Weld joint detection system and method for lower end plug of nuclear fuel rod

Abstract

The invention discloses a welding seam detection system and method for a lower end plug of a nuclear fuel rod, and relates to the technical field of nuclear power equipment manufacturing automation, the welding seam detection system comprises a data acquisition module and a data analysis module, and the data acquisition module comprises a camera, a movement mechanism and an endoscopic detection probe installed on the movement mechanism. A data analysis module is used for positioning a welding seam detection position on an environment image of a target to-be-detected lower end plug shot by a camera after a nuclear fuel assembly hovers, sending a motion control instruction to control a motion mechanism to bear an endoscopic detection probe to reach the welding seam detection position, and controlling the endoscopic detection probe to reach the welding seam detection position; the endoscope video is collected, and the small-size endoscopic detection probe goes deep into the weld joint detection position, so that the limitation of a tiny space is overcome, and the full-coverage collection of the weld joint image of the lower end plug to be detected is realized. Furthermore, the data analysis module is used for achieving welding seam anomaly detection based on endoscope videos, and the accuracy of anomaly detection results is improved.

Description

Technical Field

[0001] The present application relates to the field of nuclear power equipment manufacturing automation technology, and in particular to a weld detection system and method for the lower end plug of a nuclear fuel rod. Background Art

[0002] Nuclear fuel rods are essential components of nuclear fuel assemblies, the core components of a reactor. From manufacturing to placement in the reactor, nuclear fuel assemblies undergo numerous steps, including pellet material preparation, core processing and molding, component sealing and coating, and assembly. Quality inspections are required at each stage to ensure safe operation. Furthermore, to ensure the safe operation of nuclear power plants, prevent the leakage of radioactive materials, and avoid the detachment of the lower end plugs of nuclear fuel rods throughout the fuel assembly's lifecycle, which could lead to major nuclear safety accidents due to falling pellets, the welds of the lower end plugs of nuclear fuel rods must be inspected before new fuel assemblies are placed in the reactor to confirm the integrity of the welds, thereby improving the operational safety of nuclear power plants. Summary of the Invention

[0003] In view of the above problems, this application provides a nuclear fuel rod lower end plug weld detection system and method to achieve nuclear fuel rod lower end plug weld detection. The specific solution is as follows:

[0004] A first aspect of the present application provides a weld inspection system for a lower end plug of a nuclear fuel rod, comprising: a data acquisition module and a data analysis module, wherein the data acquisition module comprises a camera, a motion mechanism, and an endoscopic inspection probe mounted on the motion mechanism;

[0005] The camera is used to capture a sequence of environmental images after the nuclear fuel assembly is hovering, and send the sequence of environmental images to the data analysis module;

[0006] The motion mechanism is used to receive the motion control instruction issued by the data analysis module, and move the endoscopic detection probe to the weld detection position in response to the motion control instruction;

[0007] The endoscopic detection probe is used to record the endoscopic video after receiving the first instruction and send the endoscopic video to the data analysis module;

[0008] The data analysis module is used to:

[0009] receiving the environmental image sequence, generating the motion control instruction based on the environmental image sequence, and sending the motion control instruction to the motion mechanism;

[0010] In response to the endoscopic detection probe reaching the weld detection position, issuing a video recording instruction, the video recording instruction including the first instruction, the first instruction instructing the endoscopic detection probe to start recording;

[0011] The endoscope video is received, and based on the endoscope video, whether the weld state of the lower end plug to be inspected is normal is detected to obtain the weld inspection result of the lower end plug to be inspected.

[0012] In one possible implementation, the endoscopic detection probe is composed of a reflector, an endoscope, and multiple lighting modules, wherein the endoscope includes an endoscope body and an endoscope lens installed at the first end inside the endoscope body, multiple lighting modules are arranged in a ring around the endoscope lens, and the reflector is installed at the second end outside the endoscope body at a preset reflection angle.

[0013] In a possible implementation, the data analysis module is configured to receive the environmental image sequence, and when generating the motion control instruction based on the environmental image sequence, is specifically configured to:

[0014] identifying the coordinates of the weld detection position based on the environmental image sequence;

[0015] Based on the relative positional relationship between the fixed position of the camera and the real-time position of the endoscopic detection probe, the real-time position of the endoscopic detection probe is mapped into a coordinate system to obtain the coordinates of the real-time position of the endoscopic detection probe.

[0016] The motion control instruction is obtained based on the coordinates of the weld detection position and the coordinates of the real-time position of the endoscopic detection probe. The motion control instruction includes a horizontal displacement parameter, a vertical displacement parameter, and a telescopic parameter.

[0017] In one possible implementation, the motion mechanism is composed of a motor module, a telescopic mechanism, a horizontal displacement mechanism, a vertical displacement mechanism, and a rotating device. The telescopic mechanism is in the form of an electric push rod and is mounted on a fixed plate of the vertical displacement mechanism. The vertical displacement mechanism is mounted on a fixed support of the horizontal displacement mechanism. The horizontal displacement mechanism is mounted on the rotating device. The rotating device is fixed to a foundation at a preset position via a mounting bracket.

[0018] The motion mechanism is used to receive the motion control instruction issued by the data analysis module, and when responding to the motion control instruction to move the endoscopic detection probe to the weld detection position, is specifically used to:

[0019] In response to the motion control instruction, the motor module controls the horizontal displacement mechanism to perform horizontal displacement according to the horizontal displacement parameter, controls the vertical displacement mechanism to perform vertical displacement according to the vertical displacement parameter, and controls the telescopic mechanism to perform telescopic displacement according to the telescopic parameter, so as to move the endoscopic detection probe to the weld detection position.

[0020] In a possible implementation, the video recording instruction further includes a second instruction, wherein the second instruction instructs the motion mechanism to move along a preset path in the horizontal plane where the weld detection position is located;

[0021] The motion mechanism is further configured to respond to the second instruction and control the telescopic mechanism and the horizontal displacement mechanism to move along the preset path through the motor module.

[0022] In one possible implementation, the data analysis module is used to receive the endoscopic video, detect whether the weld state of the lower end plug to be inspected is normal based on the endoscopic video, and obtain the weld inspection result of the lower end plug to be inspected, specifically for:

[0023] receiving the endoscope video, and obtaining an image of a weld to be identified of the lower end plug to be inspected based on the endoscope video;

[0024] The weld image to be identified is input into an anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result. If the weld detection result is abnormal, the anomaly detection model outputs a weld identification image embedded with an abnormal image block; if the weld detection result is normal, the anomaly detection model outputs the weld image to be identified.

[0025] In a possible implementation, when the data analysis module is used to obtain the image of the weld to be identified of the lower end plug to be inspected based on the endoscopic video, it is specifically used to:

[0026] For each endoscopic video frame, converting the endoscopic video frame from a ring view image to a rectangular view image based on a polar coordinate conversion relationship;

[0027] Optimizing each pixel of the rectangular view image based on the grayscale value of each pixel in a preset neighborhood to obtain an optimized rectangular view image;

[0028] The optimized rectangular view image is subjected to displacement correction to obtain the weld image to be identified.

[0029] In one possible implementation, the data analysis module is used to input the weld image to be identified into the anomaly detection model, and obtain the weld identification image output by the anomaly detection model based on the weld detection result, specifically for:

[0030] Inputting the weld image to be identified into the anomaly detection model;

[0031] Dividing the weld image to be identified into blocks using the anomaly detection model to obtain multiple image blocks;

[0032] Extracting a feature vector of the image block using the anomaly detection model based on a multi-layer attention mechanism, substituting the feature vector of the image block into a probability density function of a multivariate Gaussian distribution to obtain a probability that the image block belongs to a normal class;

[0033] determining whether the image block is an abnormal image block based on the probability by using the abnormality detection model;

[0034] The abnormal image block is located in the weld image to be identified by using the abnormality detection model, and the abnormal image block is marked at the corresponding position to obtain the weld identification image and output it.

[0035] In a possible implementation, when the data analysis module determines whether the image block is an abnormal image block based on the probability using the abnormality detection model, it is specifically configured to:

[0036] Calculating feature similarities between an image block and its adjacent image blocks using the anomaly detection model;

[0037] Determining the weight of each of the adjacent image blocks according to the feature similarity using the anomaly detection model, wherein the weight is positively correlated with the feature similarity;

[0038] Performing a weighted average of the probabilities that the adjacent image blocks belong to the normal class based on the weights of the adjacent image blocks by the anomaly detection model to obtain a neighborhood anomaly score of the image block;

[0039] fusing the neighborhood anomaly score and the probability that the image block belongs to the normal class through the anomaly detection model to obtain an anomaly score fusion result of the image block;

[0040] If the abnormality score fusion result is greater than a probability threshold, the image block is determined to be a normal image block; if the abnormality score fusion result is not greater than a preset probability threshold, the image block is determined to be an abnormal image block.

[0041] A second aspect of the present application provides a method for detecting the weld of a nuclear fuel rod lower end plug, which is applied to a data analysis module in a nuclear fuel rod lower end plug weld detection system. The nuclear fuel rod lower end plug weld detection system includes a data acquisition module and the data analysis module, wherein the data acquisition module includes a camera, a motion mechanism, and an endoscopic detection probe mounted on the motion mechanism. The method for detecting the weld of a nuclear fuel rod lower end plug includes:

[0042] receiving a sequence of environmental images captured by the camera after the nuclear fuel assembly is hovering;

[0043] generating a motion control instruction based on the environmental image sequence, and sending the motion control instruction to the motion mechanism so that the motion mechanism moves the endoscopic detection probe to a weld detection position in response to the motion control instruction;

[0044] In response to the endoscopic detection probe reaching the weld detection position, issuing a video recording instruction, the video recording instruction including a first instruction, the first instruction instructing the endoscopic detection probe to start recording;

[0045] receiving an endoscopic video recorded by the endoscopic detection probe, and obtaining an image of a weld to be identified of the lower end plug to be detected based on the endoscopic video;

[0046] The weld image to be identified is input into an anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result. If the weld detection result shows that an anomaly exists, the weld identification image is the weld image to be identified with the abnormal position marked.

[0047] By means of the above-mentioned technical solution, the present application provides a weld inspection system and method for the lower end plug of a nuclear fuel rod. The weld inspection system includes a data acquisition module and a data analysis module, wherein the data acquisition module includes a camera, a motion mechanism, and an endoscopic inspection probe mounted on the motion mechanism. The data analysis module locates the weld inspection position based on the environmental image of the target lower end plug to be inspected, captured by the camera after hovering over the nuclear fuel assembly. A motion control instruction is issued to control the motion mechanism to carry the endoscopic inspection probe to the weld inspection position. The endoscopic inspection probe is controlled to reach the weld inspection position of the target lower end plug to be inspected, and an endoscopic video is collected. Due to the small size of the endoscopic inspection probe, it penetrates into the weld inspection position, overcoming the limitations of the small space and achieving full coverage of the image of the weld to be identified of the target lower end plug to be inspected. Furthermore, the data analysis module implements weld inspection based on the endoscopic video. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the originals and elements are not necessarily drawn to scale.

[0049] Figure 1 A schematic structural diagram of a nuclear fuel assembly provided in an embodiment of the present application;

[0050] Figure 2 A schematic structural diagram of a nuclear fuel rod lower end plug weld detection system provided in an embodiment of the present application;

[0051] Figure 3A schematic flow chart of a method for detecting a weld of a lower end plug of a nuclear fuel rod provided in an embodiment of the present application;

[0052] Figure 4 A schematic diagram of the specific structure of a nuclear fuel rod lower end plug weld detection system provided in an embodiment of the present application;

[0053] Figure 5 A schematic diagram of the specific structure of an endoscopic detection probe provided in an embodiment of the present application;

[0054] Figure 6 A flowchart of a specific implementation of a method for detecting a weld of a lower end plug of a nuclear fuel rod provided in an embodiment of the present application;

[0055] Figure 7 A schematic diagram of the conversion of an endoscopic image provided in an embodiment of the present application;

[0056] Figure 8 A schematic flow chart of a method for detecting a weld of a lower end plug of a nuclear fuel rod provided in an embodiment of the present application;

[0057] Figure 9 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] The following describes the embodiments of the present application in conjunction with the accompanying drawings. The terms used in the implementation methods of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0059] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0060] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0061] The present invention provides a system and method for detecting welds on the lower end plugs of nuclear fuel rods. The system is specifically used for non-destructive testing and abnormality identification of the welds on the lower end plugs of each nuclear fuel rod in a nuclear fuel assembly during its transfer from a fuel building to a reactor. A nuclear fuel assembly is a heat release component within a reactor consisting of a group of fuel rods and other components. It is a fuel loading and unloading unit for a reactor. Figure 1 The following is a schematic diagram of the structure of a nuclear fuel assembly: Figure 1 As shown in the figure, the tetrahedral nuclear fuel assembly consists of multiple nuclear fuel rods arranged in a 17×17 array and an upper tube seat, a lower end plug, and a lower tube seat. The four sides of the nuclear fuel assembly are 17 nuclear fuel rods, Figure 1 The structure of a single nuclear fuel rod is shown. The main body of the nuclear fuel rod is composed of a zirconium alloy cladding wrapping a core block, and the lower end is connected to a lower end plug. The main body of the nuclear fuel rod and the lower end plug are welded to form a sealed whole. The lower end plug is located at the bottom of the fuel rod and plays a role in sealing and accommodating the fuel core block, while the lower end of the fuel rod is inserted into the corresponding positioning hole and other structures on the lower tube seat. The lower tube seat realizes the positioning and support of the fuel rod. The lower tube seat serves as the bottom support of the fuel assembly and bears the weight of the fuel rod and the assembly. At the same time, the lower tube seat is the coolant inlet. The coolant flows through the gap between the fuel rods to achieve cooling. The lower tube seat indirectly supports the lower end plug by supporting the fuel rod. The two cooperate with each other in structure and function to jointly ensure the normal operation of the fuel assembly, ensure the uniform circulation of the coolant and the stability and sealing of the fuel rod.

[0062] Due to its large size and contact-based measurement requirements, existing testing equipment is only suitable for weld inspection of individual fuel rods before they are assembled into fuel assemblies. However, the distance between the fuel rods in a completed nuclear fuel assembly is only approximately 3 mm, and the distance between the rods and the lower tube seat is only approximately 10 mm, making existing testing equipment unable to perform these measurements. Therefore, embodiments of the present application provide a nuclear fuel rod lower end plug weld inspection system and method for inspecting the welds of individual nuclear fuel rod lower end plugs in assembled nuclear fuel assemblies.

[0063] Figure 2 A structural diagram of a weld detection system for a nuclear fuel rod lower end plug provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the weld inspection system includes a data acquisition module and a data analysis module. The data acquisition module includes a camera, a motion mechanism, and an endoscopic inspection probe mounted on the motion mechanism, wherein an endoscope for recording endoscopic video is arranged in the endoscopic inspection probe.

[0064] In this embodiment, the camera is used to capture a sequence of environmental images after the nuclear fuel assembly is hovering, and the sequence of environmental images is sent to the data analysis module. The motion mechanism is used to receive a motion control instruction issued by the data analysis module, and move the endoscopic detection probe to the weld detection position in response to the motion control instruction. The endoscopic detection probe is used to record an endoscopic video after receiving a first instruction, and send the endoscopic video to the data analysis module. The data analysis module is used to: receive the sequence of environmental images, generate a motion control instruction based on the sequence of environmental images, and send the motion control instruction to the motion mechanism. In response to the endoscopic detection probe reaching the weld detection position, a video recording instruction is issued, and the video recording instruction includes a first instruction, which instructs the endoscopic detection probe to start recording. The endoscopic video is received, and based on the endoscopic video, whether the weld state of the lower end plug to be inspected is normal is detected to obtain the weld inspection result of the lower end plug to be inspected.

[0065] Combine Figure 3 This paper introduces the functions and functional interactions of the weld inspection system.

[0066] Figure 3 A schematic flow chart of a method for detecting a weld of a lower end plug of a nuclear fuel rod provided in an embodiment of the present application is shown as follows: Figure 3 As shown, the method includes:

[0067] S301: After the camera hovers over the nuclear fuel assembly, it captures an environmental image sequence and sends the environmental image sequence to a data analysis module.

[0068] In this embodiment, the environment image sequence includes at least one environment image.

[0069] S302: Receive an environmental image sequence, generate a motion control instruction based on the environmental image sequence, and send the motion control instruction to a motion mechanism.

[0070] S303, the motion mechanism is used to receive the motion control instruction issued by the data analysis module, and move the endoscopic detection probe to the weld detection position in response to the motion control instruction.

[0071] S304: The data analysis module issues a video recording instruction in response to the endoscopic detection probe reaching the weld detection position.

[0072] In this embodiment, the video recording instruction includes a first instruction, and the first instruction instructs the endoscopic detection probe to start recording.

[0073] S305. After receiving the first instruction, the endoscopic detection probe is used to record the endoscopic video and send the endoscopic video to the data analysis module.

[0074] S306 , receiving the endoscope video, detecting whether the weld state of the lower end plug to be inspected is normal based on the endoscope video, and obtaining the weld inspection result of the lower end plug to be inspected.

[0075] As can be seen from the above technical solution, the present application provides a weld inspection system and method for the lower end plug of a nuclear fuel rod. The weld inspection system includes a data acquisition module and a data analysis module, wherein the data acquisition module includes a camera, a motion mechanism, and an endoscopic inspection probe mounted on the motion mechanism. The data analysis module locates the weld inspection position based on the environmental image of the target lower end plug to be inspected, captured by the camera after hovering over the nuclear fuel assembly. A motion control instruction is issued to control the motion mechanism to carry the endoscopic inspection probe to the weld inspection position. The endoscopic inspection probe is controlled to reach the weld inspection position of the target lower end plug to be inspected, and an endoscopic video is collected. Due to the small size of the endoscopic inspection probe, it penetrates the weld inspection position, overcoming the limitations of the small space and achieving full coverage of the image of the weld to be identified of the target lower end plug to be inspected. Furthermore, the data analysis module implements weld inspection based on the endoscopic video, thereby improving the accuracy of weld inspection.

[0076] Reference Figure 4 , Figure 4 A schematic diagram of the specific structure of a weld detection system for a nuclear fuel rod lower end plug provided in an embodiment of the present application is shown in FIG. Figure 4 As shown, the weld detection system of the lower end plug of the nuclear fuel rod includes a data acquisition module and a data analysis module. The data acquisition module includes a camera, a motion mechanism and an endoscopic detection probe installed on the motion mechanism. The endoscopic detection probe is arranged with an endoscope for recording endoscopic video.

[0077] In this embodiment, the camera is an industrial camera configured with preset parameters, and the industrial camera is fixed on the motion mechanism.

[0078] In this embodiment, the motion mechanism consists of a motor module, a telescopic mechanism, a horizontal displacement mechanism, a vertical displacement mechanism, and a rotating device. The telescopic mechanism, in the form of an electric push rod, is mounted on the fixed plate of the vertical displacement mechanism. The vertical displacement mechanism is mounted on the fixed support of the horizontal displacement mechanism. The horizontal displacement mechanism is mounted on the rotating device, which is fixed to the foundation at a preset position via a mounting bracket. The motor module includes linear motor modules of different sizes.

[0079] In this embodiment, the horizontal displacement mechanism, the vertical displacement mechanism, the telescopic mechanism, and the rotation device respectively realize motion control through linear motor modules of different sizes. Specifically, the rotation device realizes rotational motion parallel to the horizontal plane through the corresponding linear motor module, and the rotation device can work in the 0° and 90° states. The horizontal displacement mechanism realizes displacement motion in the horizontal direction along the fixed support through the corresponding linear motor module. The vertical displacement mechanism realizes displacement motion in the vertical direction through the corresponding linear motor module. The telescopic mechanism realizes telescopic motion perpendicular to the horizontal and vertical directions through the corresponding linear motor module. It should be noted that the number of telescopic mechanisms can be multiple. For example, four telescopic mechanisms are arranged side by side in the horizontal direction of the fixed support, and the adjacent spacing between the telescopic mechanisms is the center distance of the fuel rods. The telescopic motion of the four telescopic mechanisms is synchronous motion.

[0080] In this embodiment, the number of endoscopic detection probes is equal to the number of telescopic mechanisms. Each endoscopic detection probe is compactly arranged at the end of each telescopic mechanism in a reflective manner. The initial position of the endoscopic detection probe is in a first relative positional relationship with the industrial camera. Specifically, the end of the telescopic mechanism carries an endoscope adapter rod, and the endoscopic detection probe and the endoscope adapter rod are fixed together via a threaded connection. Figure 5 A schematic diagram of the structure of an endoscopic detection probe provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the endoscopic detection probe is composed of a reflector, an endoscope, and a lighting module, wherein the endoscope includes an endoscope body and an endoscope lens installed inside the endoscope body, multiple lighting modules are arranged in a ring around the endoscope lens, and the reflector is installed at one end of the endoscope body with a preset reflection angle to adjust the propagation direction of light.

[0081] In this embodiment, the endoscopic detection probe achieves efficient space utilization and reduced light loss through a reflective compact layout, reducing the volume of the endoscopic probe while ensuring effective light transmission and clear imaging of the endoscopic video, thereby improving the applicability of the weld detection system for the lower end plug of the nuclear fuel rod.

[0082] It should be noted that the field of view angle of the endoscopic detection probe imaging is σ sys Assuming the optical imaging distance from the endoscope probe to the bottom of the lower end plug is A, and the distance from the weld of the lower end plug on the fuel rod is B, then according to the principle of camera pinhole imaging, the field of view of the reflector fov sys =2(A+B)×tan(σ sys / 2), and the field of view must cover the entire lower end plug area and weld. On this basis, the imaging resolution should be improved as much as possible.

[0083] Furthermore, the embodiment of the present application provides a method for detecting welds of the lower end plug of a nuclear fuel rod applied to a data analysis module. Figure 6A specific implementation flow chart of a method for detecting the weld of a nuclear fuel rod lower end plug provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the method may specifically include steps S601 to S609, and these steps are described in detail below.

[0084] S601: Receive a sequence of environmental images captured by an industrial camera after the nuclear fuel assembly is hovering.

[0085] In this embodiment, the environmental image sequence includes multiple environmental images, and the environmental images include real-time images of nuclear fuel assemblies.

[0086] In this embodiment, when the nuclear fuel assembly is in a hovering and stationary state, the lower end plug positioning module is started, and a start shooting instruction is sent to the industrial camera through the lower end plug positioning module to control the industrial camera to capture an environmental image sequence at a preset image acquisition frequency and feed back the environmental image sequence.

[0087] S602: Identify the coordinates of the weld detection position based on the environmental image sequence.

[0088] In this embodiment, the weld detection position is the bottom end of the target lower end plug to be detected, that is, the gap position between the target lower end plug to be detected and the lower pipe seat, wherein the target lower end plug to be detected is any one of the lower end plugs to be detected, and the gap position is used as the starting detection position, that is, the weld detection position.

[0089] In this embodiment, the lower end plug of the leftmost nuclear fuel rod on the side of the nuclear fuel assembly facing the industrial camera is the target lower end plug to be inspected, and the weld inspection position is the gap position between the lower end plug of the leftmost nuclear fuel rod and the lower tube seat.

[0090] In this embodiment, the fuel assembly transport container is hoisted into a horizontal state, and after the container cover is opened, the fuel assembly is hoisted into a vertical state. In this state, the operator controls the lowering of the fuel assembly clamp. During this process, the industrial camera is called to capture the environmental image and the environmental image sequence is transmitted back to the back-end data analysis module. The data analysis module performs object recognition based on the environmental image. The recognized objects include the clamp, the upper tube seat and the lower tube seat. The clamp, the upper tube seat and the lower tube seat are recognized in turn. After the lower tube seat is recognized, the preparation work for the lower end plug detection is started, that is, the recognition of the relative position relationship is started.

[0091] S603 : Based on the relative positional relationship between the fixed position of the industrial camera and the real-time position of the endoscopic detection probe, map the real-time position of the endoscopic detection probe into a coordinate system to obtain the coordinates of the real-time position of the endoscopic detection probe.

[0092] S604. Based on the coordinates of the weld detection position and the coordinates of the real-time position of the endoscopic detection probe, a motion control instruction is obtained, and the motion control instruction is sent to the motion mechanism to control the motion mechanism to move the endoscopic detection probe to the weld detection position in response to the motion control instruction.

[0093] In this embodiment, the real-time position of the endoscopic detection probe is the initial position, that is, at the initial moment of each detection, the endoscopic detection probe is controlled to perform a reset operation to the initial position, or after each detection is completed, the endoscopic detection probe is controlled to perform a reset operation to the initial position.

[0094] In this embodiment, the number of lower end plugs on the target side of the fuel assembly (the side facing the industrial camera) and the pixel distribution of the lower end plugs in the image are identified through the environmental image, thereby determining the coordinates of the weld inspection position corresponding to the target lower end plug to be inspected in the coordinate system corresponding to the industrial camera. Based on the relative positional relationship between the fixed position of the industrial camera and the initial position of the endoscopic inspection probe, the initial position of the endoscopic inspection probe is mapped to the coordinate system, thereby obtaining the relative positional relationship between the coordinates of the weld inspection position and the initial position of the endoscopic inspection probe.

[0095] In this embodiment, based on the relative position relationship, a motion control instruction is generated for moving the endoscopic inspection probe to the weld inspection position. The motion control instruction includes a horizontal displacement parameter, a vertical displacement parameter, and a telescopic parameter. The horizontal displacement parameter is determined based on the relationship between the weld inspection position and the abscissa of the endoscopic inspection probe, the vertical displacement parameter is determined based on the relationship between the weld inspection position and the ordinate of the endoscopic inspection probe, and the telescopic parameter can be a preset parameter related to the length of the telescopic mechanism and the distance from the fuel assembly.

[0096] Furthermore, the motion mechanism responds to the motion control instruction, controls the horizontal displacement mechanism to perform horizontal displacement according to the horizontal displacement parameter, controls the vertical displacement mechanism to perform vertical displacement according to the vertical displacement parameter, and controls the telescopic mechanism to perform telescopic displacement according to the telescopic parameter, thereby moving the endoscopic detection probe to the weld detection position.

[0097] S605: In response to the endoscopic inspection probe reaching the weld inspection position, a video recording instruction is issued.

[0098] In this embodiment, after the endoscopic detection probe reaches the weld detection position, the video recording instruction issued includes at least a first instruction for instructing the endoscopic detection probe to start recording, and also includes a second instruction for instructing the motion mechanism to move according to a preset path in the horizontal plane where the weld detection position is located, wherein the preset path is pre-configured based on the weld detection position and the distribution of nuclear fuel rods in the nuclear fuel assembly. For example, the target lower end plug to be inspected corresponding to the weld detection position is the lower end plug corresponding to the nuclear fuel rod on the leftmost front side. Then, the preset path can be that after the telescopic mechanism is extended to the leftmost rear side, the horizontal displacement mechanism is displaced to the right by one unit length, and the telescopic mechanism is shortened to the lower end of the nuclear fuel rod adjacent to the right end of the nuclear fuel rod on the leftmost front side, and so on and so forth until it reaches the lower end of the last nuclear fuel rod.

[0099] In this embodiment, the endoscopic detection probe starts recording endoscopic video in response to the first instruction, and the motion mechanism starts moving in response to the second instruction. That is, the endoscopic detection probe takes the weld detection position as the starting detection position, and records the endoscopic video according to the preset path on the horizontal plane where the weld detection position is located, thereby traversing and detecting all the lower end plugs to be detected, thereby achieving full coverage detection of the lower end plugs to be detected.

[0100] S606 , receiving an endoscopic video, and converting each endoscopic video frame from a ring view image to a rectangular view image based on a polar coordinate conversion relationship.

[0101] In this embodiment, since the endoscopic detection probe is imaged in a polar coordinate manner with a circular layout, that is, each frame of the endoscopic video frame in the endoscope video stream is a circular view image, therefore, for each frame of the endoscopic video frame, based on the polar coordinate conversion relationship, the endoscopic video frame is converted from a circular view image to a rectangular view image.

[0102] In this embodiment, the center of the lower end plug is taken as the origin O, the preset length R and the preset angle positive direction, the width of the rectangular view after the annular view image is expanded is recorded as w, the height is recorded as h, and any pixel point P in the plane of the annular view image is taken, the angle between OP and the polar axis is recorded as θ, and the length of OP, that is, the polar diameter is recorded as γ.

[0103] Based on the polar coordinate conversion relationship, the pixel point P is mapped to the horizontal axis of the rectangular view image to obtain the horizontal coordinate of the mapped pixel point P Map the ring pixel point P to the vertical axis of the rectangular view image to obtain the vertical coordinate of the mapped pixel point of the ring pixel point P Wherein, r1 and r2 represent the inner radius and outer radius of the annular view image respectively.

[0104] Figure 7 A schematic diagram of the conversion of an endoscopic image provided in an embodiment of the present application is shown as follows: Figure 7As shown, based on the polar coordinate conversion relationship, the endoscope video frame is converted from an annular view image to a rectangular view image, that is, each pixel in the annular view image is mapped to the rectangular view image.

[0105] S607 : For a target pixel in the rectangular view image, optimize the pixel based on the grayscale values of each pixel in a preset neighborhood of the target pixel to obtain an optimized rectangular view image.

[0106] In this embodiment, the target pixel point (u x ,v y ) is any pixel point in the rectangular view image, and the method for optimizing the pixel point based on the grayscale value of each pixel point in a preset neighborhood includes:

[0107] A1. Obtain the grayscale value of each pixel in the preset neighborhood of the target pixel, where the preset neighborhood is a square area with the target pixel as the center and the preset value as the side length. Taking the preset value n as an example, the coordinates of each pixel in the preset neighborhood are marked as (u x+i ,v y+j ), where i is not greater than n and j is not greater than n. When n=2, i,j=-1,0,1.

[0108] A2. Obtain the weight of each pixel in the preset neighborhood.

[0109] In this embodiment, the pixel points (u x+i ,v y+j ) The calculation method is:

[0110]

[0111] in, Represents the weight of each pixel in the area, k i,j Indicates: pixel (u x+i ,v y+j ) and the current pixel (u x ,v y ), where i, j = -1, 0, 1.

[0112] A3. Based on the grayscale value and weight of each pixel in the preset neighborhood, a cubic polynomial is used to fit the grayscale optimization value of the target pixel.

[0113] In this embodiment, the grayscale optimization value f(u x ,v y )′ is as follows:

[0114]

[0115] Among them, f(u x+i ,v y+j ) represents the target pixel (u x ,v y ) of the pixel points in the area (u x+i ,v y+j ) grayscale value, f(u x ,v y )′ represents the target pixel (u x ,v y )The grayscale value after fitting optimization is also called grayscale optimization value.

[0116] A4. Update the pixel value of the target pixel in the rectangular view image to the grayscale optimization value.

[0117] Through A1 to A4, all pixel points on the rectangular view image are optimized to obtain an optimized rectangular view image, thereby improving the smoothness of the rectangular view image and making the defect edge more natural.

[0118] S608 , performing displacement correction on the rectangular view image to obtain a weld image to be identified.

[0119] In this embodiment, the specific method of performing displacement correction on the rectangular view image to obtain the weld image to be identified includes:

[0120] B1. Obtaining the initial strain distribution of the rectangular view image.

[0121] In this embodiment, the deformation of each pixel in the rectangular view image is estimated to obtain an initial strain distribution. Specifically, the strain can be roughly determined by comparing the positional changes of corresponding points in the expanded image with those in the ideal rectangular image. The strain distribution includes displacement components for each pixel, which include horizontal and vertical displacement components.

[0122] B2. For a pixel point, calculate the strain component of the pixel point based on the initial displacement component of the pixel point.

[0123] In this embodiment, the strain components include normal strain and shear strain.

[0124] Specifically, taking the target pixel as an example, the initial horizontal displacement component and vertical displacement component of the target pixel are u and v respectively, then the normal strain ε of the target pixel is calculated as x , ε y , and shear strain γ xy See the following formula for the method:

[0125]

[0126] B3. Based on the strain components of each pixel point, the strain field of the rectangular view image is obtained.

[0127] B4. Substitute the strain field of the rectangular view image into the preset elastic constitutive equation to obtain the stress component of each pixel.

[0128] In this embodiment, the stress components include normal stress (denoted as σ x and σ y ) and shear stress (τ xy ).

[0129] Specifically, the strain field of the rectangular view image is substituted into the preset elastic constitutive equation to obtain the stress component, as shown in the following formula:

[0130]

[0131] Wherein, E represents the elastic modulus and μ represents the Poisson's ratio, both of which can be obtained through empirical values.

[0132] B5. Substitute the stress component of each pixel into the preset equilibrium equation to obtain the displacement component of each pixel.

[0133] In this embodiment, the displacement component includes a displacement value and a direction.

[0134] Specifically, the stress component of each pixel is substituted into the preset equilibrium equation to obtain the displacement component of each pixel, as shown in the following equation:

[0135]

[0136] Furthermore, the finite element analysis method is used to solve the equation to obtain the displacement component, that is, to determine the displacement value and direction that each pixel point in the rectangular view image needs to move.

[0137] B6. For each pixel point, move the pixel point according to the displacement component to obtain the weld image to be identified.

[0138] It's important to understand that the unfolding of the annular bottom view of the lower end plug of a nuclear fuel rod can be likened to the deformation of an elastic body. Typically, the unfolded annular bottom view may deviate from its ideal rectangular shape due to the weld location and image viewing angle, much like an elastic body deforming under external force. According to the principles of elasticity, after an elastic body deforms, applying a force opposite to the deforming force can restore it to its original shape, or a near-original shape. Here, the deformation of the lower end plug image during the transformation from annular to rectangular is corrected by applying an opposing elastic force. The goal is to make the unfolded image closer to the ideal rectangle, thereby improving image quality and enabling better weld anomaly identification. The specific extent of the tensile and compressive deformations observed after the unfolding of the annular bottom view requires knowledge. Strain is a physical quantity that describes the degree of deformation of an object. By estimating the deformation of various components in the unfolded image, the strain field can be determined. The strain field represents the distribution of deformation across the entire image area. After determining the strain field, the displacement field is solved using the elastic constitutive law, which describes the relationship between stress, strain, and displacement in an elastic body. Using a known strain field and the relevant equations and principles of elastic constitutive relations, we can calculate the displacement that each point in the image should produce, thus obtaining a displacement field. This displacement field can determine the position to which each point in the image should be moved, thereby achieving image deformation correction.

[0139] S609: Input the weld image to be identified into the anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result.

[0140] In this embodiment, if the weld detection result is abnormal, the abnormality detection model outputs the weld recognition image to be identified with the abnormal position marked, that is, the weld recognition image. If the weld detection result is normal, the abnormality detection model outputs the weld image to be identified.

[0141] like Figure 7 As shown in Figure 1, after mapping each pixel in the annular view image to the rectangular view image, the gray abnormal area in the annular view image is mapped to the white rectangular area in the rectangular view image. The anomaly detection model outputs the weld image to be identified, with the white rectangular area marked.

[0142] In this embodiment, the anomaly detection model performs anomaly recognition on the image blocks of the weld image to be identified to obtain abnormal image blocks, and determines whether there is an abnormality in the weld image to be identified.

[0143] In this embodiment, the anomaly detection model uses a multi-layer attention mechanism as a feature extractor and removes some layers of the network to adapt to specific task requirements. During the training process, in order to fully utilize the structural advantages of the model, the input image is processed in a block-by-block manner. After the image is divided into multiple small blocks, each small block is processed separately, and then features are extracted from different layers of the multi-layer attention mechanism. These features from different layers contain information about the image at different levels and different levels of abstraction. These features can be used to estimate the parameters of the multivariate Gaussian distribution. The function expression is as follows:

[0144]

[0145] in, Represents the feature vector of the image block after feature extraction, η m It represents the mean vector, which reflects the center position of the multivariate Gaussian distribution in each dimension. ε represents the covariance rectangle, which is used to describe the correlation between each dimension and the degree of dispersion of the data.

[0146] In this embodiment, the method for obtaining a weld recognition image based on weld detection results using an anomaly detection model includes:

[0147] C1. Divide the weld image to be identified into blocks through the anomaly detection model to obtain multiple image blocks.

[0148] C2. Extract the feature vector of the image block through the anomaly detection model based on the multi-layer attention mechanism, substitute the feature vector of the image block into the probability density function of the multivariate Gaussian distribution, and obtain the probability that the image block belongs to the normal class.

[0149] C3. Obtain the anomaly detection result of the image block based on probability through the anomaly detection model.

[0150] In this embodiment, to improve detection accuracy and reliability, it is also possible to consider comprehensively analyzing the anomaly detection results of multiple adjacent image blocks, using a voting mechanism or a fusion method based on neighborhood information, to prevent the misjudgment of a single image block from significantly affecting the overall detection results. Specifically, the detection result of each image block is considered a "vote," and the final detection result is determined based on the voting results of multiple adjacent image blocks. For example, a threshold can be set so that only when more than a certain proportion of adjacent image blocks are detected as anomalies is the area considered anomaly.

[0151] Then, taking any image block, that is, the target image block, as an example, the method for obtaining the abnormality detection result of the image block based on probability includes:

[0152] C31. Calculate the feature similarity between the target image block and each adjacent image block.

[0153] In this embodiment, the feature similarity is determined based on the cosine distance. The higher the feature similarity, the more similar the two image blocks are.

[0154] C32. Determine the weight of each adjacent image block based on the feature similarity between the target image block and the adjacent image blocks, wherein the weight is positively correlated with the feature similarity.

[0155] C33. Based on the weights of each adjacent image block, the probability that the adjacent image blocks belong to the normal class is weighted averaged to obtain the neighborhood anomaly score of the image block. The neighborhood anomaly score and the probability that the target image block belongs to the normal class are fused to obtain the anomaly score fusion result of the target image block.

[0156] C34. If the anomaly score fusion result is greater than the probability threshold, the image block is a normal image block; if the anomaly score fusion result is not greater than the preset probability threshold, the image block is an abnormal image block.

[0157] In summary, the anomaly detection result of a single image block not only depends on its own features, but is also affected by its adjacent image blocks.

[0158] C4. Use the anomaly detection model to locate the position of the abnormal image block in the weld image to be identified, and mark the abnormal image block at the corresponding position to obtain the weld identification image and output it.

[0159] It should be noted that the setting of the probability threshold usually needs to be adjusted based on the actual application scenario and the characteristics of the dataset. Generally, a reasonable threshold can be determined by testing a certain number of normal samples and known abnormal samples and observing the probability distribution. For example, the threshold can be set to an initial value, and then a cross-validation method can be used to evaluate the accuracy, recall rate and other indicators of the model under different thresholds on the validation set. By continuously adjusting the threshold, the threshold that optimizes the model performance is found. When the calculated probability that the feature vector belongs to the normal class is less than the probability threshold, it means that the distribution of the features of the image block is significantly different from that of the normal class features, and it is likely to contain anomalies, so it is identified as an abnormal result. Conversely, if the probability is greater than or equal to the threshold, the features of the image block are considered to conform to the distribution of the normal class and it is judged to be normal.

[0160] As can be seen from the above technical solutions, the weld detection method for the lower end plug of a nuclear fuel rod provided in the embodiment of the present application has the following technical effects:

[0161] The main advantage of the anomaly detection model in this scheme is that it can be trained without the need for backpropagation and only uses normal samples to build the model, which makes it perform well when dealing with new anomalies that have not been seen before.

[0162] First, many traditional anomaly detection methods, especially those based on deep learning, typically require a large number of normal and abnormal samples for supervised or semi-supervised training. In industrial applications, obtaining a large number of abnormal samples is often difficult and costly due to the inherent uncertainty and rarity of abnormal situations. For example, some traditional neural network-based methods require continuous parameter adjustment through backpropagation to minimize the loss function, which relies on a complete dataset containing both positive and negative samples.

[0163] The anomaly detection model in this solution can be trained without backpropagation, using only normal samples to build the model. This significantly reduces reliance on abnormal samples and addresses the scarcity of abnormal samples in industrial scenarios. By learning the feature distribution from normal samples, a multivariate Gaussian distribution model is constructed to represent the normal class.

[0164] Specifically, when there are few abnormal samples, the anomaly recognition model uses a generative network model to learn the data's feature distribution. For anomaly recognition, the patterns of normal sample data are learned, and then anomalies are identified by evaluating how new test samples differ from the learned patterns. During model training, the input samples are encoded as a distribution in the latent space, and the data is reconstructed from this distribution. By minimizing the reconstruction error and the regularization term, a feature representation of normal data is learned. During the inference and testing phase, the reconstruction error of new samples can be used as an anomaly confidence level. A high reconstruction error indicates that the data differs significantly from normal data and may be anomaly. This approach does not require backpropagation to adjust model parameters because it learns the data distribution by optimizing the objective function. This simplifies the training process, reduces the time and space complexity of training, and improves model training efficiency. Furthermore, cluster analysis is performed only based on the characteristics of the positive samples themselves, eliminating the need to annotate abnormal regions to guide training. This reduces annotation costs.

[0165] Second, existing anomaly detection methods may rely on simple threshold comparisons or class labels output by a classifier to determine anomalies. For example, threshold-based methods may compare the value of a single feature with a fixed threshold to determine whether an anomaly is present. This approach lacks consideration of the overall distribution of the data and is prone to misjudgment. Classifier-based methods also struggle to accurately distinguish between normal and abnormal situations if the classifier's training data is insufficient or feature extraction is incomplete.

[0166] The anomaly detection model in this solution uses a multivariate Gaussian distribution to obtain a probabilistic representation of the normal class to determine anomalies. The input image's feature vector is substituted into the probability density function of the multivariate Gaussian distribution to calculate its probability of belonging to the normal class, which is then compared with a probability threshold. This approach fully considers the distribution of normal class data and measures the difference between the input and the normal class from a probabilistic perspective, enabling more accurate anomaly identification and improved generalization, especially for new and unseen anomalies.

[0167] Third, existing image processing techniques may analyze the entire image holistically or employ a simple block-based approach without fully leveraging the relationships between blocks. For complex weld images to be identified, this approach may not accurately locate abnormal areas and struggles to capture feature information at different scales.

[0168] The anomaly detection model in this solution processes the input image in a block-by-block manner during training, extracting features from different layers of a multi-layer attention mechanism. This block-by-block processing allows for a more detailed analysis of the image's local features. Features at different layers contain information about the image at different levels and abstractions. By combining these features to estimate the parameters of the multivariate Gaussian distribution, we can more comprehensively describe the image's feature distribution, leading to more accurate detection and location of anomalies.

[0169] It should be noted that Figure 6 This is only an optional specific implementation of the method for detecting the weld seam of the lower end plug of the nuclear fuel rod applied to the data analysis module. The data analysis module can also realize the weld seam detection of the lower end plug of the nuclear fuel rod through other specific implementation processes.

[0170] For example, in an optional embodiment, the method for obtaining an abnormality detection result of an image block based on probability is: if the probability is greater than a probability threshold, the image block is a normal image block; if the probability is not greater than a preset probability threshold, the image block is an abnormal image block.

[0171] For another example, the weld detection system of the lower end plug of a nuclear fuel rod may include multiple endoscopic detection probes installed in parallel on a motion mechanism, each endoscopic detection probe synchronously collects endoscopic video, and the data analysis module improves the accuracy of anomaly detection by performing anomaly detection on the endoscopic video taken by each endoscopic detection probe.

[0172] In summary, the weld detection method for the lower end plug of a nuclear fuel rod provided in the embodiment of the present application is summarized as follows: Figure 8 The process shown in the figure is applied to the data analysis module of the weld detection system of the lower end plug of the nuclear fuel rod. The weld detection system of the lower end plug of the nuclear fuel rod includes a data acquisition module and a data analysis module. The data acquisition module includes a camera, a motion mechanism and an endoscopic detection probe installed on the motion mechanism. Figure 8 As shown, a method for detecting the weld of a nuclear fuel rod lower end plug includes:

[0173] S801: Receive a sequence of environmental images captured by a camera after the nuclear fuel assembly hovers.

[0174] S802 : Generate a motion control instruction based on the environmental image sequence, and send the motion control instruction to the motion mechanism, so that the motion mechanism moves the endoscopic detection probe to the weld detection position in response to the motion control instruction.

[0175] S803: In response to the endoscopic inspection probe reaching the weld inspection position, a video recording instruction is issued.

[0176] In this embodiment, the video recording instruction includes a first instruction, and the first instruction instructs the endoscopic detection probe to start recording.

[0177] S804: Receive the endoscope video, and obtain the weld image to be identified of the lower end plug to be inspected based on the endoscope video.

[0178] S805: Input the weld image to be identified into the anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result.

[0179] In this embodiment, if the weld detection result shows that an abnormality exists, the weld recognition image is an image of the weld to be recognized with the abnormal position marked.

[0180] It should be noted that the specific implementation of each step of a method for detecting the weld of a nuclear fuel rod lower end plug may refer to the above embodiments.

[0181] An electronic device is also provided in an embodiment of the present application. Figure 9 As shown, Figure 9 It shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include but is not limited to fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 9 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0182] like Figure 9As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 into a random access memory (RAM) 903. When the electronic device is powered on, the RAM 903 also stores various programs and data required for the operation of the electronic device. The processing device 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0183] Typically, the following devices may be connected to the I / O interface 905: an input device 906 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 907 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 908 including, for example, a memory card, a hard disk, etc.; and a communication device 909. The communication device 909 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Figure 9 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0184] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the methods for detecting the weld of the lower end plug of a nuclear fuel rod provided in the embodiment of the present application.

[0185] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any one of the weld detection methods for the lower end plug of a nuclear fuel rod provided in an embodiment of the present application.

[0186] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0187] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

[0188] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0189] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a training device or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website, a computer, a training device or a data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. A nuclear fuel rod lower end plug weld detection system, characterized in that: include: A data acquisition module and a data analysis module, wherein the data acquisition module includes a camera, a motion mechanism, and an endoscopic detection probe mounted on the motion mechanism; The camera is used to capture a sequence of environmental images after the nuclear fuel assembly is hovering, and send the sequence of environmental images to the data analysis module; The motion mechanism is used to receive the motion control instruction issued by the data analysis module, and move the endoscopic detection probe to the weld detection position in response to the motion control instruction; The endoscopic detection probe is used to record the endoscopic video after receiving the first instruction and send the endoscopic video to the data analysis module; The data analysis module is used to: receiving the environmental image sequence, generating the motion control instruction based on the environmental image sequence, and sending the motion control instruction to the motion mechanism; In response to the endoscopic detection probe reaching the weld detection position, issuing a video recording instruction, the video recording instruction including the first instruction, the first instruction instructing the endoscopic detection probe to start recording; The endoscope video is received, and based on the endoscope video, whether the weld state of the lower end plug to be inspected is normal is detected to obtain the weld inspection result of the lower end plug to be inspected.

2. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 1, characterized in that: The endoscopic detection probe is composed of a reflector, an endoscope, and multiple lighting modules, wherein the endoscope includes an endoscope body and an endoscope lens installed at the first end inside the endoscope body, multiple lighting modules are arranged in a ring around the endoscope lens, and the reflector is installed at the second end outside the endoscope body at a preset reflection angle.

3. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 1, characterized in that: The data analysis module is used to receive the environmental image sequence, and when generating the motion control instruction based on the environmental image sequence, is specifically used to: identifying the coordinates of the weld detection position based on the environmental image sequence; Based on the relative positional relationship between the fixed position of the camera and the real-time position of the endoscopic detection probe, the real-time position of the endoscopic detection probe is mapped into a coordinate system to obtain the coordinates of the real-time position of the endoscopic detection probe. The motion control instruction is obtained based on the coordinates of the weld detection position and the coordinates of the real-time position of the endoscopic detection probe. The motion control instruction includes a horizontal displacement parameter, a vertical displacement parameter, and a telescopic parameter.

4. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 3, characterized in that: The motion mechanism is composed of a motor module, a telescopic mechanism, a horizontal displacement mechanism, a vertical displacement mechanism, and a rotating device. The telescopic mechanism is in the form of an electric push rod and is mounted on the fixed plate of the vertical displacement mechanism. The vertical displacement mechanism is mounted on the fixed support of the horizontal displacement mechanism. The horizontal displacement mechanism is mounted on the rotating device. The rotating device is fixed to the foundation at a preset position through a mounting bracket. The motion mechanism is used to receive the motion control instruction issued by the data analysis module, and when responding to the motion control instruction to move the endoscopic detection probe to the weld detection position, is specifically used to: In response to the motion control instruction, the motor module controls the horizontal displacement mechanism to perform horizontal displacement according to the horizontal displacement parameter, controls the vertical displacement mechanism to perform vertical displacement according to the vertical displacement parameter, and controls the telescopic mechanism to perform telescopic displacement according to the telescopic parameter, so as to move the endoscopic detection probe to the weld detection position.

5. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 4, characterized in that: The video recording instruction further includes a second instruction, wherein the second instruction instructs the motion mechanism to move along a preset path in the horizontal plane where the weld detection position is located; The motion mechanism is further configured to respond to the second instruction and control the telescopic mechanism and the horizontal displacement mechanism to move along the preset path through the motor module.

6. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 1, characterized in that: The data analysis module is used to receive the endoscope video, detect whether the weld state of the lower end plug to be inspected is normal based on the endoscope video, and obtain the weld inspection result of the lower end plug to be inspected, specifically for: receiving the endoscope video, and obtaining an image of a weld to be identified of the lower end plug to be inspected based on the endoscope video; The weld image to be identified is input into an anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result. If the weld detection result is abnormal, the anomaly detection model outputs a weld identification image embedded with an abnormal image block; if the weld detection result is normal, the anomaly detection model outputs the weld image to be identified.

7. The weld detection system for the lower end plug of a nuclear fuel rod according to claim 6, characterized in that: When the data analysis module is used to obtain the weld image to be identified of the lower end plug to be inspected based on the endoscopic video, it is specifically used to: For each endoscopic video frame, converting the endoscopic video frame from a ring view image to a rectangular view image based on a polar coordinate conversion relationship; Optimizing each pixel of the rectangular view image based on the grayscale value of each pixel in a preset neighborhood to obtain an optimized rectangular view image; The optimized rectangular view image is subjected to displacement correction to obtain the weld image to be identified.

8. The weld inspection system for the lower end plug of a nuclear fuel rod according to claim 7, characterized in that: The data analysis module is used to input the weld image to be identified into the anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result, specifically for: Inputting the weld image to be identified into the anomaly detection model; Dividing the weld image to be identified into blocks using the anomaly detection model to obtain multiple image blocks; Extracting a feature vector of the image block using the anomaly detection model based on a multi-layer attention mechanism, substituting the feature vector of the image block into a probability density function of a multivariate Gaussian distribution to obtain a probability that the image block belongs to a normal class; determining whether the image block is an abnormal image block based on the probability by using the abnormality detection model; The abnormal image block is located in the weld image to be identified by using the abnormality detection model, and the abnormal image block is marked at the corresponding position to obtain the weld identification image and output it.

9. The nuclear fuel rod lower end plug weld detection system according to claim 8, characterized in that: When the data analysis module determines whether the image block is an abnormal image block based on the probability by using the abnormality detection model, it is specifically used to: Calculating feature similarities between an image block and its adjacent image blocks using the anomaly detection model; Determining the weight of each of the adjacent image blocks according to the feature similarity using the anomaly detection model, wherein the weight is positively correlated with the feature similarity; Performing a weighted average of the probabilities that the adjacent image blocks belong to the normal class based on the weights of the adjacent image blocks by the anomaly detection model to obtain a neighborhood anomaly score of the image block; fusing the neighborhood anomaly score and the probability that the image block belongs to the normal class through the anomaly detection model to obtain an anomaly score fusion result of the image block; If the abnormality score fusion result is greater than a probability threshold, the image block is determined to be a normal image block; if the abnormality score fusion result is not greater than a preset probability threshold, the image block is determined to be an abnormal image block.

10. A method for detecting the weld of a nuclear fuel rod lower end plug, characterized in that: A data analysis module is applied to a weld inspection system for a nuclear fuel rod lower end plug, the weld inspection system comprising a data acquisition module and the data analysis module, wherein the data acquisition module comprises a camera, a motion mechanism, and an endoscopic detection probe mounted on the motion mechanism; The method for detecting the weld of the lower end plug of the nuclear fuel rod comprises: receiving a sequence of environmental images captured by the camera after the nuclear fuel assembly is hovering; generating a motion control instruction based on the environmental image sequence, and sending the motion control instruction to the motion mechanism so that the motion mechanism moves the endoscopic detection probe to a weld detection position in response to the motion control instruction; In response to the endoscopic detection probe reaching the weld detection position, issuing a video recording instruction, the video recording instruction including a first instruction, the first instruction instructing the endoscopic detection probe to start recording; receiving an endoscopic video recorded by the endoscopic detection probe, and obtaining an image of a weld to be identified of the lower end plug to be detected based on the endoscopic video; The weld image to be identified is input into an anomaly detection model to obtain a weld identification image output by the anomaly detection model based on the weld detection result. If the weld detection result shows that an anomaly exists, the weld identification image is the weld image to be identified with the abnormal position marked.

Images