessed image.

11. The method of claim 1, wherein: In S50, an angle plug positioning model is used to locate an angle feature and a plurality of upper end plugs of the upper tube seat, or to locate an angle feature and a plurality of lower end plugs of the lower tube seat.

12. The method of claim 11, wherein: Training the angle plug positioning model comprises the following steps: S51: using a plurality of nuclear power overhauls of videos, extracting a plurality of images covering the tube seat and the end plug, and labeling the tube seat angle feature and the end plug, and constructing an angle plug positioning data set; S52: using a deep learning framework and the angle plug positioning data set, training and testing the angle plug positioning model.

13. The method of claim 1, wherein: In S60, whether the upper tube seat angle feature box is in the lower half of the upper tube seat positioning box is determined according to the range of the upper tube seat positioning box, and the upper tube seat angle feature box is screened.

14. The method of claim 1, wherein: In S60, whether the lower tube seat angle feature box is in the upper half of the lower tube seat positioning box is determined, and the lower tube seat angle feature box is screened.

15. The method of claim 1, wherein: In S80, the set range is that the absolute value of the slope is not greater than 1.

16. The method of claim 1, wherein: In S100, the pixel distance of each upper end plug top to the reference straight line is calculated according to the upper end plug positioning box and the reference straight line, and whether the upper end plug as a whole exists upward movement is determined.

17. The method of claim 1, wherein: The pixel distance calculation method of the upper end plug includes: The center point of each upper end plug top in the horizontal axis direction is calculated, and the distance L1 of the center point to the reference straight line is calculated. The distance S1 of the two tube seat angles below the upper tube seat is obtained, and the pixel distance is equal to L1 / S1. The pixel distance of each upper end plug to the reference straight line is calculated in turn.

18. The method of claim 1, wherein: In S100, the pixel distance of each lower end plug top to the reference straight line is calculated according to the lower end plug positioning box and the reference straight line, and whether the lower end plug as a whole exists downward movement is determined.

19. The method of claim 18, wherein: The pixel distance calculation method of the lower end plug includes: The center point of each lower end plug bottom in the horizontal axis direction is calculated, and the distance L3 of the center point to the reference straight line is calculated. The distance S2 of the two tube seat angles above the lower tube seat is obtained, and the pixel distance is equal to L3 / S2. The pixel distance of each lower end plug to the reference straight line is calculated in turn.

20. The method of claim 1, wherein: S100 includes: S101: According to the pixel distance of the end plug top to the reference straight line and the pixel distance of the center point of the tube seat angle feature box, the shoulder moment coefficient of each end plug is calculated; S102: According to the shoulder moment coefficient of each end plug and the set threshold value, whether each end plug exists rod falling is determined; S103: According to the shoulder moment coefficient of each end plug and the set threshold value, whether the end plug as a whole exists movement is determined.

21. A fuel assembly end plug anomaly intelligent detection system, characterized in that, It includes: The fuel assembly feature positioning module is used for extracting the image of the four-face appearance video of the fuel assembly, pre-processing the image, and predicting the feature position in the image through the fuel assembly feature positioning model. The upper end plug anomaly detection module is used for calculating the relative distance between the upper end plug and the upper tube seat, and determining whether the upper end plug is abnormal. The lower end plug anomaly detection module is used for calculating the relative distance between the lower end plug and the lower tube seat, and determining whether the lower end plug is abnormal.

22. The intelligent detection system of fuel assembly end plug anomaly according to claim 21, characterized in that, The fuel assembly feature positioning module includes: The video input unit is used for inputting the four-face appearance video of the fuel assembly. The image extraction unit is used for extracting the four-face appearance video of the fuel assembly input by the video input unit. The image preprocessing unit is used for pre-processing the image extracted by the image extraction unit. The fuel assembly feature positioning unit is used for predicting the feature position in the image processed by the image preprocessing unit.

23. The intelligent detection system of fuel assembly end plug anomaly according to claim 22, characterized in that, The fuel assembly feature positioning unit extracts a certain number of images based on a plurality of nuclear power overhauls, labels the upper tube seat, the grid outer strip and the lower tube seat, constructs a fuel assembly feature data set, uses a deep learning framework and the fuel assembly feature data set to train and test the fuel assembly feature positioning model.

24. The intelligent detection system of fuel assembly end plug anomaly according to claim 21, characterized in that, The upper end plug anomaly detection module includes: The upper end plug anomaly detection unit is used for determining whether the y coordinate of the lower right corner point of the positioning box of the upper tube seat reaches the set threshold value, and triggering the upper end plug anomaly detection state after reaching. An upper nozzle seat positioning unit is configured to position the corner feature and the plurality of upper end plugs of the upper nozzle seat. An upper nozzle seat corner feature box judgment unit is configured to judge whether the upper nozzle seat corner feature box is in the lower half of the upper nozzle seat positioning box and to screen the upper nozzle seat corner feature box. An upper end plug reference straight line fitting unit is configured to fit a reference straight line according to the center point of the screened upper nozzle seat corner feature box and to calculate the slope and intercept of the reference straight line. A reference straight line judgment unit is configured to judge whether the slope of the reference straight line is in a set range according to the slope of the reference straight line. An upper end plug anomaly judgment unit is configured to judge whether the upper end plug is abnormal. An upper end plug pixel distance calculation unit is configured to calculate the pixel distance from the top of each upper end plug to the reference straight line. An upper end plug detection saving unit is configured to display and save the detection result according to the result of the upper end plug anomaly detection.

25. The intelligent fuel assembly end plug anomaly detection system of claim 21, wherein, The lower end plug anomaly detection module includes: A lower end plug anomaly detection unit is configured to judge whether the y-coordinate of the lower right corner point of the positioning box of the lower nozzle seat reaches a set threshold value and to trigger the lower end plug anomaly detection state after reaching the set threshold value. A lower nozzle seat positioning unit is configured to position the corner feature and the plurality of lower end plugs of the lower nozzle seat. A lower nozzle seat corner feature box judgment unit is configured to judge whether the lower nozzle seat corner feature box is in the upper half of the lower nozzle seat positioning box and to screen the lower nozzle seat corner feature box. A lower end plug reference straight line fitting unit is configured to fit a reference straight line according to the center point of the screened lower nozzle seat corner feature box and to calculate the slope and intercept of the reference straight line. A reference straight line judgment unit is configured to judge whether the slope of the reference straight line is in a set range according to the slope of the reference straight line. A lower end plug anomaly judgment unit is configured to judge whether the lower end plug is abnormal. A lower end plug pixel distance calculation unit is configured to calculate the pixel distance from the top of each lower end plug to the reference straight line. A lower end plug detection saving unit is configured to display and save the detection result according to the result of the lower end plug anomaly detection.

26. The intelligent detection system of anomalies in fuel assembly end plugs according to claim 24 or 25, characterized in that, The set threshold value is 0.8 times the height of the preprocessed image.

27. The intelligent detection system of anomalies in fuel assembly end plugs according to claim 24 or 25, characterized in that, The set range is that the absolute value of the slope is not more than 1.

28. An electronic device, comprising: The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed to implement the fuel assembly end plug anomaly intelligent detection method.

29. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer readable instructions, and the computer readable instructions are executed to implement the fuel assembly end plug anomaly intelligent detection method.

Images