Weld Defect Recognition Method and Device Based on X-ray Weld Images

Abstract

The embodiments of this specification provide a method and device for weld defect recognition based on X-ray weld images. When performing weld defect recognition, image region analysis is carried out on the X-ray weld image at the pipe weld to obtain the image region features of the X-ray weld image. The obtained image region features include region composition, region position, region brightness, region density, and region shape. Subsequently, the image region features of the X-ray weld image are provided to the weld defect recognition model for weld defect recognition. By using this weld defect recognition method, through the use of the weld defect recognition model to perform weld defect recognition based on image region features such as the region position, region brightness, region density, and region shape of the X-ray weld image, the efficiency and accuracy of weld defect recognition can be improved.

Description

Technical Field

[0001] Embodiments of this specification generally relate to the field of chemical production, and particularly to a method and device for identifying weld defects based on X-ray weld images at the welds of pipelines. Background Art

[0002] In an integrated chemical production system, a large number of metal equipment need to be deployed, and pipelines are laid between the metal equipment for connection. The laid pipelines can be as long as 1500 kilometers, for example. Therefore, multiple pipelines need to be welded together, and welding treatment is also required between the pipelines and the connected metal equipment. If there are defects in the welds at the joints, it will cause chemical production risks such as leakage of chemical raw materials. Summary of the Invention

[0003] In view of the above, embodiments of this specification provide a method and device for identifying weld defects based on X-ray weld images. By using this weld defect identification method, the efficiency and accuracy of weld defect identification can be improved by using a weld defect identification model to identify weld defects based on image region features such as the regional position, regional brightness, regional density, and regional shape of the X-ray weld image.

[0004] According to one aspect of the embodiments of this specification, a method for identifying weld defects based on X-ray weld images is provided, including: performing image region analysis on the X-ray weld image at the pipeline weld to obtain the image region features of the X-ray weld image, where the image region features include regional position, regional brightness, regional density, and regional shape; and providing the image region features of the X-ray weld image to a weld defect identification model to perform weld defect identification.

[0005] Optionally, in an example of the above aspect, providing the image region features of the X-ray weld image to a weld defect identification model to perform weld defect identification may include: providing the image region features of the X-ray weld image to a defect feature generation layer of the weld defect identification model to generate a defect feature representation of the X-ray weld image; and providing the defect feature representation of the X-ray weld image to a defect classification layer of the weld defect identification model to perform weld defect identification.

[0006] Optionally, in an example of the above aspect, the X-ray weld image includes a 360-degree weld chain image formed by stitching multiple scanned film images at different angles at the pipeline weld.

[0007] Optionally, in an example of the above aspect, the training dataset of the weld defect recognition model includes a training dataset constructed based on the actual X-ray weld image at the pipe weld and the enhanced X-ray weld image obtained by performing data augmentation on the actual X-ray weld image, and the data augmentation includes at least one of image rotation, image translation, image scaling, and filter-based image filtering.

[0008] Optionally, in an example of the above aspect, before performing image region analysis on the X-ray weld image at the pipe weld, the weld defect recognition method may further include: based on the weld shape parameters, determining whether the X-ray weld image at the pipe weld has been recognized according to the recognized X-ray weld image library, where the weld shape parameters include the build-up pitch of the welded part and the radian of the welded part. In response to the X-ray weld image not being recognized, performing image region analysis on the X-ray weld image at the pipe weld.

[0009] Optionally, in an example of the above aspect, determining whether the X-ray weld image at the pipe weld has been recognized according to the recognized X-ray weld image library based on the weld shape parameters may include: obtaining the weld shape parameters of the X-ray weld image; and based on the obtained weld shape parameters, determining whether the X-ray weld image has been recognized according to the recognized X-ray weld image library.

[0010] Optionally, in an example of the above aspect, determining whether the X-ray weld image at the pipe weld has been recognized according to the recognized X-ray weld image library based on the weld shape parameters may include: providing the X-ray weld image and the recognized X-ray weld images in the recognized X-ray weld image library to a weld image similarity determination model to determine whether the X-ray weld image has been recognized, and the weld image similarity determination model is pre-trained using the weld shape parameters as model features.

[0011] Optionally, in an example of the above aspect, the training dataset of the weld image similarity determination model includes a training dataset constructed based on the actual X-ray weld image at the pipe weld and the enhanced X-ray weld image obtained by performing data augmentation on the actual X-ray weld image, and the data augmentation includes at least one of image rotation, image translation, image scaling, and filter-based image filtering.

[0012] Optionally, in an example of the above aspect, the weld defect recognition method may further include: generating a weld defect recognition report based on the weld defect recognition result.

[0013] According to another aspect of the embodiments of the present specification, there is provided a weld defect recognition device based on X-ray weld images, including: an image region analysis unit configured to perform image region analysis on the X-ray weld image at the pipe weld to obtain the image region features of the X-ray weld image, where the image region features include region position, region brightness, region density, and region shape; and a weld defect recognition unit configured to provide the image region features of the X-ray weld image to a weld defect recognition model for weld defect recognition.

[0014] Optionally, in an example of the above aspect, the weld defect recognition device may further include: an image recognition determination unit configured to determine whether the X-ray weld image has been recognized based on the weld shape parameters according to the recognized X-ray weld image library before performing image region analysis on the X-ray weld image, where the weld shape parameters include the build-up pitch of the welded part and the arc of the welded part. Description of the Drawings

[0015] By referring to the following drawings, a further understanding of the essence and advantages of the content of the present specification can be achieved. In the drawings, similar components or features may have the same reference numerals.

[0016] Figure 1 Shows an example block diagram of a weld defect recognition system according to an embodiment of the present specification.

[0017] Figure 2 Shows an example schematic diagram of an X-ray weld image according to an embodiment of the present specification.

[0018] Figure 3 Shows an example schematic diagram of an X-ray weld image stitched together from multiple X-ray film images according to an embodiment of the present specification.

[0019] Figures 4A - 4O Shows an example schematic diagram of a defective X-ray weld image according to an embodiment of the present specification.

[0020] Figure 5 Shows an example schematic diagram of a weld defect recognition model according to an embodiment of the present specification.

[0021] Figure 6 Shows an example flowchart of a weld defect recognition method according to an embodiment of the present specification.

[0022] Figure 7 Shows an example flowchart of a process for determining whether a weld image has been recognized according to an embodiment of the present specification.

[0023] Figure 8Shows an example block diagram of a weld defect recognition device according to an embodiment of the present specification.

[0024] Figure 9 Shows an example schematic diagram of a computer system-implemented weld defect recognition device according to an embodiment of the present specification. Detailed implementation manners

[0025] Now, the subject matter described herein will be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein, and is not a limitation on the protection scope, applicability, or examples set forth in the claims. The functions and arrangements of the elements discussed can be changed without departing from the protection scope of the content of this specification. Each example can omit, substitute, or add various processes or components as needed. For example, the described method can be executed in an order different from the described order, and each step can be added, omitted, or combined. Additionally, the features described relative to some examples can also be combined in other examples.

[0026] As used herein, the term "comprising" and its variants represent open terms, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc. can refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. Unless clearly specified in the context, the definition of a term is consistent throughout the specification.

[0027] Figure 1 Shows an example block diagram of a weld defect recognition system 100 according to an embodiment of the present specification.

[0028] As Figure 1 shown, the weld defect recognition system 100 includes a weld image acquisition device 110, a recognized weld image library 120, a weld defect recognition model training device 130, a weld defect recognition model storage device 140, and a weld defect recognition device 150. The weld image acquisition device 110, the recognized weld image library 120, the weld defect recognition model training device 130, the weld defect recognition model storage device 140, and the weld defect recognition device 150 can be communicably connected via a network 160. In some embodiments, some or all of the components of the weld image acquisition device 110, the recognized weld image library 120, the weld defect recognition model training device 130, the weld defect recognition model storage device 140, and the weld defect recognition device 150 can communicate directly without communicating through the network 160.

[0029] In some embodiments, network 160 may be any one or more of a wired network or a wireless network. Examples of network 160 may include, but are not limited to, cable networks, fiber optic networks, telecommunications networks, enterprise internal networks, the Internet, local area networks (LANs), wide area networks (WANs), wireless local area networks (WLANs), metropolitan area networks (MANs), public switched telephone networks (PSTNs), Bluetooth networks, ZigBee networks, near field communication (NFC), in-device buses, in-device lines, etc., or any combination thereof.

[0030] The weld image acquisition device 110 is configured to acquire an X-ray weld image of a pipeline weld to be inspected for weld defects. A pipeline weld refers to the joint formed when two cast or forged metals are joined together to form a pipeline. The formed pipeline weld is typically a circular weld. Given the complexity of the pipeline welding site, welding must be performed manually by a welder. Due to the continuous change of the welder's posture during welding and the uniqueness of the geography, time, and space at which each weld is welded, each weld has a unique weld shape, for example, a unique weld wave, etc. The weld shape parameters of the pipeline weld may include, for example, weld edge shape parameters, such as the weldment overlaying gap and the weldment radian. The weldment overlaying gap is used to refer to the overlapping distance between two adjacent overlaying welds. The size of the weldment overlaying gap of the weld has a significant impact on the flatness of the overlaying layer surface, the chemical composition uniformity of the overlaying layer, and the dilution rate of the base metal. The smaller the weldment overlaying gap, the higher the pipeline welding quality. The weldment radian is used to refer to the radius of curvature formed on the surface of the weldment. Figure 2 An example schematic diagram of an X-ray weld image according to an embodiment of the present specification is shown. In addition, the weld shape parameters may further include weld height, weld width, and the degree of unevenness of the weld surface ripple.

[0031] In some embodiments, the weld seam image acquisition device 110 can be implemented as an X-ray scanner. When acquiring an X-ray weld seam image, the pipeline weld seam is scanned at a given angle (e.g., 0°, 120°, or 240°) to obtain an X-ray weld seam image (scanned film image) at the pipeline weld seam. Due to the limitations of the X-ray radiation source used for X-ray scanning, the X-ray weld seam image obtained by each scan is only a partial weld seam of the pipeline weld seam. In some embodiments, scanning can be performed at multiple different angles to obtain multiple scanned film images, and then the obtained multiple scanned film images are stitched together to form a 360-degree weld chain image, thereby obtaining a complete X-ray weld seam image. In some embodiments, multiple scanned film images at different angles at the pipeline weld seam can be stitched together and digitally twinned to form a 360-degree digital weld chain image as the X-ray weld seam image at the pipeline weld seam. Figure 3 The figure shows an exemplary schematic diagram of an X-ray weld seam image stitched together based on multiple X-ray film images according to an embodiment of the present specification.

[0032] The X-ray weld seam image acquired by the weld seam image acquisition device 110 is provided to the weld seam defect recognition device 150. After receiving the X-ray weld seam image, the weld seam defect recognition device 150 performs image region analysis on the X-ray weld seam image to obtain the image region features of the X-ray weld seam image, and then provides the image region features to the weld seam defect recognition model for weld seam defect recognition. The weld seam defect recognition process of the weld seam defect recognition device 150 will be described in detail later with reference to the accompanying drawings. In some embodiments, the weld seam image acquisition device 110 may also have an image preprocessing function, such as an image denoising function, etc. The weld seam image acquisition device 110 can perform image denoising processing on the acquired original X-ray weld seam image, thereby improving the image quality of the X-ray weld seam image, and providing the X-ray weld seam image after image denoising processing to the weld seam defect recognition device 150 for weld seam defect recognition, thereby improving the weld seam defect recognition accuracy.

[0033] The X-ray weld seam image for which weld seam defect recognition has been completed can be stored in the recognized weld seam image library 120 for use by the weld seam defect recognition device 150 to perform repeated recognition and identification of weld seam images. In some embodiments, the original X-ray weld seam image or the X-ray weld seam image after image preprocessing can be stored in the recognized weld seam image library 120. In some embodiments, the image feature representation (Embedding) obtained after image feature extraction of the original X-ray weld seam image or the X-ray weld seam image after image preprocessing can be stored in the recognized weld seam image library 120. The image feature representation can also be referred to as the image feature vector representation.

[0034] The weld defect recognition model used by the weld defect recognition device 150 may include, for example, a machine learning model and a deep learning model. The machine learning model may include a supervised learning model. Examples of the supervised learning model may include a linear regression model, a logistic regression model, a decision tree, a support vector machine, a BP neural network, a random forest, etc. Examples of the deep learning model may include a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), a Long Short-Term Memory (LSTM), etc. In some embodiments, the weld defect recognition model is implemented as a multi-classification model for realizing multi-defect category recognition.

[0035] Figures 4A - 4O An exemplary schematic diagram of a defective X-ray weld image according to an embodiment of the present specification is shown.

[0036] Figure 4A An exemplary schematic diagram of an X-ray weld image with a Cold Lap defect according to an embodiment of the present specification is shown. A Cold Lap defect is a weld defect that occurs when the flux metal fails to properly fuse with the base metal or previously completed welded material. In the case of Cold Lap, the arc of the weldment fails to sufficiently melt the base metal, causing a small amount of flux metal melt to flow into the base metal without achieving sufficient adhesion. If a Cold Lap defect occurs, the weld shape will be irregular and there will be areas containing inclusions in the weld in the X-ray weld image. Since the area density of the area containing inclusions will be different from the area density of the surrounding area, for example, the area density of the area containing inclusions is usually higher than the area density of the surrounding area, the Cold Lap defect can be judged based on the area density. As can be seen from the above, when there are areas with irregular edges, discontinuous lines, or significant gray-scale differences from the surrounding area in the weld image, a Cold Lap defect will occur, and thus image area features such as area shape and area density can be used to characterize the Cold Lap defect.

[0037] Figure 4BAn example schematic diagram of an X-ray weld image with a porosity defect according to an embodiment of the present specification is shown. Porosity defects are weld defects caused by gas entrapment in solid metals. On an X-ray weld image, pores can take on a variety of shapes, but typically appear as dark round or irregular spots or spots that are individual, clustered, or in rows. Sometimes, pores can be elongated and may appear to have tails. This is the result of gas trying to escape while the metal is still in a liquid state, called wormhole pores. All pores are voids in the material, and their area density (e.g., radiographic density) will be higher than the surrounding area. As can be seen from the above, image area features such as area shape, area brightness, and area density can be used to characterize porosity defects.

[0038] Figure 4C An example schematic diagram of an X-ray weld image with a Cluster Porosity defect according to an embodiment of the present specification is shown. A Cluster Porosity defect is a weld defect that occurs when the electrode of the flux is contaminated with moisture. Moisture turns into gas when heated and is trapped in the weld during the welding process. Cluster Porosity looks like regular pores in an X-ray weld image, but is tightly grouped together. As can be seen from the above, a Cluster Porosity defect occurs when there are high-density and tightly distributed dark spots in the weld image, and the shape and size uniformity of the spots, so that the Cluster Porosity defect can be characterized using image area features such as area shape, area brightness, and area density.

[0039] Figure 4D An example schematic diagram of an X-ray weld image with a slag inclusion defect according to an embodiment of the present specification is shown. Slag inclusion is a non-metallic solid material sandwiched in the weld metal or between the weld and the parent material. In the X-ray weld image, if there is a dark jagged asymmetric shape in or along the weld area, it indicates the presence of slag inclusion, so that image area features such as area shape, area position, and area density can be used to characterize the slag inclusion defect.

[0040] Figure 4EShows an exemplary schematic diagram of an X-ray weld image with an Incomplete Penetration / Lack of Penetration defect according to an embodiment of the present specification. The incomplete penetration defect is a weld defect that occurs when the welding metal fails to penetrate the joint. When the incomplete penetration defect occurs, there is a lack of permeability, resulting in natural stress concentration, which may cause cracks to propagate therefrom. If there is an incomplete penetration defect, a dark area will appear in the X-ray weld image. The edge of the dark area is clear and straight and extends downward along the pad or root surface at the center of the welded part, so that the incomplete penetration defect can be characterized by image area features such as area shape, area brightness, area position, and area density.

[0041] Figure 4F Shows an exemplary schematic diagram of an X-ray weld image with an Incomplete fusion defect according to an embodiment of the present specification. The incomplete fusion defect is a defect that occurs when the flux metal fails to properly fuse with the base metal. If the incomplete fusion defect occurs, one or more dark lines will appear in the X-ray weld image along the weld preparation area or the weld connection area and towards the weld direction, so that the incomplete fusion defect can be characterized by image area features such as area shape, area brightness, and area position.

[0042] Figure 4G Shows an exemplary schematic diagram of an X-ray weld image with an Internal Concavity / SuckBack defect according to an embodiment of the present specification. The internal concavity (suck-back) defect is a weld defect that occurs when the weld metal shrinks upon cooling and is drawn into the weld root. If the internal concavity defect occurs, similar to the incomplete penetration defect, a straight line with an irregular edge will appear in the X-ray weld image. The straight line is usually wide at the center of the weld image, so that the internal concavity defect can be characterized by image area features such as area shape and area position.

[0043] Figure 4H Shows an exemplary schematic diagram of an X-ray weld image with an External / Crown Undercut defect according to an embodiment of the present specification. The external undercut is the erosion of the base metal near the weld crown. If the external undercut defect occurs, a dark irregular line will appear in the X-ray weld image along the outer edge of the welding area, so that the external undercut defect can be characterized by image area features such as area shape, area position, and area brightness.

[0044] Figure 4IShows an exemplary schematic diagram of an X-ray weld image with an offset / mismatch defect according to an embodiment of the present specification. The offset defect is a weld defect that occurs when two parts welded together are not properly aligned. If an offset defect occurs, there will be an obvious difference in density between the two blocks, a dark straight line formed at the offset position, and inconsistent thickness in the weld area in the X-ray weld image. The difference in density is due to the difference in material thickness. The dark straight line is caused by the failure of the welding metal to fuse with the pad area. From the above description, it can be seen that image region features such as region shape, region brightness, and region density can be used to characterize the offset defect.

[0045] Figure 4J Shows an exemplary schematic diagram of an X-ray weld image with an inadequate weld reinforcement defect according to an embodiment of the present specification. The inadequate weld reinforcement defect refers to the existence of an area where the thickness of the weld metal deposited in the weld is less than the thickness of the base material. If an inadequate weld reinforcement defect occurs, in the X-ray weld image, there will be a weld area where the image density is higher (darker) than that of the surrounding base material, an area where the thickness of the weld metal is significantly less than the thickness of the base material, and the top contour of the weld is flat or concave, so that image region features such as region density and region shape can be used to characterize the inadequate weld reinforcement defect.

[0046] Figure 4K Shows an exemplary schematic diagram of an X-ray weld image with an excess weld reinforcement defect according to an embodiment of the present specification. The excess weld reinforcement defect refers to the occurrence of an area where the weld metal added in the weld exceeds the requirements of the engineering drawing and specifications. If an excess weld reinforcement defect occurs, in the X-ray weld image, there will be a weld area with a relatively light local density, a protruding weld contour, and the thickness of the weld metal significantly exceeding the thickness of the base material, etc., so that image region features such as region density and region shape can be used to characterize the excess weld reinforcement defect.

[0047] Figure 4L Shows an exemplary schematic diagram of an X-ray weld image with a crack defect according to an embodiment of the present specification. The crack defect can only be detected in the X-ray weld image when the crack propagates along the thickness change direction parallel to the X-ray beam. The crack defect presents serrated weak irregular lines, slender weak dark lines, and "tails" extending from slag inclusions or porosity defects in the X-ray weld image, so that image region features such as region shape and region brightness can be used to characterize the crack defect.

[0048] Figure 4MAn exemplary schematic diagram of an X-ray weld image with tungsten inclusion defects according to an embodiment of the present specification is shown. Tungsten inclusion defects are weld defects where tungsten is trapped in the weld due to the use of incorrect welding processes. Tungsten inclusion defects typically occur when welding using the TIG welding process. From a radiological perspective, tungsten is denser than aluminum or steel, causing tungsten to appear as a lighter-density area in the X-ray weld image with clear contours, usually having irregular or circular edges, and can be clearly distinguished from the surrounding weld metal. Thus, image region features such as region shape, region density, etc. can be used to characterize tungsten inclusion defects.

[0049] Figure 4N An exemplary schematic diagram of an X-ray weld image with oxide inclusion defects according to an embodiment of the present specification is shown. Oxide inclusions (oxide inclusion) are usually visible on the surface of the material being welded (especially aluminum). The density of oxide inclusions is generally lower than that of the surrounding material, so they appear as dark and irregularly shaped discontinuities in the X-ray weld image. These areas appear as lower-density discontinuities in the image. Thus, image region features such as region shape, region density, and region brightness can be used to characterize oxide inclusion defects.

[0050] Figure 4O An exemplary schematic diagram of an X-ray weld image with burn-through defects according to an embodiment of the present specification is shown. Burn-through defects occur when excessive heat causes excessive welding metal to penetrate the welding area. If a burn-through defect occurs, a black spot surrounded by a light-colored spherical area (icicle) appears in the X-ray weld image. Thus, image region features such as region shape, region brightness, etc. can be used to characterize burn-through defects.

[0051] From the above image analysis of defective X-ray weld images, it can be seen that if there are weld defects in the X-ray weld image, there will be region partitioning in the X-ray weld image and abnormal regions in the partitioned regions, such as abnormal region shape, abnormal region brightness, abnormal region density, etc. Thus, when identifying weld defects in an X-ray weld image, the image region features of the X-ray weld image can be selected as the model features of the weld defect recognition model for model training. After completing the training of the weld defect recognition model, image region analysis is performed on the X-ray weld image of the pipeline weld to obtain image region features, and the image region features are provided to the weld defect recognition model for weld defect recognition.

[0052] In some embodiments, the image region features used may include, for example, region position, region brightness, region density, and region shape. Region composition refers to how many regions the X-ray weld image consists of. Region position refers to the coordinate position information of each region. Region brightness refers to the brightness value of the region. Region density refers to the density value of the region, for example, radiographic density. Region shape may be, for example, the shape of the region edge, such as irregular lines, circular spots, etc. In some embodiments, the image region features may further include region composition, that is, how many regions the X-ray weld image consists of.

[0053] Figure 5 FIG. 4 shows an exemplary schematic diagram of a weld defect recognition model 500 according to an embodiment of the present specification. As Figure 5 shown, the weld defect recognition model 500 includes a defect feature generation layer 510 and a defect classification layer 520.

[0054] The defect feature generation layer 510 is used to generate a defect feature representation of the X-ray weld image based on the image region features obtained by performing image region analysis on the input X-ray weld image. The image region features of the X-ray weld image may include region position, region brightness, region density, and region shape. In some embodiments, the feature extraction layer 410 may be implemented using a convolutional neural network, and the generated defect feature representation is represented as a defect feature vector with a specific dimension, for example, a 20-dimensional defect feature vector. When generating the defect feature representation, the image region features of the X-ray weld image may be quantized, and the quantized image region features are provided to the convolutional neural network to generate the defect feature vector.

[0055] The defect classification layer 520 is used to classify weld defects based on the defect feature representation of the X-ray weld image, and thus output the weld defect category to complete the weld defect recognition. In some embodiments, the defect classification layer 520 may be implemented using a fully connected layer. In some embodiments, the defect classification layer 520 may implement multi-classification recognition, and the output result is each defect classification and the corresponding probability (confidence).

[0056] The weld defect recognition model training device 130 can pre-train a weld defect recognition model using a training data set, and store the trained weld defect recognition model in the weld defect recognition model storage device 140 for use by the weld defect recognition device 150. In some embodiments, the weld defect recognition model trained by the weld defect recognition model training device 130 can be directly deployed in the weld defect recognition device 150 without the weld defect recognition model storage device 140.

[0057] In some embodiments, the training dataset used in training the weld defect recognition model can be constructed based on the actual X-ray weld images at the pipe welds. In some embodiments, in order to expand the data scale of the training dataset and increase data diversity, in addition to using the actual X-ray weld images at the pipe welds to construct the training dataset, the actual X-ray weld images can also be data-augmented to obtain augmented X-ray weld images, and an augmented dataset can be constructed based on the augmented X-ray weld images to serve as the training dataset for data augmentation of the training dataset. The data augmentation performed on the actual X-ray weld images can include, for example, at least one of image rotation, image translation, image scaling, and filter-based image filtering.

[0058] After training the weld defect recognition model, the weld defect recognition device 150 can use the trained weld defect recognition model to perform weld defect recognition based on the X-ray weld image acquired by the weld image acquisition device 110.

[0059] Figure 6 An exemplary flowchart of a weld defect recognition method 600 according to an embodiment of the present specification is shown.

[0060] As Figure 6 shown, after acquiring the X-ray weld image at the pipe weld, at 610, based on the weld shape parameters, the identifiability of the X-ray weld image at the pipe weld is determined according to the identified X-ray weld image library, and the weld shape parameters include the build-up pitch of the welded part and the arc of the welded part.

[0061] In some embodiments, the acquired X-ray weld image can be an X-ray weld film image acquired by scanning the weld image acquisition device at a given angle. Due to the limitations of the X-ray radiation source used for X-ray scanning, this X-ray weld image is only a partial weld of the pipe weld. In some embodiments, the acquired X-ray weld image can be a 360-degree weld chain image formed by stitching multiple scanned film images at different angles at the pipe weld, for example, a 360-degree digitalized weld chain image formed by stitching multiple scanned film images at different angles at the pipe weld and performing digital twinning. In this case, since the acquired X-ray weld image is a complete weld image of the pipe weld, it is possible to prevent the occurrence of stitching the X-ray weld images acquired by scanning at different angles at pipe A to fake the X-ray weld image at pipe B.

[0062] In some embodiments, a weld image similarity determination model can be pre-trained. For example, a weld shape parameter can be used as a model feature to pre-train a weld image similarity determination model. After training the weld image similarity determination model, the X-ray weld image and the recognized X-ray weld images in the recognized X-ray weld image library are provided to the weld image similarity determination model to determine whether the X-ray weld image has been recognized. For example, the weld image similarity determination model can be used to determine the similarity between the X-ray weld image and the recognized X-ray weld images. If there is a recognized X-ray weld image with a similarity greater than a predetermined threshold, it is considered that the X-ray weld image provided to the weld defect recognition device has been recognized.

[0063] In some examples, the training dataset of the weld image similarity determination model can include a training dataset constructed based on actual X-ray weld images at the pipe weld. In some examples, the training dataset of the weld image similarity determination model can include a training dataset constructed based on actual X-ray weld images at the pipe weld and enhanced X-ray weld images obtained by data augmentation of the actual X-ray weld images. The data augmentation performed on the actual X-ray weld images can include, for example, at least one of image rotation, image translation, image scaling, and filter-based image filtering. According to the above model training method, since the training dataset for model training includes a training dataset that has been data-augmented, such as by image rotation, image translation, and image scaling, if the X-ray weld image provided to the weld image similarity determination model is obtained by data-augmenting a recognized X-ray weld image, such as by image rotation, image translation, and image scaling, the weld image similarity determination model can still determine it to be similar to the recognized X-ray weld image. This improves the ability of the weld image similarity determination model to distinguish fake new X-ray weld images obtained by data-augmenting recognized X-ray weld images, avoids repeated recognition of recognized weld images, and thus improves the efficiency of weld defect recognition.

[0064] Figure 7 An example flowchart of a weld image recognition determination process 700 according to an embodiment of the present specification is shown.

[0065] As Figure 7 shown, at 710, the weld shape parameters of the X-ray weld image to be recognized are obtained.

[0066] At 720, based on the acquired weld shape parameters, it is determined whether the X-ray weld image has been recognized according to the recognized X-ray weld image library. For example, the weld shape parameters of the X-ray weld image to be recognized can be compared with the weld shape parameters of each recognized X-ray weld image in the recognized X-ray weld image library. If it is determined based on the comparison result of the weld shape parameters that the similarity between two X-ray weld images exceeds a predetermined threshold, it is considered that there is a recognized X-ray weld image similar to the X-ray weld image to be recognized, and thus it is determined that the X-ray weld image to be recognized has been recognized.

[0067] In some embodiments, the original image data of the recognized X-ray weld image is stored in the recognized X-ray weld image library. In some examples, the digital twin image data obtained by performing digital twin on the original image data can also be stored in the recognized X-ray weld image library. In this case, the acquired weld shape parameters are the actual shape parameters of the X-ray weld image. For example, a computer image recognition algorithm can be used to perform weld shape recognition on the X-ray weld image, thereby obtaining the weld shape parameters of the X-ray weld image to be recognized.

[0068] In some embodiments, the weld shape feature vectors obtained based on the weld shape parameters of the recognized X-ray weld image are stored in the recognized X-ray weld image library. In this case, when obtaining the weld shape parameters of the X-ray weld image to be recognized, the X-ray weld image to be recognized can be provided to the weld shape feature vector extraction model to obtain the weld shape feature vector of the X-ray weld image to be recognized. Subsequently, the weld image similarity is determined based on the weld shape feature vector.

[0069] In some embodiments, significant shape feature extraction can be performed on the X-ray weld image to extract a partial X-ray weld image with significant shape features from the X-ray weld image. Subsequently, the recognition of the X-ray weld image is determined based on the extracted partial X-ray weld image.

[0070] In some embodiments, global feature comparison can be performed on the X-ray weld image to determine the image similarity with the recognized X-ray weld image. In some embodiments, local feature comparison can also be performed on the X-ray weld image to determine the image similarity with the recognized X-ray weld image. For example, the image similarity is determined based on the intercepted segment of the weld image intercepted at a given angle.

[0071] In some embodiments, the X-ray weld image library may store the identified X-ray weld images and the 360-degree complete weld digital twin image chains formed by splicing each weld image. In this case, the X-ray weld image, the identified X-ray weld images in the identified X-ray weld image library, and the 360-degree complete weld digital twin image chains formed by splicing each weld image may be provided to the weld image similarity determination model to determine whether the X-ray weld image has been identified or is part of the 360-degree complete weld.

[0072] Return to Figure 6 , after obtaining the identification determination result as above, at 620, according to the identification determination result of 610, it is determined whether the X-ray weld image at the pipe weld has been identified. In response to the X-ray weld image at the pipe weld having been identified, at 630, feedback information indicating that the defect identification of the X-ray weld image at the pipe weld has been completed is returned.

[0073] In response to the X-ray weld image at the pipe weld not being identified, at 640, image region analysis is performed on the X-ray weld image at the pipe weld to obtain the image region features of the X-ray weld image. The obtained image region features include region composition, region position, region brightness, region density, and region shape. Subsequently, at 650, the image region features of the X-ray weld image are provided to the weld defect identification model for weld defect identification.

[0074] Optionally, after identifying weld defects using the weld defect identification model, a weld defect identification report may also be generated based on the weld defect identification result. The generated weld defect identification report may include the position information of the pipe weld, whether there are weld defects, and weld defect category information, etc. The weld defect identification report may be notified to the user in the form of text, audio, and / or video.

[0075] It should be noted that in other alternative embodiments, the process of determining whether the weld image is identified shown in 610 may not be included either.

[0076] Using the above weld defect identification method, by using the weld defect identification model to perform weld defect identification based on image region features such as the region position, region brightness, region density, and region shape of the X-ray weld image, the efficiency and accuracy of weld defect identification can be improved.

[0077] Using the above weld defect identification method, by screening the identified weld images based on the uniqueness of the weld shape, repeated identification of the identified weld images can be avoided, thereby improving the efficiency of weld defect identification. At the same time, the occurrence of weld image fraud can be prevented, and the reliability of weld defect identification can be improved.

[0078] Using the above weld defect recognition method, a training data set is constructed by using the actual X-ray weld images at the pipe welds and the enhanced X-ray weld images obtained by performing data augmentation on the actual X-ray weld images when determining the model for the similarity of the training weld images. Thus, when the X-ray weld image provided to the weld image similarity determination model is obtained by performing data augmentation processing such as image rotation, image translation, image scaling, and filter-based image filtering on the recognized X-ray weld image, the weld image similarity determination model can still determine it as similar to the recognized X-ray weld image. Thereby, the discrimination ability of the weld image similarity determination model for counterfeiting new X-ray weld images by performing data augmentation processing on the recognized X-ray weld image is improved, and the repeated recognition of the recognized weld image is avoided, thereby improving the weld defect recognition efficiency.

[0079] Figure 8 FIG. shows an example block diagram of a weld defect recognition device 800 according to an embodiment of the present specification. As Figure 8 shown, the weld defect recognition device 800 includes a weld image recognition determination unit 810, an image region analysis unit 820, and a weld defect recognition unit 830.

[0080] The weld image recognition determination unit 810 is configured to determine whether the X-ray weld image at the pipe weld has been recognized based on the weld shape parameters according to the recognized X-ray weld image library, and the weld shape parameters include the build-up pitch of the welded part and the arc of the welded part. The operation of the weld image recognition determination unit 810 can refer to the operations described above with reference to Figure 6 610 and Figure 7 described above.

[0081] The image region analysis unit 820 is configured to perform image region analysis on the X-ray weld image at the pipe weld to obtain the image region features of the X-ray weld image, and the obtained image region features include region position, region brightness, region density, and region shape. The operation of the image region analysis unit 820 can refer to the operations described above with reference to Figure 6 640 described above.

[0082] The weld defect recognition unit 830 is configured to provide the image region features of the X-ray weld image to the weld defect recognition model for weld defect recognition. The operation of the weld defect recognition unit 830 can refer to the operations described above with reference to Figure 6 650 described above.

[0083] It should be noted that in other alternative embodiments, the weld defect recognition device 800 may not include the weld image recognition determination unit 810.

[0084] As described above with reference toFigures 1 to 8 A weld defect identification method and a weld defect identification device according to an embodiment of the present specification are described. The above-mentioned weld defect identification device can be implemented by hardware, or can be implemented by software or a combination of hardware and software.

[0085] Figure 9 FIG. shows an exemplary schematic diagram of a weld defect identification device 900 implemented based on a computer system according to an embodiment of the present specification. As Figure 9 shown, the weld defect identification device 900 may include at least one processor 910, a memory (e.g., a non-volatile memory) 920, a memory 930, and a communication interface 940, and at least one processor 910, the memory 920, the memory 930, and the communication interface 940 are connected together via a bus 960. At least one processor 910 executes at least one computer-readable instruction stored or encoded in the memory (i.e., the above-mentioned elements implemented in software form).

[0086] In one embodiment, computer-executable instructions are stored in the memory, which when executed cause at least one processor 910 to: perform image region analysis on an X-ray weld image at a pipeline weld to obtain image region features of the X-ray weld image, and the obtained image region features include region position, region brightness, region density, and region shape; and provide the image region features of the X-ray weld image to a weld defect identification model for weld defect identification.

[0087] It should be understood that the computer-executable instructions stored in the memory, when executed, cause at least one processor 910 to perform the various operations and functions described above in the various embodiments of the present specification in combination with Figures 1 - 8 description.

[0088] According to one embodiment, a program product such as a machine-readable medium (e.g., a non-transitory machine-readable medium) is provided. The machine-readable medium may have instructions (i.e., the above-mentioned elements implemented in software form), which when executed by the machine, cause the machine to perform the various operations and functions described above in the various embodiments of the present specification in combination with Figures 1 - 8 description. Specifically, a system or device equipped with a readable storage medium may be provided, on which software program code for implementing the functions of any one of the above-mentioned embodiments is stored, and the computer or processor of the system or device reads and executes the instructions stored in the readable storage medium.

[0089] In this case, the program code read from the readable medium itself can implement the functions of any one of the above-mentioned embodiments, so the machine-readable code and the readable storage medium storing the machine-readable code constitute a part of the present invention.

[0090] Examples of readable storage media include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD-RW), magnetic tapes, non-volatile memory cards, and ROMs. Optionally, program code can be downloaded from a server computer or the cloud via a communication network.

[0091] According to one embodiment, there is provided a computer program product including a computer program which, when executed by a processor, causes the processor to perform the various operations and functions described above in connection with the various embodiments of this specification. Figures 1 - 8 Description of the various operations and functions.

[0092] Those skilled in the art should understand that various modifications and variations can be made to the above-disclosed embodiments without departing from the essence of the invention. Therefore, the scope of protection of the present invention should be defined by the appended claims.

[0093] It should be noted that not all steps and units in the above processes and system structure diagrams are necessary, and some steps or units can be ignored according to actual needs. The execution order of each step is not fixed and can be determined as required. The device structures described in the above embodiments can be physical structures or logical structures. That is, some units may be implemented by the same physical entity, or some units may be implemented by multiple physical entities separately, or some components in multiple independent devices may be jointly implemented.

[0094] In the above embodiments, the hardware units or modules can be implemented mechanically or electrically. For example, a hardware unit, module, or processor can include permanent dedicated circuits or logic (such as a dedicated processor, FPGA, or ASIC) to perform corresponding operations. The hardware unit or processor can also include programmable logic or circuits (such as a general-purpose processor or other programmable processors) that can be temporarily set by software to perform corresponding operations. The specific implementation method (mechanical method, or dedicated permanent circuit, or temporarily set circuit) can be determined based on cost and time considerations.

[0095] The specific embodiments described above in conjunction with the accompanying drawings describe exemplary embodiments, but do not represent all embodiments that can be implemented or fall within the scope of protection of the claims. The term "exemplary" used throughout this specification means "serving as an example, instance, or illustration" and does not mean "preferred" or "advantageous" over other embodiments. For the purpose of providing an understanding of the described technology, the specific embodiments include specific details. However, the technology can be practiced without these specific details. In some instances, well-known structures and devices are shown in block diagram form in order to avoid obscuring the concepts of the described embodiments.

[0096] The foregoing description of the disclosure has been provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the scope of the disclosure. Thus, the disclosure is not limited to the examples and designs described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying weld defects based on X-ray weld images, comprising: Providing the X-ray weld image at the pipe weld, the identified X-ray weld images in the identified X-ray weld image library, and the 360-degree complete weld digital twin image chain formed by splicing the identified X-ray weld images to a weld image similarity determination model to determine whether the X-ray weld image has been identified. The weld image similarity determination model is pre-trained using the weld shape parameters of the X-ray weld image as model features. The weld shape parameters include the build-up pitch of the welded part and the arc of the welded part. The training dataset of the weld image similarity determination model includes a training dataset constructed based on the actual X-ray weld image at the pipe weld and the enhanced X-ray weld image obtained by data augmentation of the actual X-ray weld image. The data augmentation includes at least one of image rotation, image translation, image scaling, and filter-based image filtering; In response to the X-ray weld image not being identified, performing image region analysis on the X-ray weld image to obtain the image region features of the X-ray weld image. The image region features include region position, region brightness, region density, and region shape. The region position includes the coordinate position information of the image region. The region brightness includes the brightness value of the image region. The region density includes the radiographic density value of the image region. The region shape includes the region edge shape of the image region; and Providing the image region features of the X-ray weld image to a weld defect identification model to perform weld defect identification. The weld defect identification model is implemented as a multi-classification model. The identified multi-defect categories include cold lap defect, porosity defect, dense porosity defect, slag inclusion defect, incomplete penetration defect, lack of fusion defect, internal concavity defect, external undercut defect, misalignment defect, incomplete fusion defect, excessive weld surplus defect, crack defect, tungsten inclusion defect, oxygen inclusion defect, and burn-through defect.

2. The weld defect recognition method according to claim 1, wherein, Providing the image region features of the X-ray weld image to a weld defect identification model to perform weld defect identification includes: Providing the image region features of the X-ray weld image to the defect feature generation layer of the weld defect identification model to generate a defect feature representation of the X-ray weld image; and Providing the defect feature representation of the X-ray weld image to the defect classification layer of the weld defect identification model to perform weld defect identification.

3. The weld defect recognition method according to claim 1, wherein, The X-ray weld image includes a 360-degree weld chain image formed by splicing multiple scanned film images at different angles at the pipe weld.

4. The weld defect identification method according to claim 1, wherein, The training dataset of the weld defect identification model includes a training dataset constructed based on the actual X-ray weld image at the pipe weld and the enhanced X-ray weld image obtained by data augmentation of the actual X-ray weld image. The data augmentation includes at least one of image rotation, image translation, image scaling, and filter-based image filtering.

5. The weld defect identification method according to claim 1, further comprising: Generating a weld defect identification report based on the weld defect identification result.

6. A weld defect recognition device based on X-ray weld images, comprising: An image recognition determination unit configured to provide the X-ray weld image at the pipe weld, the recognized X-ray weld images in the recognized X-ray weld image library, and the 360-degree complete weld digital twin image chain stitched together by the recognized X-ray weld images to a weld image similarity determination model to determine whether the X-ray weld image has been recognized. The weld image similarity determination model is pre-trained using the weld shape parameters of the X-ray weld image as model features. The weld shape parameters include the build-up pitch of the welded part and the arc of the welded part. The training dataset of the weld image similarity determination model includes a training dataset constructed based on the actual X-ray weld images at the pipe weld and the enhanced X-ray weld images obtained by data augmentation of the actual X-ray weld images. The data augmentation includes at least one of image rotation, image translation, image scaling, and filter-based image filtering; An image region analysis unit configured to, in response to the X-ray weld image not being recognized, perform image region analysis on the X-ray weld image to obtain the image region features of the X-ray weld image. The image region features include region position, region brightness, region density, and region shape. The region position includes the coordinate position information of the image region, the region brightness includes the brightness value of the image region, the region density includes the radiographic density value of the image region, and the region shape includes the region edge shape of the image region; and A weld defect recognition unit configured to provide the image region features of the X-ray weld image to a weld defect recognition model for weld defect recognition. The weld defect recognition model is implemented as a multi-classification model, and the recognized multi-defect categories include cold lap defect, porosity defect, dense porosity defect, slag inclusion defect, incomplete penetration defect, lack of fusion defect, internal concavity defect, external undercut defect, misalignment defect, incomplete fusion defect, excessive weld allowance defect, crack defect, tungsten inclusion defect, oxygen inclusion defect, and burn-through defect.

Images