CN114140419A

CN114140419A - Nuclear fuel assembly appearance defect identification method, system and terminal in underwater irradiation environment

Info

Publication number: CN114140419A
Application number: CN202111421277.8A
Authority: CN
Inventors: 章济; 许小进; 叶琛; 谢晨江; 顾清; 施国龙; 王菁华; 马战龙; 简海林; 陈树
Original assignee: State Nuclear Power Plant Service Co Ltd
Current assignee: State Nuclear Power Plant Service Co Ltd
Priority date: 2021-11-26
Filing date: 2021-11-26
Publication date: 2022-03-04

Abstract

The invention provides a method, a system and a terminal for automatically identifying appearance defects of a nuclear fuel assembly in an underwater irradiation environment, which are suitable for a scene of quickly visually inspecting the fuel assembly in a overhaul unloading period of a pressurized water reactor nuclear power station, not only abandons the traditional method for detecting the appearance defects of the fuel assembly based on manual visual appearance, realizes automatic intelligent detection, but also can effectively reduce the interference of the environment on shooting by adopting an image processing technology and obtain more accurate defect identification results; the obtained defect identification result accurately marks the position of the suspected defect, the inspection is carried out without looking back the video after unloading, and the inspection is completed in one step, so that the inspection efficiency is improved, the risk of abnormal conditions of the fuel is reduced, the integrity of all the fuels can be intelligently inspected in real time in the unloading project, the time of a main line is shortened, and the overhaul process is accelerated.

Description

Nuclear fuel assembly appearance defect identification method, system and terminal in underwater irradiation environment

Technical Field

The application relates to the technical field of underwater image processing, in particular to a method, a system and a terminal for identifying appearance defects of a nuclear fuel assembly in an underwater irradiation environment.

Background

When the fuel assembly is in a high-radiation, high-temperature and high-pressure environment during operation in the stack, radiation deformation, foreign matter abrasion and the like can occur. Or fuel scraping is highly likely to occur due to deformation of the fuel assembly during discharge. And the nuclear power plant can perform appearance inspection on the fuel assemblies discharged in the last cycle during each refueling overhaul period, so that the influence on the stacking reuse is prevented. Therefore, great care is required and any small defect is very likely to be worsened to cause an accident.

The inspection device adopted in China at present is a four-side appearance inspection device of a fuel assembly, and the adopted detection method is a manual visual detection method, namely, four paths of videos of an irradiation-resistant underwater camera are checked by an operator who obtains a nondestructive testing visual secondary certificate and correspond to four sides of the assembly. It is difficult for the general people to concentrate on a thing for a long time, and the process extremely tests the attention and the reaction speed of the people. In the examination time of 3 days, people pay attention to four shooting paths for a long time, and the problems that attention is not focused, one person considers four groups of pictures simultaneously, and the like are easy to occur. The inspection contents are mechanical integrity and surface condition, defects need to be found as soon as possible, and damaged fuel assemblies need to be repaired or replaced in advance, so that the method has important significance for safe and economic operation of the nuclear power station.

The conventional detection method depends on the experience and proficiency of workers, detection standards are not unified, and the accuracy and the detection efficiency cannot be guaranteed due to the fact that manual detection is uncertain. Originally, the video recording is carried out while the manual checking is carried out roughly once. After fuel lift, review video and then carefully inspect for defects.

There is a great uncertainty due to the abnormal damage of the fuel assembly, and the overhaul of each plant is the only time window to check the fuel. The time of the fuel check has a great influence on the schedule of the overhaul. Therefore, the result of the fuel assembly inspection is one of the conditions for scientific and efficient operation of the power station, and objective economic benefits are created for the power station.

The existing nuclear fuel rod appearance defect detection method based on deep learning does not comprise an underwater image processing part, does not have an underwater irradiation environment, and does not comprise a defect automatic identification function which can be added to a fuel visual inspection system in a spent fuel pool.

Disclosure of Invention

In view of the above shortcomings of the prior art, the present application aims to provide a method, a system and a terminal for identifying an appearance defect of a nuclear fuel assembly in an underwater irradiation environment, which solve the above problems of interference inspection in the prior art, such as unclear video, frame loss, low visual observation efficiency, low identification degree and the like caused by water flow disturbance.

In order to achieve the above and other related objects, the present application provides a method for identifying an apparent defect of a nuclear fuel assembly in an underwater irradiation environment, comprising: synchronously taking appearance videos of the nuclear fuel assembly shot by a radiation-resistant underwater camera; performing image preprocessing on the nuclear fuel assembly appearance video; and based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing.

In one or more embodiments of the present application, the method for constructing the automatic defect identification model includes: performing image data enhancement on the plurality of defect images to obtain a defect image training set subjected to defect labeling; training by using the defect image training set to obtain an image recognition model; carrying out light weight deployment on the image recognition model on a lower computer through transfer learning so as to obtain an automatic defect recognition model; (ii) a The defect automatic identification model is evaluated using a fuel inspection video taken by a radiation tolerant underwater camera.

In one or more embodiments of the present application, the appearance defect identification result includes: detecting images with framing marks and corresponding defect types; wherein the defect types include: one or more of wear, cracks, annular grooves, swelling, and scratch defects.

In one or more embodiments of the present application, the manner of synchronously retrieving the video of the appearance of the nuclear fuel assembly captured by the radiation-resistant underwater camera includes: through an interface protocol, the nuclear fuel assembly appearance video is synchronously retrieved while the nuclear fuel assembly appearance video shot by the radiation-resistant underwater camera is stored.

In one or more embodiments of the present application, the manner of obtaining a corresponding appearance defect recognition result from an image-preprocessed nuclear fuel assembly appearance video based on the constructed defect automatic recognition model includes: and inputting the appearance video of the nuclear fuel assembly subjected to image preprocessing into the defect automatic identification model, and accelerating by the GPU to obtain a corresponding appearance defect identification result.

In one or more embodiments of the present application, the manner of obtaining a corresponding appearance defect recognition result from an image-preprocessed nuclear fuel assembly appearance video based on the constructed defect automatic recognition model includes: inputting the appearance video of the nuclear fuel assembly subjected to image preprocessing into the automatic defect identification model to obtain a defect result; and inputting the defect result into an added defect discriminator to discriminate the defect result so as to output a corresponding appearance defect identification result.

In one or more embodiments of the present application, the inputting the defect result to the added defect discriminator for discriminating the defect result to output the corresponding apparent defect recognition result includes: and inputting the defect result to an added defect discriminator to discriminate the final defect type based on the set defect discrimination rule, and outputting an appearance defect identification result.

To achieve the above and other related objects, the present application provides a system for identifying appearance defects of a nuclear fuel assembly in an underwater irradiation environment, comprising: the video calling module is used for synchronously calling the appearance video of the nuclear fuel assembly shot by the radiation-resistant underwater camera; the image preprocessing module is connected with the video calling module and used for preprocessing the images of the appearance videos of the nuclear fuel assembly; and the defect identification module is connected with the image preprocessing module and used for acquiring a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing based on the constructed defect automatic identification model.

To achieve the above and other related objects, the present invention provides a terminal for identifying apparent defects of a nuclear fuel assembly in an underwater irradiation environment, comprising: a memory for storing a computer program; and the processor is used for executing the nuclear fuel assembly appearance defect identification method of the underwater irradiation environment.

As described above, according to the method, the system and the terminal for identifying the appearance defects of the nuclear fuel assembly in the underwater irradiation environment, the appearance videos of the nuclear fuel assembly shot by the radiation-resistant underwater camera are synchronously taken, and the appearance videos of the nuclear fuel assembly are subjected to image preprocessing; and then based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing. The invention is particularly suitable for the pressurized water reactor nuclear power station to carry out quick visual inspection on the fuel assembly during overhaul unloading; the traditional method for detecting the appearance defects of the fuel assembly based on manual visual appearance is abandoned, automatic intelligent detection is realized, the interference of the environment to shooting can be effectively reduced by adopting an image processing technology, and a more accurate defect identification result is obtained; the obtained defect identification result accurately marks the position of the suspected defect, the inspection is carried out without looking back the video after unloading, and the inspection is completed in one step, so that the inspection efficiency is improved, the risk of abnormal conditions of the fuel is reduced, the integrity of all the fuels can be intelligently inspected in real time in the unloading project, the time of a main line is shortened, and the overhaul process is accelerated.

Drawings

Fig. 1 is a schematic flow chart of a method for identifying appearance defects of a nuclear fuel assembly in an underwater irradiation environment according to an embodiment of the present application.

Fig. 2 is a schematic flow chart of a method for identifying appearance defects of a nuclear fuel assembly in an underwater irradiation environment according to an embodiment of the present application.

Fig. 3 is a schematic diagram illustrating a process of constructing an automatic defect identification model according to an embodiment of the present application.

Fig. 4 shows a schematic structural diagram of a nuclear fuel assembly appearance defect identification system of an underwater irradiation environment in an embodiment of the application.

Fig. 5 shows a schematic structural diagram of a nuclear fuel assembly appearance defect identification terminal of an underwater irradiation environment in an embodiment of the application.

Detailed Description

The following description of the embodiments of the present application is provided by way of specific examples, and other advantages and effects of the present application will be readily apparent to those skilled in the art from the disclosure herein. The present application is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present application. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

Embodiments of the present application will be described in detail below with reference to the accompanying drawings so that those skilled in the art to which the present application pertains can easily carry out the present application. The present application may be embodied in many different forms and is not limited to the embodiments described herein.

In order to clearly explain the present application, components that are not related to the description are omitted, and the same reference numerals are given to the same or similar components throughout the specification.

Throughout the specification, when a component is referred to as being "connected" to another component, this includes not only the case of being "directly connected" but also the case of being "indirectly connected" with another element interposed therebetween. In addition, when a component is referred to as "including" a certain constituent element, unless otherwise stated, it means that the component may include other constituent elements, without excluding other constituent elements.

When an element is referred to as being "on" another element, it can be directly on the other element, or intervening elements may also be present. When a component is referred to as being "directly on" another component, there are no intervening components present.

Although the terms first, second, etc. may be used herein to describe various elements in some instances, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, the first interface and the second interface, etc. are described. Also, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms "comprises," "comprising," "includes" and/or "including," when used in this specification, specify the presence of stated features, steps, operations, elements, components, items, species, and/or groups, but do not preclude the presence, or addition of one or more other features, steps, operations, elements, components, species, and/or groups thereof. The terms "or" and/or "as used herein are to be construed as inclusive or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: a; b; c; a and B; a and C; b and C; A. b and C ". An exception to this definition will occur only when a combination of elements, functions, steps or operations are inherently mutually exclusive in some way.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the singular forms "a", "an" and "the" include plural forms as long as the words do not expressly indicate a contrary meaning. The term "comprises/comprising" when used in this specification is taken to specify the presence of stated features, regions, integers, steps, operations, elements, and/or components, but does not exclude the presence or addition of other features, regions, integers, steps, operations, elements, and/or components.

Although not defined differently, including technical and scientific terms used herein, all terms have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. Terms defined in commonly used dictionaries are to be additionally interpreted as having meanings consistent with those of related art documents and the contents of the present prompts, and must not be excessively interpreted as having ideal or very formulaic meanings unless defined.

The invention provides a nuclear fuel assembly appearance defect identification method of an underwater irradiation environment, which comprises the steps of synchronously taking a nuclear fuel assembly appearance video shot by a radiation-resistant underwater camera, and carrying out image preprocessing on the nuclear fuel assembly appearance video; and then based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing. The invention is particularly suitable for the pressurized water reactor nuclear power station to carry out quick visual inspection on the fuel assembly during overhaul unloading; the traditional method for detecting the appearance defects of the fuel assembly based on manual visual appearance is abandoned, automatic intelligent detection is realized, the interference of the environment to shooting can be effectively reduced through an image processing technology, and a more accurate defect identification result is obtained; the obtained defect identification result accurately marks the position of the suspected defect, the inspection is carried out without looking back the video after unloading, and the inspection is completed in one step, so that the inspection efficiency is improved, the risk of abnormal conditions of the fuel is reduced, the integrity of all the fuels can be intelligently inspected in real time in the unloading project, the time of a main line is shortened, and the overhaul process is accelerated.

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings so that those skilled in the art can easily implement the embodiments of the present invention. The present invention may be embodied in many different forms and is not limited to the embodiments described herein.

Fig. 1 is a schematic flow chart showing a method for identifying apparent defects of a nuclear fuel assembly in an underwater irradiation environment according to an embodiment of the present invention.

The method comprises the following steps:

step S11: a nuclear fuel assembly appearance video captured by a radiation tolerant underwater camera is synchronously retrieved.

Optionally, an appearance video of the nuclear fuel assembly under the underwater irradiation environment, which is shot by the radiation-resistant underwater camera in real time, is synchronously taken; it should be noted that the shooting environment and the real-time detection environment are consistent. The radiation-resistant underwater camera can effectively reduce the interference of the environment to shooting.

In one embodiment, four paths of video streams shot by four paths of high-definition radiation-resistant cameras are synchronously taken; for example, four radiation-resistant underwater cameras are used for shooting. The four radiation-resistant underwater cameras are installed above a spent fuel pool transportation channel in a shooting mode. Each camera is oriented perpendicular to 4 faces of the fuel assembly. The four sides of the fuel assembly are visually inspected during lifting of the assembly from the fuel carrier. One of the four cameras is adjustable for viewing the bottom of the assembly. During the transfer of the fuel assemblies from the RPV to the spent fuel pool.

Optionally, four paths of video streams shot by the irradiation-resistant underwater camera are synchronously taken through an interface protocol. It should be noted that the interface protocol is determined based on the actual environment, for example, the video stream is called by the CSI-2 protocol corresponding to the camera serial interface.

Optionally, considering that the conservative decision in the kernel security culture does not affect the storage of the original video, step S11 includes: synchronously retrieving the appearance video of the nuclear fuel assembly while storing the appearance video of the nuclear fuel assembly shot by the radiation-resistant underwater camera through an interface protocol; specifically, in the video access process, the video is stored and simultaneously four paths of video streams shot by the irradiation-resistant underwater camera are synchronously taken through an interface protocol;

by the mode, the defect identification can be carried out while the video is stored, and the storage of the original video is not influenced.

Step S12: and performing image preprocessing on the nuclear fuel assembly appearance video.

Optionally, as the subsequent identification efficiency is improved, the background noise and turbulence influence on the appearance video of the nuclear fuel assembly need to be eliminated; in order to prevent the video noise and turbulence influence from causing false detection and false detection on subsequent defect detection, distinguishing the nuclear fuel assembly appearance video main body from the background, and then applying a turbulence image blind deconvolution model to eliminate the influence of turbulence;

preferably, the nuclear fuel assembly appearance video subject is distinguished from the background by using HDR + multiframe fusion technology of Google Camera. According to the characteristics of turbulence, a turbulence image blind deconvolution model is used, and a conjugate gradient numerical optimization method is used for alternately and iteratively solving a restored image, so that a group of clear videos with noise and thermal turbulence removed are obtained.

Step S13: and based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing.

Optionally, as shown in fig. 2, the method for constructing the automatic defect identification model includes:

step S21: performing image data enhancement on the plurality of defect images to obtain a defect image training set subjected to defect labeling; it should be noted that the defect image may be collected in advance, or may be shot by a high-definition radiation-resistant underwater camera, so that the interference of the environment on the shooting can be effectively reduced, and a clearer training or testing data set for obtaining the defect can be obtained. For example, a defect picture is taken as a defect image by using a defective fuel rod self-made by a national nuclear operating Qingpu experiment base;

step S22: training by using the defect image training set to obtain an image recognition model; specifically, training a visual neural network by using the defect image training set to obtain an image recognition model; preferably, the image recognition model can also be tested by using the acquisition test set;

step S23: deploying the image recognition model to a lower computer to obtain an automatic defect recognition model; specifically, the image recognition model is deployed in a light weight mode on a lower computer through transfer learning to obtain an automatic defect recognition model so as to be more suitable for the use environment of the automatic defect recognition system;

step S24: the defect automatic identification model is evaluated using a fuel inspection video taken by a radiation tolerant underwater camera. Specifically, the automatic defect identification model is evaluated by using a fuel inspection video acquired by a radiation-resistant underwater camera in the real environment of the nuclear power station, namely, the automatic defect identification model is verified by inputting the fuel inspection video into the automatic defect identification model deployed on a lower computer to obtain an appearance defect identification result and comparing the appearance defect identification result with a defect identification result of manual detection. For example, defect identification is performed by using a set of fuel inspection videos taken during Shandong nuclear major repair for verification. Suspected defects are accurately marked on the video. And the similar synchronous image processing with each component delay less than 1 second is realized, and the characteristics of the original image cannot be covered by the sharpening processing. The appearance defects of the fuel assemblies are effectively identified. The requirements of the fuel assembly on the appearance inspection of the underwater television in the underwater irradiation environment are met.

Alternatively, for the case of limited defect pictures, too little data means easy overfitting; therefore, data enhancement is carried out aiming at the graph training of small sample capacity; using reinforcement learning to find the optimal image transformation strategy from the data itself, learning different reinforcement methods for different tasks, such as various geometric transformations, the color transformation strategy has all proven helpful to improve the generalization capability of the model.

Therefore, step S21 includes:

respectively enhancing data of each defect image to obtain a plurality of enhanced defect images with defects; specifically, image cropping, noise adding, turning, scaling deformation, random erasing, contrast adjustment and the like are performed on an existing defective image to obtain a large number of multiple defective images with defects respectively;

labeling each enhanced defect image by a detector, and acquiring each enhanced defect image subjected to defect labeling to obtain a defect image training set; the detection personnel who participate in image labeling all work for many years and have rich experience need another detection personnel to recheck after each image is labeled.

It should be noted that image data enhancement and image training can be performed on the cloud service platform.

Optionally, the image data is enhanced by using an AutoML technology of G1 edge Brain, and the functions used by the AutoML technology include a search space, a policy function, and a reward index. The functionality may learn different enhancement methods for different tasks.

Optionally, the defect type of the enhanced defect image is a defect of a homemade fuel rod operated by a national nuclear, and the defect type includes: one or more of wear (120 microns, 350 microns deep), cracks, annular grooves, swelling, and scratches.

Optionally, step S22 includes: and (3) a visual-based neural network detection task, namely training a set of labeled defect images. The target identification framework adopts a CNN algorithm, improves a depth residual convolution neural network structure ResNet, serves as a convolution neural network for automatically extracting characteristics, and constructs a neural network structure and sets a loss function. And starting to train on a cloud computing service platform by using the defect image training set, and reducing a loss function through multiple rounds of iteration to obtain an image recognition model.

Optionally, GPU neural network acceleration may be performed while the neural network trains the model.

Optionally, step S23 includes: and carrying out lightweight deployment on the image recognition model on a lower computer through Transfer learning, and implementing the image recognition model in a way that client software provides a use interface of system functions.

Optionally, step S24: evaluating the automatic defect identification model by using a fuel inspection video shot by a radiation-resistant underwater camera, determining whether to participate in use based on an evaluation result, and if the evaluation result meets the requirement, adopting the model to perform subsequent defect identification; if the defect identification model does not meet the requirement, rebuilding to obtain a defect automatic identification model meeting the requirement; that is to say, the finally used automatic defect identification models are all actually measured by nuclear power plant fuel inspection videos and meet the requirements; for example, a fuel inspection video shot during Shandong nuclear power overhaul is applied to defect identification, and the position of a suspected defect is accurately marked according to the result to achieve the target. The synchronous inspection does not need to be carried out by looking back the video after unloading, but is carried out in one step. The inspection efficiency is improved, and the risk of abnormal conditions of the fuel is reduced.

Optionally, the appearance defect identification result includes: detecting images with framing marks and corresponding defect types; wherein the defect types include: one or more of wear, cracks, annular grooves, swelling, and scratch defects.

Optionally, in the detection process, the image recognition model often outputs defect results regarding a plurality of defect categories appearing at the same position, so that the final defect result needs to be determined; therefore, step S13 includes:

inputting the appearance video of the nuclear fuel assembly subjected to image preprocessing into the automatic defect identification model to obtain a defect result;

and inputting the defect result into an added defect discriminator to discriminate the defect result so as to output a corresponding appearance defect identification result.

Further, in an embodiment, the manner of determining the final defect recognition result by the defect determiner includes:

and inputting the defect result to an added defect discriminator to discriminate the final defect type based on the set defect discrimination rule, and outputting an appearance defect identification result.

Preferentially, the set defect judgment rule can judge the type of the defect according to the probability of all defect types at the same position in the defect result; for example, the probability of all defect types at the same position (weight that can be set in conjunction with the defect type) can be calculated by using a defect discrimination algorithm, and the defect type can be determined based on the obtained calculation result.

In addition, for defect types with larger difference of probability values, the defect types can be directly distinguished by setting a probability threshold; for example, setting the probability threshold to 70%, the model output for the A part defect type is: abrasion, with a probability of 30%; cracking with a probability of 70%; namely, the judgment result is the crack;

certainly, the finally judged defect types can be one or more, and the judging method can be directly judged through numerical values or can be judged through a trained judging algorithm; for example, the model output has defect types for part a: an annular groove with a probability of 50%; wear, with a probability of 50%; the annular groove and wear are also determined.

It should be noted that, a defect discriminator is added, and a selectable button is set on the display interface.

In order to better describe the method for identifying the appearance defects of the nuclear fuel assembly in the underwater irradiation environment, the following specific embodiments are provided;

example 1: a nuclear detection fuel rod appearance defect automatic identification method of underwater irradiation environment; FIG. 3 is a schematic diagram of a construction process of an applied automatic defect identification model;

the automatic defect identification model of the embodiment is designed on an open source machine learning platform tenorflow of Google, and is realized by programming Python, c + +, opencv and Pythrch, and a Windows operating system is adopted;

the method comprises the following steps:

step 1: synchronously calling four paths of videos shot by an irradiation-resistant underwater camera through an interface protocol so as to more clearly obtain a test data set of defects and a real fuel inspection video during the nuclear power overhaul;

step 2: performing data enhancement on the existing fuel defect picture by adopting the AutoML technology of G1 edge Brain, wherein the data enhancement comprises image marking, image cutting, noise adding and contrast adjustment of the existing fuel defect picture, and the image marking, the image cutting, the noise adding and the contrast adjustment are used as a training set of the fuel defect picture;

and 3, step 3: and (3) carrying out a visual-based neural network detection task, and carrying out a training set on the labeled fuel defect pictures. The target identification framework adopts a CNN algorithm, and improves a depth residual convolution neural network structure ResNet as a convolution neural network for automatically extracting features. And constructing a neural network structure and setting a loss function. And starting to train on the cloud computing service platform. And after 500 iterations, the loss function is reduced to obtain an image recognition model. After inspection, the performance of the final model in the test set is excellent;

and 4, step 4: and carrying out lightweight deployment on the trained model on a lower computer through Transfer learning, and providing a use interface mode of system functions through client software to finish implementation. And 5, step 5: and verifying the automatic defect identification model by adopting a real fuel inspection video during nuclear power overhaul, and accurately marking suspected defects on the video. And the similar synchronous image processing with each component delay less than 1 second is realized, and the characteristics of the original image cannot be covered by the sharpening processing. The appearance defects of the fuel assemblies are effectively identified. The requirements of the fuel assembly on the appearance inspection of the underwater television in the underwater irradiation environment are met.

And 6, step 6: synchronously calling four paths of video streams shot by a radiation-resistant underwater camera while storing the four paths of video streams through an interface protocol;

and 7, step 7: the HDR + multi-frame fusion technology of Google Camera is adopted to distinguish the four paths of video stream bodies from the background. According to the characteristics of turbulence, a turbulence image blind deconvolution model is used, and a conjugate gradient numerical optimization method is used for alternately and iteratively solving a restored image to obtain a group of clear videos with noise and thermal turbulence removed;

and 8, step 8: and inputting the appearance video of the nuclear fuel assembly subjected to image preprocessing into the defect automatic identification model, accelerating the model by a GPU, setting selectable buttons on a display interface by an added multi-person discriminator, and outputting the detection image frame with the frame selection mark. The output result meets the expectation, and the position of the suspected defect is selected out.

The defect identification method is used for preprocessing the problems of irradiation snowflake points, water flow disturbance of underwater videos, feature capture in a high-speed motion state and the like, and overcomes factors of interference detection, such as unclear videos, lost frames, low visual observation efficiency, low identification degree and the like caused by water flow disturbance. Effectively and intelligently identify the appearance defects of the fuel assemblies and meet the inspection requirements of the fuel assemblies.

The invention provides a nuclear fuel assembly appearance defect identification system of an underwater irradiation environment, which is similar to the principle of the embodiment.

Specific embodiments are provided below in conjunction with the attached figures:

fig. 4 shows a schematic structural diagram of a nuclear fuel assembly appearance defect identification system of an underwater irradiation environment in an embodiment of the invention.

The system comprises:

a video retrieving module 41, configured to retrieve a nuclear fuel assembly appearance video captured by the radiation-resistant underwater camera synchronously;

the image preprocessing module 42 is connected to the video retrieving module 41 and is used for performing image preprocessing on the nuclear fuel assembly appearance video;

and the defect identification module 43 is connected with the image preprocessing module 42 and is used for obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing based on the constructed defect automatic identification model.

It should be noted that the division of each module in the embodiment of the system in fig. 2 is only a division of a logical function, and all or part of the actual implementation may be integrated into one physical entity or may be physically separated. And these modules can be realized in the form of software called by processing element; or may be implemented entirely in hardware; part of the modules can be realized in a software calling mode through a processing element, and part of the modules can be realized in a hardware mode;

for example, the modules may be one or more integrated circuits configured to implement the above methods, such as: one or more Application Specific Integrated Circuits (ASICs), or one or more microprocessors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs), among others. For another example, when one of the above modules is implemented in the form of a Processing element scheduler code, the Processing element may be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

Therefore, since the implementation principle of the system for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment has been described in the foregoing embodiments, repeated descriptions are omitted here.

Optionally, the method for constructing the automatic defect identification model includes: performing image data enhancement on the plurality of defect images to obtain a defect image training set subjected to defect labeling; training by using the defect image training set to obtain an image recognition model; carrying out light weight deployment on the image recognition model on a lower computer through transfer learning so as to obtain an automatic defect recognition model; (ii) a The defect automatic identification model is evaluated using a fuel inspection video taken by a radiation tolerant underwater camera.

Optionally, the method for performing image data enhancement on a plurality of defect images to obtain a defect image training set subjected to defect labeling includes: respectively enhancing data of each defect image to obtain a plurality of enhanced defect images with defects; and acquiring each enhanced defect image subjected to defect labeling to obtain a defect image training set.

Optionally, the manner of determining the final defect identification result by the defect determiner includes: and judging the corresponding final defect type according to the defect result output by the image recognition model based on the set defect judgment rule, and outputting an appearance defect recognition result.

Optionally, the video retrieving module 41 is configured to retrieve, through an interface protocol, the nuclear fuel assembly appearance video synchronously while storing the nuclear fuel assembly appearance video captured by the radiation-resistant underwater camera.

Optionally, the defect identification module 43 is configured to input the nuclear fuel assembly appearance video subjected to image preprocessing into the defect automatic identification model, and obtain a corresponding appearance defect identification result through GPU acceleration.

Optionally, the defect identification module 43 is configured to input the nuclear fuel assembly appearance video subjected to image preprocessing into the defect automatic identification model, so as to obtain a defect result; and inputting the defect result into an added defect discriminator to discriminate the defect result so as to output a corresponding appearance defect identification result.

Optionally, the defect identification module 43 is configured to input the defect result to an added defect discriminator to discriminate a final defect type based on a set defect discrimination rule, and output an appearance defect identification result.

Fig. 5 shows a schematic structural diagram of a nuclear fuel assembly appearance defect identification terminal 50 of an underwater irradiation environment in an embodiment of the invention.

The terminal 50 for identifying the appearance defects of the nuclear fuel assembly in the underwater irradiation environment comprises: a memory 51 and a processor 52, the memory 51 for storing computer programs; the processor 52 runs a computer program to implement the method for identifying the appearance defect of the nuclear fuel assembly in the underwater irradiation environment as shown in fig. 1.

Optionally, the number of the memories 51 may be one or more, the number of the processors 52 may be one or more, and fig. 5 is an example.

Optionally, the processor 52 in the terminal 50 for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment loads one or more instructions corresponding to the processes of the application program into the memory 51 according to the steps shown in fig. 1, and the processor 52 runs the application program stored in the first memory 51, so as to implement various functions in the method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment shown in fig. 1.

Optionally, the memory 51 may include, but is not limited to, a high speed random access memory, a non-volatile memory. Such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid-state storage devices; the Processor 52 may include, but is not limited to, a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.

Optionally, the Processor 52 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.

The invention also provides a computer readable storage medium storing a computer program which, when running, implements the method for identifying apparent defects of a nuclear fuel assembly in an underwater irradiation environment as shown in fig. 1. The computer-readable storage medium may include, but is not limited to, floppy diskettes, optical disks, CD-ROMs (compact disc-read only memories), magneto-optical disks, ROMs (read-only memories), RAMs (random access memories), EPROMs (erasable programmable read only memories), EEPROMs (electrically erasable programmable read only memories), magnetic or optical cards, flash memory, or other type of media/machine-readable medium suitable for storing machine-executable instructions. The computer readable storage medium may be a product that is not accessed by the computer device or may be a component that is used by an accessed computer device.

In summary, according to the method, the system and the terminal for identifying the appearance defects of the nuclear fuel assembly in the underwater irradiation environment, the appearance videos of the nuclear fuel assembly shot by the radiation-resistant underwater camera are synchronously taken, and image preprocessing is performed on the appearance videos of the nuclear fuel assembly; and then based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing. The invention is particularly suitable for the pressurized water reactor nuclear power station to carry out quick visual inspection on the fuel assembly during overhaul unloading; the traditional method for detecting the appearance defects of the fuel assembly based on manual visual appearance is abandoned, automatic intelligent detection is realized, the interference of the environment to shooting can be effectively reduced by adopting an image processing technology, and a more accurate defect identification result is obtained; the obtained defect identification result accurately marks the position of the suspected defect, the inspection is carried out without looking back the video after unloading, and the inspection is completed in one step, so that the inspection efficiency is improved, the risk of abnormal conditions of the fuel is reduced, the integrity of all the fuels can be intelligently inspected in real time in the unloading project, the time of a main line is shortened, and the overhaul process is accelerated. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.

The above embodiments are merely illustrative of the principles and utilities of the present application and are not intended to limit the application. Any person skilled in the art can modify or change the above-described embodiments without departing from the spirit and scope of the present application. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical concepts disclosed in the present application shall be covered by the claims of the present application.

Claims

1. A nuclear fuel assembly appearance defect identification method of an underwater irradiation environment is characterized by comprising the following steps:

synchronously taking appearance videos of the nuclear fuel assembly shot by a radiation-resistant underwater camera;

performing image preprocessing on the nuclear fuel assembly appearance video;

and based on the constructed defect automatic identification model, obtaining a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing.

2. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to claim 1, wherein the automatic defect identification model is constructed in a manner that:

performing image data enhancement on the plurality of defect images to obtain a defect image training set subjected to defect labeling;

training by using the defect image training set to obtain an image recognition model;

carrying out light weight deployment on the image recognition model on a lower computer through transfer learning so as to obtain an automatic defect recognition model; the defect automatic identification model is evaluated using a fuel inspection video taken by a radiation tolerant underwater camera.

3. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to any one of claims 1 or 2, wherein the appearance defect identification result comprises the following steps: detecting images with framing marks and corresponding defect types;

wherein the defect types include: one or more of wear, cracks, annular grooves, swelling, and scratch defects.

4. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to claim 1, wherein the manner of synchronously retrieving the appearance videos of the nuclear fuel assemblies shot by the radiation-resistant underwater camera comprises the following steps:

through an interface protocol, the nuclear fuel assembly appearance video is synchronously retrieved while the nuclear fuel assembly appearance video shot by the radiation-resistant underwater camera is stored.

5. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to claim 1, wherein the image preprocessing of the nuclear fuel assembly appearance video comprises:

and removing noise from the appearance video of the nuclear fuel assembly, and eliminating turbulence influence by using a turbulence image blind deconvolution model.

6. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to claim 1, wherein the mode of obtaining the corresponding appearance defect identification results from the nuclear fuel assembly appearance videos subjected to image preprocessing based on the constructed defect automatic identification model comprises the following steps:

and inputting the appearance video of the nuclear fuel assembly subjected to image preprocessing into the defect automatic identification model, and accelerating by the GPU to obtain a corresponding appearance defect identification result.

7. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment as claimed in claim 1 or 6, wherein the manner of obtaining the corresponding appearance defect identification results from the nuclear fuel assembly appearance videos subjected to image preprocessing based on the constructed defect automatic identification model comprises:

8. The method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment as claimed in claim 7, wherein the inputting the defect results into the added defect discriminator for discrimination of the defect results and the outputting the corresponding appearance defect identification results comprises:

9. A system for identifying cosmetic defects in a nuclear fuel assembly in an underwater irradiation environment, comprising:

the video calling module is used for synchronously calling the appearance video of the nuclear fuel assembly shot by the radiation-resistant underwater camera;

the image preprocessing module is connected with the video calling module and used for preprocessing the images of the appearance videos of the nuclear fuel assembly;

and the defect identification module is connected with the image preprocessing module and used for acquiring a corresponding appearance defect identification result from the nuclear fuel assembly appearance video subjected to image preprocessing based on the constructed defect automatic identification model.

10. A nuclear fuel assembly appearance defect identification terminal of an underwater irradiation environment, comprising:

a memory for storing a computer program;

a processor for executing the method for identifying the appearance defects of the nuclear fuel assemblies in the underwater irradiation environment according to any one of claims 1 to 8.