Training data acquisition device, welding system, and training data acquisition method

Abstract

An AI model generation device 2 is provided with: a first input device 24 that receives a weld scar specification input from a user specifying the shape of a convex weld bead; and a processor 22 that creates defectless image data indicating a 3D image including a weld bead in accordance with the weld scar specification input to the first input device 24, and creates training image data indicating a 3D image including a weld defect and the weld bead on the basis of the defectless image data and defective image data indicating a 3D image including the weld defect.

Description

Learning data acquisition device, welding system, and learning data acquisition method

[0001] The present disclosure relates to a learning data acquisition device, a welding system, and a learning data acquisition method for acquiring image data used in machine learning.

[0002] Patent Literature 1 discloses a learning device that generates a model through machine learning using training data. This learning device determines whether each piece of training data collected by a user is appropriate or not, and excludes training data that is determined to be inappropriate.

[0003] International Publication No. 2019 / 187594

[0004] It is conceivable to create a model that outputs welding defect information, including the location of welding defects, based on image data of a 3D image by machine learning. To generate such a model, training data is required, which includes multiple data sets: image data of 3D images of the area around the weld mark captured by a 3D sensor, and welding defect information to be output based on the image data. One conceivable method for creating the training data is to acquire 3D images of a weld mark that does not include welding defects and 3D images of welding defects by measurement with a 3D sensor, process them, and combine them. However, this method requires forming and measuring a weld mark that does not include welding defects, which is time-consuming and requires a long work time.

[0005] The present disclosure has been made in consideration of these points, and its purpose is to reduce the effort and work time required to acquire learning data.

[0006] In order to achieve the above object, the present disclosure provides a learning data acquisition device that acquires learning image data to be used in machine learning for generating a model, the model outputting welding defect information including the location of welding defects based on image data of a 3D image, and the learning data acquisition device is characterized by comprising: an input unit that accepts a user's welding defect identification input that specifies at least one of the shape and size of a convex or concave welding defect; a welding defect image creation step that creates welding defect image data showing a 3D image including the welding defect in accordance with the welding defect identification input to the input unit; and a learning image creation step that creates learning image data showing a 3D image including the welding defect and the welding defect based on the welding defect image data created by the welding defect image creation step and defect-present image data showing a 3D image including the welding defect.

[0007] As a result, the learning data acquisition unit creates weld mark image data in response to the weld mark identification input, so there is no need to form weld marks that do not contain welding defects by welding or to measure the shapes of the weld marks with a 3D sensor in order to acquire the weld mark image data, thereby reducing the effort and work time required to acquire learning image data.

[0008] According to the present disclosure, the effort and work time required to acquire learning data can be reduced.

[0009] FIG. 1 is a block diagram illustrating a configuration of a welding system including an AI model generation device as a learning data acquisition device according to an embodiment of the present disclosure. FIG. 2 is an explanatory diagram illustrating an example of an image with a defect in which a mark is added to an area identified by annotation data. FIG. 3 is an explanatory diagram illustrating an example of an image with a defect in which a mark is added to an area where a welding defect is formed, identified by segmentation data. FIG. 4 is an explanatory diagram illustrating an image without a defect. FIG. 5 is an explanatory diagram illustrating an example of an image in which a defect-free image has been processed by solid-coloring a weld bead formation area based on segmentation data. FIG. 6 is a flowchart illustrating an operation of creating a defect detection model by the AI ​​model generation device. FIG. 7 is an explanatory diagram illustrating an example of an image prompting a user to input the width, length, and thickness of a base material. FIG. 8 is an explanatory diagram illustrating an example of an image prompting a user to input a designation for the cross-sectional shape of a weld bead. FIG. 9 is an explanatory diagram illustrating an example of an image prompting a user to input a designation for the top surface shape of a weld bead. FIG. 10 is an explanatory diagram illustrating an example of an image prompting a user to input a designation for the side surface shape of a weld bead. FIG. 11 is an explanatory diagram illustrating a method for creating segmentation data of defect-free image data.

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. The following description of the preferred embodiments is merely exemplary in nature and is not intended to limit the present invention, its applications, or its uses.

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] 1 shows a welding system 1. The welding system 1 includes an AI model generation device 2 as a learning data acquisition device according to an embodiment of the present disclosure, and a welding device 3 that performs welding.

[0013] The AI ​​model generation device 2 generates a defect detection model through machine learning. The AI ​​model generation device 2 also acquires learning image data that constitutes learning data used in the machine learning for generating the defect detection model.

[0014] The defect detection model is an object detection model that predicts and outputs welding defect information based on image data of an input 3D image. The defect detection model is represented by a CNN (Convolutional Neural Network). The welding defect information includes category information indicating the type of welding defect in the 3D image represented by the image data, and position information indicating the position of the welding defect. In detail, the category information indicates whether the welding defect is a hole, pit, spatter, undercut, protrusion, etc. A pit is an opening on the surface of a convex weld bead as a welding mark. The position information indicates the position of a bounding box surrounding the welding defect. Note that the bounding box is a rectangular boundary that surrounds an object such as a welding defect.

[0015] In addition, the AI ​​model generation device 2 creates defective welding information for each training image data, and generates a defect detection model through machine learning using multiple data sets of the training image data and the defective welding information as training data.

[0016] Specifically, the AI ​​model generation device 2 includes a memory device 21, a processor 22 as a learning data acquisition unit, a first output device 23, and a first input device 24 as an input unit.

[0017] The storage device 21 includes a defect-containing image data storage unit 211 , a defect-free image data storage unit 212 , an annotation data storage unit 213 , a segmentation data storage unit 214 , a learning image data storage unit 215 , and a parameter storage unit 216 .

[0018] The defect image data storage unit 211 stores, as defect image data, image data of a 3D image acquired by capturing an image of the area around the welding point PW including the welding defect using the 3D sensor 36 described later.

[0019] The defect-free image data storage unit 212 stores defect-free image data as welding mark image data generated by the processor 22 .

[0020] The annotation data storage unit 213 stores annotation data for the defect-containing image data stored in the defect-containing image data storage unit 211. This annotation data specifies a rectangular area (bounding box) including a welding defect in the defect-containing image based on the defect-containing image data. The annotation data is specified by the user inputting defect information to the first input device 24 while the defect-containing image is being output to the first output device 23.

[0021] The segmentation data storage unit 214 stores segmentation data of the defect-containing image data stored in the defect-containing image data storage unit 211. This segmentation data specifies the boundary between a region where a welding defect has occurred and a region where a welding defect has not occurred in the defect-containing image based on the defect-containing image data. This segmentation data is specified by a boundary specification input by the user to the first input device 24 while the defect-containing image is being output to the first output device 23.

[0022] The segmentation data storage unit 214 also stores segmentation data of the defect-free image data stored in the defect-free image data storage unit 212. This segmentation data specifies the boundary between areas where welding defects have occurred and areas where welding defects have not occurred in the defect-free image based on the defect-free image data. This segmentation data is created by the defect-free data processing unit 222.

[0023] The learning image data storage unit 215 stores learning image data created by the processor 22 .

[0024] The parameter storage unit 216 stores parameters that specify a fault detection model generated by the processor 22, which will be described later.

[0025] The processor 22 has a defect-containing data processing unit 221, a defect-free data processing unit 222, a learning data creation unit 223, and an AI model generation unit 224.

[0026] The defect data processing unit 221 receives defect image data indicating a 3D image acquired by capturing an image of the area around the welding point PW, including a welding defect, using the 3D sensor 36 described below from the welding device 3, and stores the defect image data in the defect image data storage unit 211. The defect data processing unit 221 also performs a data format conversion process on the defect image data stored in the defect image data storage unit 211, and stores the defect image data in the converted data format in the defect image data storage unit 211. The data format conversion process performed here is a process of converting the data format from CSV data to NPY data.

[0027] Furthermore, the defect-containing data processing unit 221 outputs a defect-containing image based on the defect-containing image data stored in the defect-containing image data storage unit 211 to the first output device 23. In this state, the first input device 24 accepts a user's input of defect information specifying the type and position of a welding defect in the defect-containing image. When the user finishes inputting the defect information, the defect-containing data processing unit 221 associates annotation data identified based on the user's input of the defect information to the first input device 24 with the defect-containing image data of each image and stores the annotation data in the annotation data storage unit 213. For example, when a user specifies the position and type of a welding defect by inputting defect information while the defect-containing image is being output to the first output device 23, an image in which a square mark is added to the welding defect is output to the first output device 23, as shown in FIG. 2 .

[0028] Furthermore, the defect-containing data processing unit 221, while outputting the defect-containing image to the first output device 23, accepts a boundary identification input from the user that identifies the boundary between the defective weld region and the non-defective weld region in the defect-containing image. In response to the boundary identification input, the defect-containing data processing unit 221 associates segmentation data indicating the boundary between the defective weld region and the non-defective weld region with the defect-containing image data of each image and stores it in the segmentation data storage unit 214. For example, while outputting the defect-containing image to the first output device 23, if the user identifies the boundary between the defective weld region and the non-defective weld region by the boundary identification input, an image in which the defective weld region is shaded is output to the first output device 23, as shown in FIG. 3. On the actual screen of the first output device 23, the pit region indicated by diagonal hatching in FIG. 3 and the spatter region indicated by black in FIG. 3 are shaded in different colors.

[0029] The defect-free data processing unit 222 receives a base material designation input from the user to the first input device 24, which designates the size of the base material W on which the weld bead will be formed, and a weld mark identification input from the user to the first input device 24, which designates the shape of the weld bead. The defect-free data processing unit 222 then generates defect-free image data representing a defect-free image, which is a 3D image including the weld bead and the base material W, in accordance with the base material designation input and the weld mark identification input, and stores the defect-free image data in the defect-free image data storage unit 212. FIG. 4 illustrates an example of a defect-free image represented by the defect-free image data. The defect-free data processing unit 222 also automatically generates segmentation data identifying the boundary between a weld bead (weld mark) formation area and a non-weld mark formation area, and stores the segmentation data in the segmentation data storage unit 214. In FIG. 5, areas identified as weld mark formation areas by the segmentation data of the defect-free image are hatched with diagonal lines. The defect-free data processing unit 222 is preferably capable of creating defect-free image data that shows defect-free images including weld beads with a height of 0.5 mm or less, and more preferably capable of creating defect-free image data that shows defect-free images including weld beads with a height of 0.3 mm or less.

[0030] The learning data creation unit 223 creates learning image data based on the defect-containing image data stored in the defect-containing image data storage unit 211, the defect-free image data stored in the defect-free image data storage unit 212, the annotation data for the defect-containing image data stored in the annotation data storage unit 213, and the segmentation data for the defect-containing image data and the defect-free image data stored in the segmentation data storage unit 214. Specifically, data expansion is performed on image data representing rectangular areas specified by the annotation data among the 3D images represented by the defect-containing image data. The data expansion here involves obtaining multiple image data from a single image by performing one or more of the following processes on the image data of each rectangular area: inversion, transposition, enlargement or reduction, and synthesis of the welding defect formation area. The welding defect formation area can be identified by the segmentation data. The learning data creation unit 223 then creates learning image data of multiple types of 3D images using the multiple image data obtained by data expansion. The 3D images represented by these multiple types of learning image data are images in which welding defects of 3D images represented by multiple image data obtained by data expansion are pasted onto images without defects in various combinations and patterns. The learning data creation unit 223 then stores learning data consisting of multiple data sets of the created learning image data and welding defect information corresponding to the learning image data in the learning image data storage unit 215.

[0031] The AI ​​model generation unit 225 generates a fault detection model by performing machine learning using the training data stored in the training image data storage unit 215. Specifically, the AI ​​model generation unit 225 specifies the weights and biases of each node constituting the CNN that expresses the fault detection model as parameters that specify the fault detection model. Then, the AI ​​model generation unit 225 stores the specified parameters in the parameter storage unit 216.

[0032] The first output device 23 outputs an image based on the image data stored in the defective image data storage unit 211. The first output device 23 is configured, for example, by a liquid crystal monitor. Furthermore, as will be described in detail later, the first output device 23 outputs an image prompting the user to enter various inputs.

[0033] The first input device 24 receives the defect information input, the base material designation input, and the welding mark identification input from the user, and transmits a signal corresponding to the received input to the processor 22 .

[0034] The welding device 3 has a welding torch 31, a wire feeder (not shown), a welding power source 32, an output control unit 33, a robot arm 34, a robot control unit 35, a 3D sensor 36, a computer 37 serving as a defect information calculation unit, a second output device 38, and a second input device 39. When power is supplied from the welding power source 32 to the welding wire WI held by the welding torch 31, an arc is generated between the tip of the welding wire WI and the base metal W, and the base metal W is heated to perform arc welding. Note that the welding device 3 has other components and equipment such as piping and gas cylinders for supplying shielding gas to the welding torch 31, but for convenience of explanation, these are not shown or described.

[0035] Output control unit 33 is connected to welding power source 32 and a wire feeder (not shown) and controls the welding output of welding torch 31, in other words, the power supplied to welding wire WI and the power supply time, in accordance with predetermined welding conditions. Output control unit 33 also controls the feed speed and feed amount of welding wire WI fed from the wire feeder (not shown) to welding torch 31. Note that the welding conditions may be input directly to output control unit 33 via an input unit (not shown), or may be selected from a welding program separately read from a recording medium or the like. Welding torch 31, wire feeder (not shown), welding power source 32, and output control unit 33 constitute a welding unit that performs welding.

[0036] Robot arm 34 is a known articulated robot that holds welding torch 31 at its tip and is connected to robot control unit 35. Robot control unit 35 controls the operation of robot arm 34 so that the tip of welding torch 31, in other words, the tip of welding wire WI held by welding torch 31, moves to a desired position while tracing a predetermined welding trajectory.

[0037] The 3D sensor 36 is attached to the welding torch 31 and measures the welded portion PW of the base metal W, specifically, the shape of the welded portion PW, to acquire shape data. The 3D sensor 36 is a three-dimensional shape measurement sensor composed of, for example, a laser light source (not shown) configured to scan the surface of the base metal W and a camera (not shown) that captures the reflection trajectory of the laser light projected on the surface of the base metal W (hereinafter, sometimes referred to as a shape line). The 3D sensor 36 scans the entire welded portion PW of the base metal W with a laser beam, and the camera captures the laser beam reflected by the welded portion PW, thereby measuring the shape of the welded portion PW. The 3D sensor 36 is configured to measure the shape of not only the welded portion PW but also a predetermined range around it. This is to evaluate the presence or absence of spatter, etc. The camera has a CCD or CMOS image sensor as an imaging element. The configuration of the 3D sensor 36 is not limited to the above, and other configurations are possible. For example, an optical interferometer may be used instead of the camera.

[0038] The computer 37 executes software implemented on a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) to realize the functions of a plurality of functional blocks within the computer 37. The computer 37 has an image processing unit 371, a welding defect information calculation unit 372, and a welding defect display image generation unit 373.

[0039] The image processing unit 371 receives the shape data acquired by the 3D sensor 36 and, based on this shape data, acquires measurement image data that shows a 3D image including the welding point PW. For example, the image processing unit 371 acquires point cloud data of the shape lines captured by the 3D sensor 36. The image processing unit 371 also corrects the inclination and distortion of the base portion of the welding point PW relative to a predetermined reference plane, for example, the installation surface of the base material W, by statistically processing the point cloud data, and acquires measurement image data that includes the welding point PW.

[0040] The defective welding information calculation unit 372 calculates defective welding information based on the measurement image data acquired by the image processing unit 371, using the defect detection model generated by the AI ​​model generation device 2. The calculation of the defective welding information may be performed when a predetermined input operation is performed on the second input device 39.

[0041] The defective welding display image generating unit 373 calculates a pixel value of each pixel of the defective welding display image displaying the defective welding information, based on the defective welding information calculated by the defective welding information calculating unit 372. The information specifying the defective welding display image may be compressed using a JPEG format or the like.

[0042] The second output device 38 displays the poor welding display image based on the pixel value of each pixel calculated by the poor welding display image generating unit 373. The second output device 38 is configured by a liquid crystal monitor or the like.

[0043] The second input device 39 accepts a predetermined input operation by the user to cause the computer 37 to start calculating the defective welding information.

[0044] Next, the operation of creating a fault detection model by the AI ​​model generation device 2 will be described with reference to the flowchart of FIG.

[0045] First, in (S101), the user starts the power supply to the welding power source 32 of the welding device 3 and performs arc welding to form a weld bead along with a weld defect on the base material W. Then, the user causes the 3D sensor 36 to capture an image of the welded portion PW. Thereafter, the image processing unit 371 acquires defect image data that indicates an image including a weld defect based on the shape data acquired by the 3D sensor 36.

[0046] Next, in (S102), the defect data processing unit 221 of the AI ​​model generation device 2 receives the defect image data acquired in (S101) from the welding device 3, converts the data format from CSV data to NPY data, and stores it in the defect image data storage unit 211.

[0047] Next, in (S103), the defect-containing data processing unit 221 outputs a defect-containing image based on the defect-containing image data stored in the defect-containing image data storage unit 211 to the first output device 23. In this state, the first input device 24 accepts a user's input of defect information specifying the type and position of a welding defect in the defect-containing image. After the user has finished inputting the defect information, the defect-containing data processing unit 221 associates annotation data identified based on the user's input of the defect information to the first input device 24 with the defect-containing image data of each image and stores the annotation data in the annotation data storage unit 213. For example, if the user specifies the position and type of a welding defect by inputting defect information while the defect-containing image is being output to the first output device 23, an image in which a square mark is added to the welding defect is output to the first output device 23, as shown in FIG. 2 .

[0048] Next, in (S104), the defect-containing data processing unit 221 outputs a defect-containing image based on the defect-containing image data stored in the defect-containing image data storage unit 211 to the first output device 23. Then, in this state, the defect-containing data processing unit 221 receives a boundary identification input from the user that identifies the boundary between the defective weld formation area and the non-defective weld formation area in the defect-containing image. In response to the boundary identification input, the defect-containing data processing unit 221 associates segmentation data indicating the boundary between the defective weld formation area and the non-defective weld formation area with the defect-containing image data of each image and stores it in the segmentation data storage unit 214. For example, when the user identifies the boundary between the defective weld formation area and the non-defective weld formation area by the boundary identification input while the defect-containing image is being output to the first output device 23, an image in which the defective weld formation area is painted is output to the first output device 23, as shown in FIG. 3 .

[0049] Next, in (S105), when the user inputs that a base material W will be added, the defect-free data processing unit 222 causes the first output device 23 to output an image prompting the user to input a base material designation, specifying the size of the base material W on which the weld bead will be formed. The first output device 23 displays, for example, an image as shown in FIG. 7 , prompting the user to input the width, length, and thickness of the plate-shaped base material W. The user can input the width, length, and thickness of the base material W to the first input device 24 and select "add," thereby performing a base material designation input that specifies the size of the base material W. When the user inputs the base material designation input, the defect-free data processing unit 222 accepts this base material designation input.

[0050] Next, in (S106), when the user inputs a request to add a weld bead, the defect-free data processing unit 222 outputs to the first output device 23 an image prompting the user to input a weld mark specification to specify the shape of the weld bead. The defect-free data processing unit 222 displays, for example, an image such as that shown in FIG. 8 on the first output device 23, prompting the user to input a cross-sectional shape of the weld bead. In this state, when the user uses the first input device 24 to select a type of cross-sectional shape ("chevron" in FIG. 8), specify coordinates of each point, and select "Next," the defect-free data processing unit 222 displays, for example, an image such as that shown in FIG. 9 on the first output device 23, prompting the user to input a top surface shape of the weld bead. In this state, when the user uses the first input device 24 to specify coordinates of each point of the top surface shape and select "Next," the defect-free data processing unit 222 displays, for example, an image such as that shown in FIG. 10 on the first output device 23, prompting the user to input a side surface shape of the weld bead. In this state, the user specifies the coordinates of each point of the side shape using the first input device 24 and selects "Next." In this way, the user can perform a weld mark specifying input that specifies the shape of the weld bead using the first input device 24. Note that an input that specifies the position of the weld bead on the base material W may also be performed as a weld mark specifying input.

[0051] Thereafter, in (S107), in response to a predetermined input by the user, the defect-free data processing unit 222 creates defect-free image data that shows an image without defects, such as that shown in Fig. 4. In the defect-free image, a weld bead having a shape specified by the weld mark identification input in (S106) is formed on the base material W having a size specified by the base material designation input in (S105).

[0052] Thereafter, in (S108), the defect-free data processing unit 222 automatically creates segmentation data that identifies the boundary between the weld bead formation area and the weld bead non-formation area based on the defect-free image data, and stores the data in the segmentation data storage unit 214. For example, as shown in Fig. 11, in the defect-free image, squares within five squares from the end of the area (the dotted area) that protrudes from the base material W are identified vertically every square and horizontally every four squares, and a straight line L connecting the identified squares is set as the boundary.

[0053] Next, in (S109), the learning data creation unit 223 creates learning image data based on the image data with defects stored in the image data storage unit 211, the image data without defects stored in the image data storage unit 212, the annotation data for the image data with defects stored in the annotation data storage unit 213, and the segmentation data for the image data with defects and the image data without defects stored in the segmentation data storage unit 214. At this time, as described above, the learning data creation unit 223 performs data extension on image data representing an image of a rectangular area specified by the annotation data among the 3D images represented by the image data with defects. The learning data creation unit 223 creates learning image data of multiple types of 3D images using multiple image data obtained by data extension. The 3D images represented by these multiple types of learning image data are images in which welding defects from 3D images represented by the multiple image data obtained by data extension are pasted onto defect-free images in various combinations and patterns. The 3D images represented by each learning image data are images including welding defects and weld beads. Then, the learning data creation unit 223 stores the learning data, which consists of multiple data sets of the created learning image data and the welding defect information corresponding to the learning image data, in the learning image data storage unit 215.

[0054] Next, in (S110), the AI ​​model generation unit 225 generates a defect detection model by performing machine learning using the learning data stored in the learning image data storage unit 215. Then, the AI ​​model generation unit 225 stores parameters that specify the generated defect detection model in the parameter storage unit 216. The parameters stored in the parameter storage unit 216 are sent to the defective welding information calculation unit 372 of the welding device 3. The defective welding information calculation unit 372 stores the parameters.

[0055] In this way, because processor 22 creates defect-free image data in response to the base material designation input and the weld mark identification input, in order to acquire defect-free image data, it is not necessary to form a weld bead that does not include a welding defect by welding or to measure the shape of the weld bead using 3D sensor 36. This reduces the effort and work time required to acquire learning data.

[0056] Furthermore, because the processor 22 automatically creates segmentation data for defect-free image data, the user does not need to perform input work to identify the boundary between the weld mark formation area and the non-formation area while visually checking the defect-free image in order to create the segmentation data, which further reduces the effort and work time required to acquire learning data.

[0057] Furthermore, when the protrusion height of the weld bead formed by welding from the base material W is 0.5 mm or less, particularly when the protrusion height of the weld bead from the base material W is 0.3 mm or less, it is difficult to accurately measure the shape of the weld bead using the 3D sensor 36. This problem becomes particularly pronounced when the robot arm 34 vibrates. Even if the measurement by the 3D sensor 36 is accurate, a user viewing a 3D image based on the measurement results may be unable to recognize the boundary between the weld bead formation area and the weld bead non-formation area, making segmentation impossible. However, even in these cases, in this embodiment, the processor 22 can appropriately create defect-free image data and segmentation data of the defect-free image data.

[0058] The user then starts the power supply to the welding power source 32 of the welding device 3 and performs arc welding to form a weld bead on the base material W along with a weld defect. The 3D sensor 36 then captures an image of the welded portion PW, and the image processing unit 371 receives the shape data acquired by the 3D sensor 36. The image processing unit 371 then converts this shape data into image data of a welding image including the welded portion PW. The poor welding information calculation unit 372 then calculates (predicts) poor welding information based on the image data generated by the image processing unit 371 using a defect detection model identified by parameters sent from the AI ​​model generation device 2. The poor welding display image generation unit 373 calculates the pixel value of each pixel of the poor welding display image displaying the poor welding information based on the poor welding information calculated by the poor welding information calculation unit 372. The second output device 38 then displays the poor welding display image based on the pixel value of each pixel calculated by the poor welding display image generation unit 373.

[0059] In the above embodiment, the welding device 3 used to form the weld bead to obtain image data with defects in (S101) and the welding device 3 used to form the weld bead to obtain image data to be input into the defect detection model are a common welding device, but they may also be different welding devices.

[0060] In the above embodiment, the functions of the defect-containing data processing unit 221, the defect-free data processing unit 222, and the learning data creation unit 223, and the function of the AI ​​model generation unit 224 are realized by one processor 22, but they may be realized by different processors. In other words, the functions of the processor 22 in the above embodiment may be realized by multiple processors.

[0061] In the above embodiment, the present invention is applied to a welding system having a welding section that performs arc welding. However, the present invention can also be applied to welding systems in which the welding section performs welding without supplying a filler material such as a filler or wire, such as TIG welding, plasma welding, laser welding, or resistance welding. In the above embodiment, the weld beads included in the defect-free image are convex, but they may also be concave. In such a case, the defect-free data processing unit 222 can preferably generate defect-free image data that shows a defect-free image including a weld bead with a depth of 0.5 mm or less, and more preferably generate defect-free image data that shows a defect-free image including a weld bead with a depth of 0.3 mm or less. In the above embodiment, the defect-free data processing unit 222 generates defect-free image data that shows an image including a weld bead. However, the defect-free image data that shows an image including other weld marks, such as dents caused by a resistance welding welding gun, instead of a weld bead may be generated depending on the welding method.

[0062] Furthermore, in the above embodiment, the welding mark specifying input specifies the shape of the weld bead, but it may also specify the shape and size of the weld bead, or only the size of the weld bead.

[0063] The learning data acquisition device, welding system, and learning data acquisition method disclosed herein can reduce the effort and work time required to acquire learning data, and are useful as a learning data acquisition device, welding system, and learning data acquisition method for acquiring image data to be used in machine learning.

[0064] 2 AI model generation device (learning data acquisition device) 3 Welding device 22 Processor (learning data acquisition unit) 24 First input device (input unit) 31 Welding torch (welding portion) 32 Welding power source (welding portion) 33 Output control unit (welding portion) 36 3D sensor 37 Computer (defect information calculation unit)

Claims

1. A learning data acquisition device that acquires learning image data used for machine learning to generate a model, the model outputs welding defect information including the location of a welding defect based on image data of a 3D image, the learning data acquisition device comprising: an input unit that accepts a user's welding mark identification input specifying at least one of a convex or concave weld mark shape and a size; a welding mark image creation step that creates welding mark image data showing a 3D image including the weld mark in response to the welding mark identification input to the input unit; and a learning image creation step that creates learning image data showing a 3D image including the weld defect and the weld mark based on the welding mark image data created by the welding mark image creation step and defect-present image data showing a 3D image including the weld defect.

2. A learning data acquisition device as described in claim 1, characterized in that the learning data acquisition unit further executes a segmentation step of automatically creating segmentation data that identifies the boundary between the formation area and non-formation area of ​​the weld mark based on the weld mark image data.

3. The learning data acquisition device according to claim 2, wherein the height or depth of the weld mark in the 3D image shown by the weld mark image data is 0.5 mm or less.

4. The learning data acquisition device according to claim 3, wherein the height or depth of the welding mark in the 3D image shown by the welding mark image data is 0.3 mm or less.

5. A welding system comprising the learning data acquisition device according to any one of claims 1 to 4 and a welding device, wherein the learning data acquisition unit further executes a model generation step of generating the model by performing machine learning using a plurality of data sets of the learning image data and defective welding information corresponding to the learning image data, and the welding device has: a welding part where welding is performed; a 3D sensor that measures the shape of the welded part and acquires shape data; a step of acquiring measured image data showing a 3D image including the welded part based on the shape data acquired by the 3D sensor; and a defect information calculation unit that executes a step of calculating the defective welding information based on the measured image data using the model.

6. A welding system according to claim 5, characterized in that the welding section performs welding without supplying a filler metal.

7. A welding system according to claim 6, characterized in that the welding performed by the welding section is TIG welding, plasma welding, laser welding or resistance welding.

8. A learning data acquisition method for acquiring image data used in machine learning to generate a model, wherein the model outputs welding defect information including the location of a welding defect based on image data of a 3D image, and the learning data acquisition method is characterized by executing the following steps: an input step for accepting a user's welding mark identification input specifying at least one of the shape and size of a convex or concave welding mark; a welding mark image creation step for creating welding mark image data showing a 3D image including the welding mark in accordance with the welding mark identification input; and a learning image data creation step for creating learning image data showing a 3D image including the welding defect and the welding mark based on the welding mark image data created by the welding mark image creation step and defect-present image data showing a 3D image including the welding defect.