Nuclear power plant pipeline weld joint quality detection method, device and equipment and storage medium

Abstract

The invention relates to the technical field of welding seam detection, in particular to a nuclear power plant pipeline welding seam quality detection method, device and equipment and a storage medium. A welding seam area image of a pipeline welding seam area of a nuclear power plant is obtained through negative film digitization or digital rays, and the collected welding seam area image is preprocessed; and finally, the defect category of the weld seam image is automatically identified through the trained weld seam defect detection model, and the weld seam quality detection efficiency of the nuclear power station pipeline is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of weld detection, and in particular to a method, device, equipment and storage medium for detecting the quality of welds in nuclear power plant pipelines. Background Art

[0002] The safety of nuclear power plant equipment is directly related to the safe operation of the nuclear power plant. The quality of welds, as the connection parts of key equipment in nuclear power plants, directly affects the reliability and safety of the equipment. The quality inspection of welds plays a vital role in the construction and operation of nuclear power plants. At the same time, the weld inspection process of pipelines in nuclear power plants must not affect the safe operation of the equipment. Traditional radiographic inspection film evaluation usually relies on manual work, and operators need to have rich inspection experience and high professional skills. However, the manual film evaluation process is labor-intensive and time-consuming, and is easily affected by factors such as subjective judgment, fatigue, and distraction, which can easily lead to missed or misjudgment of defects. The automatic evaluation system for radiographic flaw detection results based on AI automatic recognition technology can effectively assist in film evaluation. It uses deep learning algorithms to intelligently identify and analyze defects in radiographic films, significantly improving the accuracy and consistency of defect identification and effectively reducing the risk of human error.

[0003] The above content is only used to assist in understanding the technical solution of the present invention and does not constitute an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of the present invention is to provide a method, device, equipment and storage medium for nuclear power plant pipeline weld quality inspection, aiming to solve the technical problems in the existing technology of low efficiency of manual evaluation of nuclear power plant pipeline weld negatives, occasional omissions and errors in evaluation, which pose hidden dangers to the safe operation of nuclear power plants.

[0005] To achieve the above object, the present invention provides a method for detecting the quality of nuclear power plant pipeline welds, the method comprising the following steps:

[0006] Performing film scanning and / or digital radiography on the nuclear power plant pipeline welds to obtain weld area images of the pipeline weld areas of the nuclear power plant;

[0007] Preprocessing the weld area image to obtain a target weld area image;

[0008] Extracting a target weld image from the target weld area image;

[0009] By using the trained weld defect detection model, defect detection is performed on the weld area of ​​the nuclear power plant pipeline according to the target weld image to obtain a weld detection result.

[0010] Optionally, performing film scanning and / or digital radiography on the nuclear power plant pipeline welds to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant includes:

[0011] In response to an image scanning request, determining a scanning mode, scanning parameters, and image identification information according to the image scanning request;

[0012] Scanning the weld X-ray film according to preset scanning parameters and the scanning mode to obtain an initial weld area image;

[0013] Performing image quality detection on the initial weld area image;

[0014] After the quality inspection is passed, the initial weld area image is named according to the image identification information to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant;

[0015] and / or

[0016] Obtaining weld parameters of a pipeline weld area of ​​the nuclear power plant;

[0017] In response to an image acquisition request, determining an acquisition time and a number of superimposed frames of an image acquisition device based on the image acquisition request, and generating image identification information;

[0018] Adjusting the voltage, current and exposure time of the image acquisition device based on the weld parameters;

[0019] The image acquisition device is driven based on the voltage, current, exposure time, acquisition time and number of superimposed frames to acquire a weld area image of a pipeline weld area of ​​the nuclear power plant.

[0020] Optionally, preprocessing the weld area image to obtain a target weld area image includes:

[0021] Denoising the weld area image using a preset filtering model;

[0022] Segmenting the weld area image to obtain a plurality of pixel blocks of the same size;

[0023] Count the grayscale histogram of each pixel block separately;

[0024] Performing peak clipping on the grayscale histogram of each pixel block based on a preset grayscale threshold to obtain a target grayscale histogram;

[0025] Calculating the cumulative distribution function of each pixel block based on the target grayscale histogram;

[0026] Performing grayscale mapping on each pixel block according to the cumulative distribution function to obtain a balanced local area image;

[0027] The balanced local area images corresponding to each pixel block are fused through a bilinear interpolation model to obtain an overall balanced image, and the overall balanced image is output as the target weld area image.

[0028] Optionally, extracting the target weld image from the target weld area image includes:

[0029] Extracting edge information from the target weld area image using an edge detection model to obtain a target weld area image;

[0030] Constructing a corresponding weld image matrix according to the target weld area image;

[0031] Performing grayscale conversion on the weld image matrix to obtain a grayscale conversion image matrix;

[0032] Performing Gaussian filtering on the grayscale conversion image matrix to obtain a filtered image matrix;

[0033] Performing threshold segmentation on pixel points in the filtered image matrix to obtain a segmented image matrix;

[0034] Traversing the target pixel points that meet the grayscale mutation condition in the segmented image matrix;

[0035] Generate a sudden pixel line segment based on the target pixel point, and determine the weld position according to the sudden pixel line segment;

[0036] The target weld area image is segmented based on the weld position to obtain a target weld image.

[0037] Optionally, the method of performing defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using the trained weld defect detection model to obtain a weld detection result includes:

[0038] Performing pixel clustering analysis on the target weld image by using a mixed Gaussian model, and extracting global features and local features of the target weld image according to the pixel clustering analysis results;

[0039] Inputting the global features and the local features into a feature pyramid network model respectively to obtain a multi-scale defect prediction result, wherein the defect prediction result includes at least candidate defect bounding boxes, confidence scores corresponding to each candidate defect bounding box, defect categories, and probability values ​​corresponding to the defect categories;

[0040] Perform non-maximum suppression on each candidate defect bounding box to filter out duplicate detection boxes and obtain the target defect detection box;

[0041] Filtering the defect detection results based on a preset confidence threshold and a preset probability threshold to obtain a target defect category;

[0042] A weld inspection result is generated according to the target defect detection frame and the target defect category.

[0043] Optionally, before performing defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using the trained weld defect detection model and obtaining the weld detection result, the method further includes:

[0044] Acquire a weld defect image sample, wherein the weld defect image sample is marked with an actual defect frame and a defect category identifier;

[0045] Inputting the weld defect image sample into an initial weld defect detection model for model training to obtain a defect prediction frame and a defect prediction category;

[0046] Calculating a bounding box loss value, a classification loss value, and a confidence loss value according to the defect prediction box, the defect prediction category, the actual defect box, and the defect category identifier;

[0047] generating an overall loss function based on the bounding box loss value, the classification loss value, and the confidence loss value;

[0048] The model parameters of the initial weld defect detection model are adjusted based on the overall loss function until the overall loss function is less than a preset loss threshold, and the trained weld defect detection model is output.

[0049] In addition, to achieve the above-mentioned purpose, the present invention also provides a nuclear power plant pipeline weld quality detection device, the nuclear power plant pipeline weld quality detection device comprising:

[0050] An acquisition module, configured to perform film scanning and / or digital ray acquisition on the nuclear power plant pipeline welds to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant;

[0051] A preprocessing module, configured to preprocess the weld area image to obtain a target weld area image;

[0052] An extraction module, configured to extract a target weld image from the target weld area image;

[0053] The detection module is used to perform defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using a trained weld defect detection model to obtain a weld detection result.

[0054] In addition, to achieve the above-mentioned objectives, the present invention also proposes a nuclear power plant pipeline weld quality detection device, which includes: a memory, a processor, and a nuclear power plant pipeline weld quality detection program stored in the memory and executable on the processor, wherein the nuclear power plant pipeline weld quality detection program is configured to implement the steps of the nuclear power plant pipeline weld quality detection method described above.

[0055] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which a nuclear power plant pipeline weld quality detection program is stored. When the nuclear power plant pipeline weld quality detection program is executed by a processor, the steps of the nuclear power plant pipeline weld quality detection method as described above are implemented.

[0056] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the nuclear power plant pipeline weld quality detection method as described above.

[0057] The present invention performs film scanning and / or digital ray acquisition on nuclear power plant pipeline welds to obtain weld area images of the pipeline weld areas of the nuclear power plant; preprocesses the weld area images to obtain target weld area images; extracts target weld images from the target weld area images; and uses a trained weld defect detection model to perform defect detection on the nuclear power plant pipeline weld areas according to the target weld images to obtain weld detection results. The weld area images of the nuclear power plant pipeline weld areas are acquired through a multi-dimensional acquisition method, adapting to different environments or pipeline parameters, thereby improving the robustness of weld detection in different environments without physically affecting the pipelines. The acquired weld area images are preprocessed to enhance the image clarity and detail, thereby improving the accuracy of subsequent defect detection. Finally, the trained weld defect detection model is used to automatically identify defect categories in the weld images, thereby improving the efficiency of weld quality inspection of nuclear power plant pipelines and avoiding the technical problems in the prior art of low efficiency and occasional missed or erroneous evaluation of weld films of nuclear power plant pipelines, which pose hidden dangers to the safe operation of nuclear power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0060] Figure 1 This is a flow chart of a first embodiment of a method for detecting weld quality in nuclear power plant pipelines according to the present invention;

[0061] Figure 2 This is a system functional architecture diagram of a weld quality detection solution for a nuclear power plant pipeline weld quality detection method according to an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of a process for obtaining a weld area image by using a film scanning technology in one embodiment of a method for inspecting weld quality of pipelines in nuclear power plants according to the present invention;

[0063] Figure 4 This is a flow chart of a second embodiment of a method for detecting weld quality in nuclear power plant pipelines according to the present invention;

[0064] Figure 5 A schematic diagram of the data management architecture of an embodiment of a nuclear power plant pipeline weld quality inspection method according to the present invention;

[0065] Figure 6 This is a structural block diagram of a first embodiment of a nuclear power plant pipeline weld quality inspection device according to the present invention;

[0066] Figure 7 It is a structural schematic diagram of a nuclear power plant pipeline weld quality detection device in the hardware operating environment involved in an embodiment of the present invention.

[0067] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0068] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0069] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0070] Based on this, the embodiment of the present invention provides a nuclear power plant pipeline weld quality detection method, referring to Figure 1 , Figure 1 The figure is a flow chart of a first embodiment of a method for detecting weld quality of pipelines in nuclear power plants according to the present invention.

[0071] In this embodiment, the nuclear power plant pipeline weld quality detection method includes:

[0072] Step S10: Film scanning and / or digital ray collection to obtain a digital image of the weld area of ​​the pipeline weld area of ​​the nuclear power plant.

[0073] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device capable of implementing the above functions, a control computer, etc. The following describes this embodiment and the following embodiments using a control computer as an example.

[0074] It should be understood that the weld quality inspection of nuclear power plant pipelines is different from that of traditional equipment. The weld quality inspection of nuclear power plant pipelines has strict requirements and a large amount of data. Traditional technologies that use manual inspection or sample inspection to detect welding quality may have a physical impact on the operation of equipment in nuclear power plants, resulting in serious consequences.

[0075] Based on this, in order to obtain the weld area image of the pipeline weld area of ​​the nuclear power plant without affecting the safe operation of the nuclear power plant equipment, this embodiment proposes a weld quality detection solution, referring to Figure 2 , Figure 2 This is a functional architecture diagram of the weld quality inspection solution of this embodiment. This embodiment also provides two image acquisition methods, namely film scanning and digital ray acquisition, to adapt to different environments or pipeline parameters and improve the robustness of weld inspection in different environments.

[0076] Furthermore, the performing of film scanning and / or digital radiography on the nuclear power plant pipeline welds to obtain weld area images of the pipeline weld areas of the nuclear power plant includes:

[0077] In response to an image scanning request, determining a scanning mode, scanning parameters, and image identification information according to the image scanning request;

[0078] Scanning the weld X-ray film according to preset scanning parameters and the scanning mode to obtain an initial weld area image;

[0079] Performing image quality detection on the initial weld area image;

[0080] After the quality inspection is passed, the initial weld area image is named according to the image identification information to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant;

[0081] and / or

[0082] Obtaining weld parameters of a pipeline weld area of ​​the nuclear power plant;

[0083] In response to an image acquisition request, determining an acquisition time and a number of superimposed frames of an image acquisition device based on the image acquisition request, and generating image identification information;

[0084] Adjusting the voltage, current and exposure time of the image acquisition device based on the weld parameters;

[0085] The image acquisition device is driven based on the voltage, current, exposure time, acquisition time and number of superimposed frames to acquire a weld area image of a pipeline weld area of ​​the nuclear power plant.

[0086] In the specific implementation, refer to Figure 3 , Figure 3 The figure is a schematic diagram of the process of obtaining weld area images through film scanning technology in this embodiment. In this embodiment, the pipelines in nuclear power plants are larger in area than those in traditional equipment, and the welds are distributed over a wider area. To improve the efficiency of weld quality inspection, this embodiment can uniformly collect weld X-ray films for the wide pipeline area and then perform subsequent quality inspections. In this process, since weld X-ray films are generally physical films, in order to obtain high-quality digital images, this embodiment uses film scanning technology to convert the physical weld X-ray films into high-quality digital weld area images.

[0087] Film scanning refers to placing the stored X-ray films of the welds of nuclear power plant pipelines on the scanning platform of the scanning equipment after receiving an image scanning request. During this process, it is necessary to ensure that the film surface is free of stains, wrinkles or impurities to avoid affecting the image quality during the scanning process. The scanning mode, scanning parameters and image identification information are determined according to the image scanning request. The scanning modes include: single-sheet scanning or multiple-sheet continuous scanning mode. The single-sheet mode is suitable for detailed inspection, and the continuous scanning is suitable for batch processing of large-scale films.

[0088] Scanning parameters include: film size (length, width), scanning sampling interval (such as 25, 50, 75) and other parameters. Image identification information includes overhaul number, weld number and other identification information, providing a basis for subsequent image classification and management.

[0089] After the scan is completed, the image is quality checked. The quality check includes checking whether the edges of the film are complete, whether the image is consistent with the original film, and ensuring that the image is free of blur, distortion, or unclear edges. If there are any problems, an automatic prompt will be given and a rescan will be required. Naming the initial weld area image according to the image identification information means using OCR technology combined with the overhaul number and weld number to recognize the identification characters in the image and automatically generate an image file name. According to preset rules, the scan results are saved to a specified storage path for subsequent image processing and weld quality inspection. The default name of the image file can be "overhaul number / weld number / ".

[0090] It is understandable that due to the narrow space in the area where some pipelines are located, it is difficult to directly collect the sealing image. Therefore, this embodiment uses digital ray acquisition technology to collect high-quality digital ray images of the weld area in this area to improve the operating safety of nuclear power plant equipment.

[0091] Weld parameters include the wall thickness and material of the weld. Adjust the voltage, current, and exposure time of the X-ray machine according to factors such as the wall thickness and material of the weld to ensure uniform and clear image exposure and avoid image noise or blur due to underexposure or overexposure. Set the acquisition time and number of superimposed frames of the image acquisition device to ensure high image quality without obvious noise information.

[0092] In addition, before image acquisition, image identification information such as overhaul number and weld number can be entered according to the image acquisition request. If this information is not entered, the image cannot be saved. After image acquisition is completed, the radiographic image of the weld area is displayed in real time, allowing fine-tuning of the image window width and window position to optimize contrast and brightness. If the image quality is not ideal (such as too dark, blurred or insufficient contrast), it can be immediately re-shot by adjusting the X-ray machine parameters, current, voltage, etc., or adjusting the acquisition time and other parameters of the flat-panel detector. Image naming, based on the overhaul number and weld number as the main naming information, automatically generates a serial number according to the shooting order, and combines the image names to ensure unified naming standards and facilitate subsequent data storage, query and management. After the image is named, the image is saved to the specified storage path according to the preset rules. If the user does not explicitly specify the path, the system will construct a default storage path based on the overhaul number and weld number. In addition to traditional X-ray machines, other types of rays (such as gamma rays) are also supported to expand the system's detection range for different materials and different wall thicknesses.

[0093] Step S20: pre-processing the weld area image to obtain a target weld area image.

[0094] In this embodiment, the weld area image is preprocessed to remove noise in the image, adjust the image contrast, ensure that the weld image details are clear, and facilitate subsequent model to identify defects in the image.

[0095] Furthermore, the preprocessing of the weld area image to obtain a target weld area image includes:

[0096] Denoising the weld area image using a preset filtering model;

[0097] Segmenting the weld area image to obtain a plurality of pixel blocks of the same size;

[0098] Count the grayscale histogram of each pixel block separately;

[0099] Performing peak clipping on the grayscale histogram of each pixel block based on a preset grayscale threshold to obtain a target grayscale histogram;

[0100] Calculating the cumulative distribution function of each pixel block based on the target grayscale histogram;

[0101] Performing grayscale mapping on each pixel block according to the cumulative distribution function to obtain a balanced local area image;

[0102] The balanced local area images corresponding to each pixel block are fused through a bilinear interpolation model to obtain an overall balanced image, and the overall balanced image is output as the target weld area image.

[0103] In a specific implementation, in order to denoise the weld area graphics, this embodiment can adopt filtering algorithms such as median filtering, mean filtering or bilateral filtering to effectively eliminate random noise introduced by the X-ray detector and environmental noise.

[0104] In addition, in order to improve the contrast between different areas of the image and avoid the over-enhancement phenomenon that may be introduced by global equalization, so that the details of the weld area are more prominent, this embodiment adopts an adaptive histogram equalization method to adjust the image quality of the weld area image, which is specifically manifested as follows:

[0105] The weld area image is divided into several non-overlapping sub-areas, such as the commonly used 8×8 or 16×16 pixel blocks; the grayscale histogram is calculated for each sub-area, and the frequency of occurrence of each grayscale level pixel in the area is counted, providing basic statistical data for subsequent local equalization; a preset grayscale threshold is set, and the grayscale histogram is subjected to peak clipping processing, which means that the part of the grayscale histogram where the number of grayscale pixels exceeds the threshold will be truncated, and the truncated pixel values ​​will be evenly distributed to all grayscale levels of the histogram.

[0106] The cumulative distribution function of each sub-region is calculated according to the contrast-limited histogram. The original grayscale value is mapped using the cumulative distribution function in each sub-region to obtain the equalized local area image.

[0107] The cumulative distribution function is calculated as:

[0108]

[0109] Where h(i) is the grayscale histogram and N is the total number of pixels in the image.

[0110] The mapping formula for mapping the original grayscale value is:

[0111] s = round((L-1)×CDF(r))

[0112] Wherein, L is the total number of gray levels (e.g., 256), r is the original pixel gray value, CDF(r) is the corresponding cumulative distribution function value, and s is the mapped pixel gray value.

[0113] Finally, the bilinear interpolation method is used to fuse the processing results of each sub-region. For each pixel, interpolation calculation is performed according to the mapping function of the sub-region where it is located and the adjacent sub-region to obtain a smooth and natural overall equalized image, and the overall equalized image is used as the target weld area image.

[0114] Step S30: extracting a target weld image from the target weld region image.

[0115] In a specific implementation, since there may be large-area environmental factors in the captured image, such as equipment or pipelines, when performing weld quality inspection, this environmental factor will bring certain interference to the model detection process. In order to improve the efficiency of the subsequent weld defect detection model in detecting weld quality, this embodiment can extract the weld part in the target weld area image, remove the background part as much as possible, and improve the efficiency of weld quality inspection.

[0116] Furthermore, extracting the target weld image from the target weld area image includes:

[0117] Extracting edge information from the target weld area image using an edge detection model to obtain a target weld area image;

[0118] Constructing a corresponding weld image matrix according to the target weld area image;

[0119] Performing grayscale conversion on the weld image matrix to obtain a grayscale conversion image matrix;

[0120] Performing Gaussian filtering on the grayscale conversion image matrix to obtain a filtered image matrix;

[0121] Performing threshold segmentation on pixel points in the filtered image matrix to obtain a segmented image matrix;

[0122] Traversing the target pixel points that meet the grayscale mutation condition in the segmented image matrix;

[0123] Generate a sudden pixel line segment based on the target pixel point, and determine the weld position according to the sudden pixel line segment;

[0124] The target weld area image is segmented based on the weld position to obtain a target weld image.

[0125] In the specific implementation, by adopting Sobel, Canny and other gradient-based edge detection algorithms, and combining preprocessing, parameter adaptive adjustment and post-processing fusion strategies, the edge information of the weld area can be accurately extracted. Clear and accurate weld contour and feature information is the key prerequisite for subsequent defect location, type judgment and quality assessment.

[0126] Specifically, the Sobel operator can quickly extract edge information within the weld area, and has a good detection effect on continuous and relatively smooth weld boundaries, which is suitable for preliminary positioning of the weld area; or the Canny algorithm can detect subtle changes in the weld edge, which is particularly suitable for detecting local weak edges caused by welding process or material unevenness. Its dual-threshold detection mechanism can retain the true weld edge while preventing noise interference.

[0127] In the specific implementation, the weld image matrix constructed based on the target weld area image is:

[0128]

[0129] Among them, {r,g,b} (x,y) is the RGB pixel value corresponding to the (x, y) coordinate position in the image matrix.

[0130] Threshold segmentation of pixels in the filtered image matrix means correcting pixel values ​​greater than a certain threshold to a first threshold, and correcting pixel values ​​less than or equal to a certain threshold to a second threshold. Generally, the first threshold is 255 and the second threshold is 0.

[0131] In this embodiment, the grayscale mutation condition refers to the pixel points that meet the grayscale mutation formula. The grayscale mutation formula is:

[0132]

[0133] Among them, d (x,y) is the pixel at the (x, y) coordinate position in the grayscale value mutation point image matrix Pd; a is a set of natural numbers satisfying (a·Δk); y=a·Δk means y≤n and is a positive integer multiple of Δk; Δk is the offset row number for interval search, which is generally a constant of 20.

[0134] This embodiment obtains a weld area image of the pipeline weld area of ​​the nuclear power plant by performing film scanning and / or digital ray acquisition on the welds of the nuclear power plant pipeline; preprocesses the weld area image to obtain a target weld area image; extracts a target weld image from the target weld area image; uses a trained weld defect detection model to perform defect detection on the nuclear power plant pipeline weld area according to the target weld image to obtain a weld detection result, obtains a weld area digital image of the pipeline weld area of ​​the nuclear power plant through a multi-dimensional acquisition method, and preprocesses the acquired weld area image to improve the clarity and detail of the image, thereby improving the accuracy of subsequent defect detection, and finally automatically identifies the defect category of the weld image through the trained weld defect detection model, thereby improving the detection efficiency of the weld quality film of the nuclear power plant pipeline.

[0135] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 4 Step S40 includes:

[0136] Step S401: performing pixel clustering analysis on the target weld image using a mixture Gaussian model, and extracting global features and local features of the target weld image based on the pixel clustering analysis results.

[0137] Step S402: inputting the global features and the local features into a feature pyramid network model respectively to obtain multi-scale defect prediction results.

[0138] Step S403: performing non-maximum suppression processing on each candidate defect bounding box to filter out duplicate detection boxes and obtain a target defect detection box.

[0139] Step S404: screening the defect detection results based on a preset confidence threshold and a preset probability threshold to obtain a target defect category.

[0140] Step S405: generating a weld inspection result according to the target defect inspection frame and the target defect category.

[0141] It should be noted that global features are used to characterize the overall structural information of the weld area, and local features characterize the specific details and defect characteristics in the weld. Global features are generally used for the detection of large-scale defects, and local features are mainly used for weld quality detection of small and medium-scale defects.

[0142] Pixel cluster analysis refers to analyzing the probability of each pixel in the target weld image belonging to a certain class. For example, the detailed judgment process of whether a certain pixel belongs to a defective pixel or a normal pixel is as follows (taking defective pixels as an example): if the defective pixels contain a total of K classes, that is, K Gaussian subclasses, then the probability of the pixel color value belonging to each subclass is analyzed at the same time, and then the sum of these K probabilities is calculated. The pixel classification formula in the pixel cluster analysis process is:

[0143]

[0144] Where K is the number of subclasses, x is the RGB color value of the pixel in the target weld image, μ is the data mean, and π k is the prior probability of the k-th Gaussian distribution.

[0145] In its specific implementation, the weld recognition process refers to normalizing the target weld image processed by the image enhancement and analysis module and adjusting it to the fixed input size (608×608) required by the model for subsequent network processing. Based on the edge detection and weld area segmentation results in the preprocessing stage, the weld area in the image is marked or cropped to reduce background interference and improve detection efficiency. Through the Feature Pyramid Networks (FPN), feature information of different scales is integrated to ensure that large, medium and small-scale defects can be effectively detected.

[0146] In the detection head, a convolutional layer is used to output prediction results at multiple detection scales. Each prediction contains a bounding box, confidence, and category probability information. Non-maximum suppression is performed on the candidate bounding boxes output by the detection head, and duplicate detection boxes with large overlaps are filtered out to retain the optimal detection results. The detection results are then filtered according to the preset confidence threshold and category probability threshold to ensure that only high-confidence defect detection results are output.

[0147] Optionally, before performing defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using the trained weld defect detection model and obtaining the weld detection result, the method further includes:

[0148] Acquire a weld defect image sample, wherein the weld defect image sample is marked with an actual defect frame and a defect category identifier;

[0149] Inputting the weld defect image sample into an initial weld defect detection model for model training to obtain a defect prediction frame and a defect prediction category;

[0150] Calculating a bounding box loss value, a classification loss value, and a confidence loss value according to the defect prediction box, the defect prediction category, the actual defect box, and the defect category identifier;

[0151] generating an overall loss function based on the bounding box loss value, the classification loss value, and the confidence loss value;

[0152] The model parameters of the initial weld defect detection model are adjusted based on the overall loss function until the overall loss function is less than a preset loss threshold, and the trained weld defect detection model is output.

[0153] In a specific implementation, during the model training process, this embodiment uses loss functions in multiple dimensions to evaluate the effect of model training and iteratively optimize it.

[0154] Bounding box loss function (LCIoU): CIoU loss is used to measure the deviation between the predicted box and the real box.

[0155]

[0156] L CIoU =1-CIOU

[0157] Among them, d is the distance between the center point of the predicted box and the real box, c is the diagonal distance of the minimum enclosing rectangle, and is the similarity factor of the aspect ratio.

[0158] Classification loss function (Lcls): uses cross entropy to calculate the difference between the predicted category and the true category;

[0159]

[0160] Among them, yi is the true category label, pi is the probability of the predicted category, and C is the number of categories.

[0161] Confidence loss function (Lconf): measures the confidence prediction error of whether the candidate box contains the target.

[0162]

[0163] Among them, i refers to the prior frame number, j refers to the real frame number, p refers to the category number, p = 0 means background, Taking 1 means that the i-th prior bbox matches the j-th GT box, and the category of this GT box is p. represents the predicted probability of the i-th search box corresponding to category p. It is important to note that the first half of the formula is the loss of positive samples (Pos), that is, the loss of being classified as a certain category (excluding background), and the second half is the loss of negative samples (Neg), that is, the loss of being classified as background.

[0164] The formula for generating the overall loss function based on the bounding box loss value, classification loss value, and confidence loss value is:

[0165] L=L CIoU +λ cls L cls +λ conf L conf

[0166] Among them, λ cls and λ conf is the weight coefficient corresponding to each loss function, which is determined by adjusting the parameters of the validation set.

[0167] In this embodiment, in order to facilitate the management of the quality inspection data of each pipeline weld, in this embodiment, each pipeline weld can also be archived and managed according to the image identification information mentioned above.

[0168] Specifically, refer to Figure 5 , Figure 5 This is the architecture diagram of data management in this embodiment. Data archiving is carried out in accordance with the hierarchical management method of "electric field-inspection company-unit-inspection number-file number-weld number". Each level serves as a key dimension of data index to facilitate data classification, statistics and subsequent retrieval. Among them, the electric field corresponds to each nuclear power plant or power plant area; the inspection company is used to distinguish different inspection units or inspection service providers; the unit corresponds to a specific unit to further refine the data archiving; the inspection number is the unique identifier of each inspection task to ensure the timeliness and independence of the inspection data; the file number is the file identifier associated with the inspection report and historical records; the weld number is used to finally refine to each weld and store specific inspection images and assessment information.

[0169] A hierarchical index structure is constructed using the database to ensure that data is automatically stored according to the above-mentioned hierarchical relationship during archiving. A distributed file system is used to store large-size images and original inspection data, while metadata and index information are saved in the database. Permissions are configured according to different levels and user roles (such as inspection companies, crew managers, auditors, etc.) to ensure that sensitive data can be securely accessed and operated under hierarchical management, and data encryption, backup, and logging are implemented to meet security and traceability requirements.

[0170] This embodiment performs pixel clustering analysis on the target weld image through a mixed Gaussian model, and extracts global features and local features of the target weld image based on the pixel clustering analysis results; the global features and the local features are respectively input into a feature pyramid network model to obtain multi-scale defect prediction results, which include at least candidate defect bounding boxes, confidence levels corresponding to each candidate defect bounding box, defect categories, and probability values ​​corresponding to the defect categories; non-maximum suppression processing is performed on each candidate defect bounding box to filter out duplicate detection frames to obtain a target defect detection frame; the defect detection results are screened based on a preset confidence threshold and a preset probability threshold to obtain a target defect category; a weld detection result is generated based on the target defect detection frame and the target defect category, and weld quality detection at multiple scales is performed on the weld image through global features and local features to improve the accuracy of weld monitoring.

[0171] This application also provides a nuclear power plant pipeline weld quality detection device, please refer to Figure 6 , the nuclear power plant pipeline weld quality detection device includes:

[0172] The acquisition module 10 is used to perform film scanning and / or digital ray acquisition on the welds of the pipelines of the nuclear power plant to obtain weld area images of the weld areas of the pipelines of the nuclear power plant.

[0173] The preprocessing module 20 is used to preprocess the weld area image to obtain a target weld area image.

[0174] The extraction module 30 is used to extract the target weld image from the target weld area image.

[0175] The detection module 40 is used to perform defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using a trained weld defect detection model to obtain a weld detection result.

[0176] This embodiment obtains a weld area image of the pipeline weld area of ​​the nuclear power plant by performing film scanning and / or digital ray acquisition on the nuclear power plant pipeline weld; preprocesses the weld area image to obtain a target weld area image; extracts a target weld image from the target weld area image; uses a trained weld defect detection model to perform defect detection on the nuclear power plant pipeline weld area according to the target weld image to obtain a weld detection result, obtains a weld area image of the pipeline weld area of ​​the nuclear power plant through a multi-dimensional acquisition method, and preprocesses the acquired weld area image to improve the clarity and detail of the image, thereby improving the accuracy of subsequent defect detection, and finally automatically identifies the defect category of the weld image through the trained weld defect detection model, thereby improving the detection efficiency of the weld quality film of the nuclear power plant pipeline.

[0177] In one embodiment, the acquisition module 10 is further used to respond to an image scanning request, determine a scanning mode, scanning parameters and image identification information according to the image scanning request; perform an image scan on the weld X-ray film according to preset scanning parameters and the scanning mode to obtain an initial weld area image; perform image quality detection on the initial weld area image; after the quality detection passes, perform image naming on the initial weld area image according to the image identification information to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant; and / or obtain weld parameters of the pipeline weld area of ​​the nuclear power plant; in response to the image acquisition request, determine the acquisition time and the number of superimposed frames of the image acquisition device based on the image acquisition request, and generate image identification information; adjust the voltage, current and exposure time of the image acquisition device based on the weld parameters; drive the image acquisition device based on the voltage, current, exposure time, acquisition time and number of superimposed frames to obtain the weld area image of the pipeline weld area of ​​the nuclear power plant.

[0178] In one embodiment, the preprocessing module 20 is further used to denoise the weld area image through a preset filtering model; perform image segmentation on the weld area image to obtain multiple pixel blocks of the same size; respectively calculate the grayscale histogram of each pixel block; perform peak clipping on the grayscale histogram of each pixel block based on a preset grayscale threshold to obtain a target grayscale histogram; calculate the cumulative distribution function of each pixel block based on the target grayscale histogram; perform grayscale mapping on each pixel block according to the cumulative distribution function to obtain a balanced local area image; fuse the balanced local area images corresponding to each pixel block through a bilinear interpolation model to obtain an overall balanced image, and output the overall balanced image as the target weld area image.

[0179] In one embodiment, the extraction module 30 is further used to extract edge information in the target weld area image through an edge detection model to obtain a target weld area image; construct a corresponding weld image matrix based on the target weld area image; perform grayscale conversion on the weld image matrix to obtain a grayscale conversion image matrix; perform Gaussian filtering on the grayscale conversion image matrix to obtain a filtered image matrix; perform threshold segmentation on the pixel points in the filtered image matrix to obtain a segmented image matrix; traverse the target pixel points in the segmented image matrix that meet the grayscale mutation condition; generate a mutation pixel line segment based on the target pixel point, and determine the weld position based on the mutation pixel line segment; segment the target weld area image based on the weld position to obtain a target weld image.

[0180] In one embodiment, the detection module 40 is further used to perform pixel clustering analysis on the target weld image through a mixed Gaussian model, and extract global features and local features of the target weld image based on the pixel clustering analysis results; input the global features and the local features into a feature pyramid network model respectively to obtain multi-scale defect prediction results, wherein the defect prediction results include at least candidate defect bounding boxes, confidence levels corresponding to each candidate defect bounding box, defect categories, and probability values ​​corresponding to the defect categories; perform non-maximum suppression processing on each candidate defect bounding box to filter out duplicate detection boxes to obtain a target defect detection box; screen the defect detection results based on a preset confidence threshold and a preset probability threshold to obtain a target defect category; and generate a weld detection result based on the target defect detection box and the target defect category.

[0181] In one embodiment, the detection module 40 is also used to obtain weld defect image samples, in which the weld defect image samples are marked with actual defect boxes and defect category identifiers; the weld defect image samples are input into the initial weld defect detection model for model training to obtain defect prediction boxes and defect prediction categories; the bounding box loss value, classification loss value and confidence loss value are calculated based on the defect prediction box, defect prediction category, actual defect box and defect category identifier; an overall loss function is generated based on the bounding box loss value, classification loss value and confidence loss value; the model parameters of the initial weld defect detection model are adjusted based on the overall loss function until the overall loss function is less than a preset loss threshold, and the trained weld defect detection model is output.

[0182] The present application provides a nuclear power plant pipeline weld quality detection device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the nuclear power plant pipeline weld quality detection method of the above-mentioned embodiment 1.

[0183] Reference below Figure 7, which shows a schematic structural diagram of a nuclear power plant pipeline weld quality inspection device suitable for implementing embodiments of the present application. The nuclear power plant pipeline weld quality inspection device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7 The nuclear power plant pipeline weld quality inspection equipment shown is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

[0184] like Figure 7 As shown, the nuclear power plant pipeline weld quality inspection equipment may include a processing device 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the nuclear power plant pipeline weld quality inspection equipment. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input device 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage device 1003 including, for example, a magnetic tape, hard disk, etc.; and communication device 1009. Communication device 1009 can allow the nuclear power plant pipeline weld quality inspection equipment to communicate wirelessly or wired with other equipment to exchange data. Although the figure shows a nuclear power plant pipeline weld quality inspection equipment with various systems, it should be understood that implementation or presence of all the illustrated systems is not required. More or fewer systems may alternatively be implemented or present.

[0185] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0186] The nuclear power plant pipeline weld quality inspection device provided in this application utilizes the nuclear power plant pipeline weld quality inspection method described in the aforementioned embodiment, thereby resolving the technical issues surrounding nuclear power plant pipeline weld quality inspection. Compared to the prior art, the nuclear power plant pipeline weld quality inspection device provided in this application achieves the same beneficial effects as the nuclear power plant pipeline weld quality inspection method described in the aforementioned embodiment. The other technical features of this nuclear power plant pipeline weld quality inspection device are the same as those disclosed in the aforementioned embodiment and are not further elaborated upon here.

[0187] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0188] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0189] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, a computer program) stored thereon, the computer-readable program instructions being used to execute the nuclear power plant pipeline weld quality inspection method in the above-mentioned embodiment.

[0190] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0191] The computer-readable storage medium may be included in the nuclear power plant pipeline weld quality detection equipment; or it may exist independently without being assembled into the nuclear power plant pipeline weld quality detection equipment.

[0192] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the nuclear power plant pipeline weld quality detection equipment, the nuclear power plant pipeline weld quality detection equipment is enabled to perform: nuclear power plant pipeline weld quality detection.

[0193] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Python, Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0194] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0195] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0196] The computer-readable storage medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned method for inspecting the quality of nuclear power plant pipeline welds. This computer-readable storage medium can address the technical challenges of inspecting the quality of nuclear power plant pipeline welds. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are similar to those of the method for inspecting the quality of nuclear power plant pipeline welds provided in the aforementioned embodiments, and are not further elaborated here.

[0197] The present application also provides a computer program product, comprising a computer program, which implements the steps of the above-mentioned nuclear power plant pipeline weld quality detection method when executed by a processor.

[0198] The computer program product provided in this application can solve the technical problem of nuclear power plant pipeline weld quality inspection. Compared with the existing technology, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the nuclear power plant pipeline weld quality inspection method provided in the above embodiment, and will not be repeated here.

[0199] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for detecting the quality of pipeline welds in a nuclear power plant, characterized in that: The nuclear power plant pipeline weld quality detection method comprises: Performing film scanning and / or digital radiography on the nuclear power plant pipeline welds to obtain weld area images of the pipeline weld areas of the nuclear power plant; Preprocessing the weld area image to obtain a target weld area image; Extracting a target weld image from the target weld area image; By using the trained weld defect detection model, defect detection is performed on the weld area of ​​the nuclear power plant pipeline according to the target weld image to obtain a weld detection result.

2. The nuclear power plant pipeline weld quality inspection method according to claim 1, characterized in that: The method of performing film scanning and / or digital radiography on the nuclear power plant pipeline welds to obtain a weld area image of the nuclear power plant pipeline weld area includes: In response to an image scanning request, determining a scanning mode, scanning parameters, and image identification information according to the image scanning request; Scanning the weld X-ray film according to preset scanning parameters and the scanning mode to obtain an initial weld area image; Performing image quality detection on the initial weld area image; After the quality inspection is passed, the initial weld area image is named according to the image identification information to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant; and / or Obtaining weld parameters of a pipeline weld area of ​​the nuclear power plant; In response to an image acquisition request, determining an acquisition time and a number of superimposed frames of an image acquisition device based on the image acquisition request, and generating image identification information; Adjusting the voltage, current and exposure time of the image acquisition device based on the weld parameters; The image acquisition device is driven based on the voltage, current, exposure time, acquisition time and number of superimposed frames to acquire a weld area image of a pipeline weld area of ​​the nuclear power plant.

3. The nuclear power plant pipeline weld quality inspection method according to claim 1, characterized in that: The preprocessing of the weld area image to obtain a target weld area image includes: Denoising the weld area image using a preset filtering model; Segmenting the weld area image to obtain a plurality of pixel blocks of the same size; Count the grayscale histogram of each pixel block separately; Performing peak clipping on the grayscale histogram of each pixel block based on a preset grayscale threshold to obtain a target grayscale histogram; Calculating the cumulative distribution function of each pixel block based on the target grayscale histogram; Performing grayscale mapping on each pixel block according to the cumulative distribution function to obtain a balanced local area image; The balanced local area images corresponding to each pixel block are fused through a bilinear interpolation model to obtain an overall balanced image, and the overall balanced image is output as the target weld area image.

4. The nuclear power plant pipeline weld quality inspection method according to claim 1, characterized in that: The extracting of the target weld image from the target weld area image comprises: Extracting edge information from the target weld area image using an edge detection model to obtain a target weld area image; Constructing a corresponding weld image matrix according to the target weld area image; Performing grayscale conversion on the weld image matrix to obtain a grayscale conversion image matrix; Performing Gaussian filtering on the grayscale conversion image matrix to obtain a filtered image matrix; Performing threshold segmentation on pixel points in the filtered image matrix to obtain a segmented image matrix; Traversing the target pixel points that meet the grayscale mutation condition in the segmented image matrix; Generate a sudden pixel line segment based on the target pixel point, and determine the weld position according to the sudden pixel line segment; The target weld area image is segmented based on the weld position to obtain a target weld image.

5. The nuclear power plant pipeline weld quality inspection method according to claim 1, characterized in that: The trained weld defect detection model is used to perform defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image to obtain a weld detection result, including: Performing pixel clustering analysis on the target weld image by using a mixed Gaussian model, and extracting global features and local features of the target weld image according to the pixel clustering analysis results; Inputting the global features and the local features into a feature pyramid network model respectively to obtain a multi-scale defect prediction result, wherein the defect prediction result includes at least candidate defect bounding boxes, confidence scores corresponding to each candidate defect bounding box, defect categories, and probability values ​​corresponding to the defect categories; Perform non-maximum suppression on each candidate defect bounding box to filter out duplicate detection boxes and obtain the target defect detection box; Filtering the defect detection results based on a preset confidence threshold and a preset probability threshold to obtain a target defect category; A weld inspection result is generated according to the target defect detection frame and the target defect category.

6. The nuclear power plant pipeline weld quality inspection method according to claim 5, characterized in that: Before performing defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using the trained weld defect detection model and obtaining the weld detection result, the method further includes: Acquire a weld defect image sample, wherein the weld defect image sample is marked with an actual defect frame and a defect category identifier; Inputting the weld defect image sample into an initial weld defect detection model for model training to obtain a defect prediction frame and a defect prediction category; Calculating a bounding box loss value, a classification loss value, and a confidence loss value according to the defect prediction box, the defect prediction category, the actual defect box, and the defect category identifier; generating an overall loss function based on the bounding box loss value, the classification loss value, and the confidence loss value; The model parameters of the initial weld defect detection model are adjusted based on the overall loss function until the overall loss function is less than a preset loss threshold, and the trained weld defect detection model is output.

7. A nuclear power plant pipeline weld quality detection device, characterized in that: The nuclear power plant pipeline weld quality detection device comprises: An acquisition module, configured to perform film scanning and / or digital ray acquisition on the nuclear power plant pipeline welds to obtain a weld area image of the pipeline weld area of ​​the nuclear power plant; A preprocessing module, configured to preprocess the weld area image to obtain a target weld area image; An extraction module, configured to extract a target weld image from the target weld area image; The detection module is used to perform defect detection on the weld area of ​​the nuclear power plant pipeline according to the target weld image using a trained weld defect detection model to obtain a weld detection result.

8. A nuclear power plant pipeline weld quality inspection equipment, characterized in that: The nuclear power plant pipeline weld quality inspection device includes: a memory, a processor, and a nuclear power plant pipeline weld quality inspection program stored in the memory and executable on the processor, wherein the nuclear power plant pipeline weld quality inspection program is configured to implement the nuclear power plant pipeline weld quality inspection method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a nuclear power plant pipeline weld quality detection program, which, when executed by a processor, implements the nuclear power plant pipeline weld quality detection method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the nuclear power plant pipeline weld quality inspection method according to any one of claims 1 to 6 are implemented.

Images