Method for detecting welding defects, method for evaluating welding defects, control device, welding system, and detection program

Abstract

Welding defects are detected based on the factors that cause them. [Solution] A welding defect detection method comprises: a first step of acquiring at least an image of the welding area during welding; a second step of extracting information relating to at least porosity defects from the welding defects based on the image; a third step of classifying the information relating to porosity defects into two or more types of porosity defect occurrences; and a fourth step of extracting at least one type of porosity defect occurrence and calculating quantitative data relating to the porosity defect.

Description

Technical Field

[0001] The present invention relates to a method for detecting welding defects, a method for evaluating welding defects, a control device, a welding system, and a detection program.

Background Art

[0002] During arc welding, welding defects may occur due to various factors. Welding defects affect the quality of the welded structure to be fabricated. Therefore, conventionally, in order to improve welding quality, it has been required to detect welding defects during welding in real time. The higher the accuracy of this detection, the more it can be utilized for improving welding quality, such as suppressing the occurrence of welding defects or traceability after welding.

[0003] Patent Document 1 describes a welding observation device. The welding observation device is a welding observation device that observes a molten portion in arc welding, and includes a first imaging unit that captures an image of a molten pool formed in the molten portion, and a first detection unit that detects bubbles generated in the molten pool from the image of the molten pool captured by the first imaging unit.

[0004] Patent Document 2 describes a welding observation device. The welding observation device is a welding observation device that observes a molten portion in arc welding, and includes a bubble detection unit that detects bubbles generated in the molten portion, a cumulative number calculation unit that calculates the cumulative number of bubbles detected by the bubble detection unit for each analysis section having a predetermined movement distance in arc welding, and a welding defect detection unit that detects the occurrence of welding defects based on the cumulative number for each analysis section.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

[0006] Welding defects can occur due to a variety of factors. These include porosity defects such as pits and blowholes. One cause of porosity defects is poor shielding gas, such as wind effects or nozzle blockage due to spatter. Other causes of porosity defects include those related to plating (e.g., galvanized steel sheets) and those caused by moisture or oil adhering to the steel sheet surface. Even for welding defects other than porosity defects, multiple contributing factors may exist.

[0007] To utilize the detection results of welding defects to suppress their occurrence, it is necessary to appropriately control the welding system based on the detection results. However, this control requires not only detection accuracy but also an understanding of the cause of the defect. Furthermore, when using the results for post-weld traceability, clarifying the cause of the welding defect allows for more detailed information to be obtained. However, conventional technologies do not consider means of detecting welding defects by identifying the cause.

[0008] The present invention has been made in view of the above-mentioned problems, and its objective is to provide a welding defect detection method, a welding defect evaluation method, a control device, a welding system, and a detection program that can detect welding defects for each factor causing welding defects. [Means for solving the problem]

[0009] The present invention consists of the following configuration.

[0010] (1) A method for detecting welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, The method is characterized by comprising a fourth step of extracting at least one type of pore defect occurrence and calculating quantitative data related to the pore defect, Methods for detecting welding defects.

[0011] (2) A method for evaluating welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, A fourth step involves extracting at least one type of pore defect and calculating quantitative data related to the pore defect, A fifth step in which, based on the aforementioned quantitative data, information on welding defects after welding, the amount of any element contained in the molten metal, or the amount of deposits on the workpiece surface are predicted, The invention is characterized by comprising a sixth step of evaluating the presence or absence of porosity defects based on the welding defect information and predicted values ​​of arbitrary elemental amounts or deposit amounts. Methods for evaluating welding defects.

[0012] (3) A control device for detecting welding defects, comprising a control unit, The control unit, during welding, At least obtain images of the welded area, Based on the aforementioned image, information relating to at least porosity defects among the welding defects is extracted. Based on the information relating to the aforementioned pore defects, classify them into two or more types of pore defect occurrences. The method is characterized by extracting at least one type of pore defect and calculating quantitative data related to the pore defect. Control device.

[0013] (4) A welding system for detecting welding defects, Equipped with a control device, During welding, the control device at least acquires an image of the welding location, based on the image, extracts information related to at least porosity defects among the welding defects, classifies the information related to the porosity defects into two or more types of occurrence types of porosity defects, extracts at least one type of occurrence type of the porosity defects, and calculates quantitative data related to the porosity defects, and is characterized in that welding system.

[0014] (5) A detection program for detecting welding defects, during welding, to the control device, at least has a function of acquiring an image of the welding location, based on the image, has a function of extracting information related to at least porosity defects among the welding defects, has a function of classifying the information related to the porosity defects into two or more types of occurrence types of porosity defects, has a function of extracting at least one type of occurrence type of the porosity defects and calculating quantitative data related to the porosity defects, and is characterized by realizing detection program.

[0015] (6) A method for detecting welding defects, during welding, a first step of at least acquiring an image of the welding location, based on the image, a third step of classifying the information related to the welding defects into two or more types of occurrence types of welding defects, a fourth step of extracting at least one type of occurrence type of the welding defects and calculating quantitative data related to the welding defects, and is characterized by having method for detecting welding defects. <000 103>[

Advantages of the Invention

[0016] According to the present invention, welding defects can be detected for each factor causing the occurrence of welding defects.

Brief Description of the Drawings

[0017] [Figure 1] Figure 1 is a schematic diagram showing an example of the configuration of the welding system according to this embodiment. [Figure 2] Figure 2 is a perspective view illustrating the arrangement of the imaging device according to this embodiment. [Figure 3] Figure 3 is a perspective view illustrating the arrangement of multiple imaging devices according to this embodiment. [Figure 4] Figure 4 is a block diagram showing an example configuration of a data processing device according to this embodiment. [Figure 5] Figure 5 illustrates the welding defect detection process according to this embodiment using a pre-trained model. [Figure 6] Figure 6 shows an example of the output of the bubble quantity according to this embodiment. [Figure 7] Figure 7 is a diagram illustrating the display screen provided by the display unit according to this embodiment. [Modes for carrying out the invention]

[0018] Hereinafter, a welding system according to one embodiment of the present invention will be described with reference to the drawings. In each drawing, the same components will be given the same reference numeral to indicate their correspondence. The welding system according to the present invention is suitable for a portable welding robot, but is not limited to the configuration of this embodiment. For example, it may be a 6-axis welding robot or an automatic welding device having a drive unit such as a trolley. In this embodiment, as an example, single-sided welding in a horizontal position is selected, but is not limited to this, and the construction method and welding position are not particularly limited.

[0019] Figure 1 is a schematic diagram showing an example of the configuration of a welding system according to this embodiment. The welding system 50 includes a portable welding robot 100, a feed device 300, a welding power supply 400, a shielding gas supply source 500, a robot control device 600, an imaging device 700, and a data processing device 800. As described above, when applying the features of this embodiment to an automatic welding device having a drive unit such as a 6-axis welding robot or a trolley, further configurations may be included to match those configurations. In addition, each part constituting the welding system 50 is connected in a communicative manner by various wired or wireless communication methods. The communication method here is not limited to one, and multiple communication methods may be combined for connection.

[0020] (Robot control device) The robot control device 600 is connected to the portable welding robot 100 by a robot control cable 610 and to the welding power supply 400 by a power supply control cable 620. The robot control device 600 has a data holding unit 601 that holds teaching data that defines in advance the operation pattern of the portable welding robot 100, welding start position, welding end position, construction conditions, welding conditions, parameter table (hereinafter also referred to as "condition DB"), etc. Based on this teaching data, it sends command information as commands to the portable welding robot 100 and the welding power supply 400 to control the operation of the portable welding robot 100 and the welding conditions.

[0021] Furthermore, the robot control device 600 may include a groove shape information calculation unit 602 that calculates groove shape information from detection data obtained by performing sensing such as touch sensing before welding, and a welding condition acquisition unit 603 that performs stack design based on at least one of the groove shape information and the condition DB, and sets or corrects the welding conditions of the teaching data for each pass. The groove shape information calculation unit 602 and the welding condition acquisition unit 603 constitute the control unit 604. The control unit 604 is configured using, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a DSP (Digital Signal Processor), or an FPGA (Field Programmable Gate Array). In this embodiment, welding condition setting using sensing, which is most preferable from the viewpoint of efficiency, is applied, but it is not limited to this, and for example, welding condition setting may be performed by direct teaching. Furthermore, in this invention, the sensing position or teaching position (hereinafter also referred to as the "teaching position" or "teaching point") of the first layer (hereinafter also referred to as the "first layer") may be any position on the welding start side. Note that any position on the welding start side refers to the welding start position or the vicinity of the welding start position. In this embodiment, the welding start position is used from the viewpoint of accuracy of groove shape information, but the welding conditions at the welding start position may be set based on groove shape information in the vicinity of the welding start position.

[0022] Lamination design refers to determining the appropriate number of layers and passes for a given groove. A layer refers to the number of times the weld bead is layered in the thickness direction of the plate. In lamination design for multi-layer welding, it is generally assumed that the layer height per layer is kept constant. Therefore, it is preferable to perform lamination design based on groove shape information and set or correct the welding conditions based on the calculated lamination design information. Here, lamination design information includes the number of layers, the number of passes, and the layer height (≒ height of one pass), but the welding conditions are determined based on at least the calculated layer height value. For example, if sensing reveals that the gap is larger at the end of welding than at the start of welding, the gap width will widen as welding progresses, so it is necessary to increase the amount of weld material to keep the layer height constant. In this case, welding conditions related to the amount of weld material include wire feeding speed and welding speed, and these welding conditions should be set or corrected based on groove shape information and lamination design information.

[0023] The welding power supply 400, under the command of the robot control device 600, supplies power to the welding wire 211 and the workpiece Wo, which are consumable electrodes, thereby generating an arc between the welding wire 211 and the workpiece Wo. Power from the welding power supply 400 is sent to the feeder 300 via the power cable 410, and from the feeder 300 to the welding torch (hereinafter referred to as "torch") 200 via the conduit tube 420. Power from the welding power supply 400 is then supplied to the welding wire 211 via the contact tip at the tip of the torch 200. The current during welding may be DC or AC, and its waveform is not particularly specified. Therefore, the current may be a pulse such as a square wave or a triangular wave.

[0024] Furthermore, the welding power supply 400 is configured such that, for example, power cable 410 is connected to the torch 200 as the positive electrode, and power cable 430 is connected to the workpiece Wo as the negative electrode. Note that this is the case when welding with reverse polarity; when welding with positive polarity, the power cable for the positive electrode should be connected to the workpiece Wo, and the power cable for the negative electrode should be connected to the torch 200.

[0025] (Shielding gas supply source) The shielding gas supply source 500 consists of a container filled with shielding gas, ancillary components such as valves, etc. Shielding gas is supplied from the shielding gas supply source 500 to the supply device 300 via a gas tube 510. The shielding gas supplied to the supply device 300 is then supplied to the torch 200 via a conduit tube 420. The shielding gas supplied to the torch 200 flows through the torch 200, is guided to the nozzle 210, and is ejected from the tip of the torch 200. Examples of shielding gases used in this embodiment include argon (Ar), carbon dioxide (CO2), or a mixture thereof.

[0026] In this embodiment, the conduit tube 420 has a conductive path formed on the outer sheath side of the tube for functioning as a power cable, a protective tube for protecting the welding wire 211 is arranged inside the tube, and a flow path for shielding gas is formed. However, the conduit tube 420 is not limited to this, and for example, a bundle of power supply cables and shielding gas supply hoses can be used with a protective tube for supplying the welding wire 211 to the torch 200 at its center. Alternatively, for example, the tube for supplying the welding wire 211 and shielding gas and the power cable can be installed separately.

[0027] (Feeding device) The feeding device 300 feeds the welding wire 211 to the torch 200. The welding wire 211 fed by the feeding device 300 is not particularly limited and is selected according to the properties of the workpiece Wo and the welding method, for example, solid wire or flux-cored wire can be used. The wire diameter of the welding wire is not particularly limited, but in this embodiment, the preferred wire diameter is 1.6 mm at the upper limit and 0.9 mm at the lower limit.

[0028] Furthermore, in this embodiment, a touch sensor is used as the detection means to sense the surface of the groove, etc., by applying a voltage between the workpiece Wo and the welding wire 211 and utilizing the voltage drop phenomenon that occurs when the welding wire 211 comes into contact with the workpiece Wo. The detection means is not limited to the touch sensor of this embodiment, and an image sensor, laser sensor, etc., or a combination of these detection means may be used, but it is preferable to use the touch sensor of this embodiment due to the simplicity of the device configuration.

[0029] (Imaging device) The imaging device 700 (hereinafter also referred to as the "visual sensor" or "camera") is composed of, for example, a camera equipped with a CMOS (Complementary Metal-Oxide-Semiconductor) as a visual sensor. The imaging device 700 may be directly attached to the portable welding robot 100, or it may be fixed to a specific location in the surrounding area. The imaging device 701, which will be described later with reference to Figure 3, may have a similar configuration to the imaging device 700.

[0030] Image information captured by the imaging device 700 and the imaging device 701 (described later) is transmitted to the data processing device 800 and used by the data processing device 800. At this time, the data processing device 800 may, for example, select arbitrary images from the captured image information at predetermined intervals and use them for processing described later. The selection method and selection settings here may be switched according to, for example, the configuration and functions of the imaging device and the performance of the data processing device 800.

[0031] In this embodiment, an imaging device 700 is directly attached to and fixed to a portable welding robot 100, and a moving image is captured as a welding image, such that the imaging range includes at least the welding area containing the molten pool as an object (target) to be included in the image data. Here, "welding area" refers to any location on the welding line. In addition, a moving image may be captured as a welding image, such that the imaging range further includes the workpiece Wo, welding wire 211, and arc.

[0032] Various imaging settings related to welding images may be predetermined or switched according to the operating conditions of the welding system 50. Examples of imaging settings include frame rate, number of pixels in the image, resolution, and shutter speed.

[0033] Figure 2 is a perspective view illustrating the arrangement of the imaging device 700 according to this embodiment. Note that the orientation of the torch 200 and the groove differs depending on the welding position, so the orientation shown in Figure 2 is just one example. In this embodiment, the workpiece Wo is a butt joint. The workpiece Wo consists of two metal plates that are butted together with a groove in between. In this embodiment, a horizontal orientation is given as an example, in which case the upper plate workpiece will be referred to as W1 (hereinafter referred to as "upper plate W1") and the lower plate workpiece as W2 (hereinafter referred to as "lower plate W2"). ​​A ceramic backing material 14 is attached to the back side of the two butted workpieces W1 and W2. Note that a metal backing material may be used as the backing material 14, or a configuration without a backing material may be used. Therefore, the material of the backing material is not particularly limited and may differ depending on the material of the workpiece Wo, etc. In a butt joint, arc welding is performed in one direction along the groove. In the following, the direction in which welding progresses is referred to as the "weld line direction." In Figure 2, the direction in which welding progresses is indicated by an arrow. For this reason, the imaging device 700 is located behind the torch 200. The position of the imaging device 700 is not limited to behind the torch 200, but when detecting porosity defects, as described later, placing the imaging device 700 behind the torch 200 can improve the accuracy of porosity defect detection.

[0034] Figure 3 is a perspective view illustrating the arrangement of the multiple imaging devices 700 and 701 according to this embodiment. Parts similar to those shown in the embodiment in Figure 2 will not be described. In Figure 3, in addition to imaging device 700, imaging device 701 is further arranged. Imaging device 701 is located in front of the torch 200. Imaging device 701, positioned in front of the torch 200, is mainly used for welding speed control, and if it is to be used in conjunction with defect detection according to the present invention, it should be arranged as shown in Figure 3. Imaging device 700, positioned behind the torch 200, is mainly used for detecting welding defects according to the embodiment of this disclosure.

[0035] The visual sensors of the imaging device 700 and imaging device 701 in this embodiment can continuously capture still images of, for example, 1024 × 768 pixels. In other words, the imaging device 700 and imaging device 701 can capture welding images as moving images. The resolution of the still images that can be captured by the imaging device 700 and imaging device 701 is not particularly limited. For example, the imaging device 700 and imaging device 701 may each acquire welding images with different resolutions. Furthermore, before inputting into the trained model described later, preprocessing such as extracting an arbitrary feature region from the captured welding image may be performed for the purpose of reducing processing time. The arbitrary feature region may be a fixed-size range arranged so that a predetermined area is located in the center. Also, the size of the arbitrary feature region may be changed according to the welding conditions.

[0036] (Data processing device) Figure 4 is a block diagram showing an example configuration of the data processing device 800 according to this embodiment. The data processing device 800 is a control device, and is composed of, for example, a computer. The computer is composed of a main unit 810, an input unit 820, and a display unit 830. The main unit 810 is composed of a CPU 811, a GPU (Graphical Processing Unit) 812, a ROM 813, a RAM 814, a non-volatile storage device 815, an input / output interface 816, a communication interface 817, a video output interface 818, and a calculation unit 819. The CPU 811, GPU 812, ROM 813, RAM 814, non-volatile storage device 815, input / output interface 816, communication interface 817, video output interface 818, and calculation unit 819 are connected to each other so as to be able to communicate with each other by buses or signal lines.

[0037] The non-volatile storage device 815 stores a learning program 815A that performs machine learning using predetermined training data, a trained model 815B generated through the execution of the learning program 815A, a quantitative data generation program 815C that generates quantitative data described later using the trained model 815B, and image data 815D. In addition to these, the non-volatile storage device 815 also has an operating system and application programs installed.

[0038] The data processing unit 800 implements various functions through the execution of programs by the CPU 811 and GPU 812. In this embodiment, the data processing unit 800 implements the function of generating a trained model by machine learning and the function of performing various processing using the trained model. The details of these functions will be described later. Alternatively, the data processing unit 800 may be divided into two parts: one for generating a trained model and another for performing control processing based on information output from the trained model during actual welding. From the standpoint of versatility, it is more preferable to divide the data processing unit 800 according to each function. The GPU 812 is used as the arithmetic unit when executing the learning program 815A and the quantitative data generation program 815C. The ROM 813 stores the BIOS (Basic Input Output System) and other information that is executed by the CPU 811. The RAM 814 is used as the working area for programs read from the non-volatile storage device 815.

[0039] The input / output interface 816 is connected to the input unit 820, which consists of a keyboard, mouse, etc. The imaging device 700 and imaging device 701, which are vision sensors, are also connected to the input / output interface 816. Image data output from imaging devices 700 and 701 is provided to the CPU 811 via the input / output interface 816. The communication interface 817 is a communication module for wired or wireless communication. The video output interface 818 is connected to the display unit 830, which consists of, for example, a liquid crystal display or an organic EL (Electro-Luminescence) display, and outputs a video signal to the display unit 830 according to the video data provided by the CPU 811. The calculation unit 819 works in cooperation with the CPU 811 and GPU 812 to perform various processes, such as calculating quantitative data related to bubble defects according to this embodiment and calculating geometric quantities used to control the welding speed.

[0040] (Generating a pre-trained model) The trained model 815B in this embodiment is composed of a convolutional neural network and includes multiple convolutional layers and multiple pooling layers. However, the configuration of the convolutional neural network is not limited to the above, and the number of layers and configuration may be different.

[0041] The trained model 815B takes information based on welding images output from the imaging device 700 and imaging device 701 as input data. As will be described later, the input data obtained from imaging device 700 outputs information from which quantitative data related to porosity defects can be derived for each type (hereinafter also referred to as "occurrence type") that is separated according to the cause of porosity defect occurrence. The input data obtained from imaging device 701 outputs information from which geometric quantities used for controlling the welding speed can be derived. Hereinafter, this embodiment relating to imaging device 700 will be described. The original information input to the trained model 815B is an image, in which at least the molten pool is reflected as an object, i.e., the target. The image may also reflect welding wire, arc, and other objects. By inputting information based on imaging information, the trained model 815B outputs information from which quantitative data related to porosity defects for each occurrence type can be derived. The information based on imaging information may be the image itself, or it may be information obtained by applying some kind of information processing, such as preprocessing, to the image. The quantitative data information for each type of pore defect may be the quantitative data itself. The information from which quantitative data related to each type of pore defect can be derived may be information from which quantitative data can be derived by applying some kind of information processing to that information.

[0042] The data processing device 800 calculates quantitative data of any type of pore defect by inputting information based on the images captured from the imaging device 700 into a trained model 815B.

[0043] (Detection of welding defects) The welding system 50 of this disclosure detects welding defects. Welding defects include, for example, pits and blowholes. The imaging device 700 captures images of the weld area in real time. The data processing device 800 acquires the captured images of the weld area. In this specification, the captured images of the weld area may be abbreviated as welding images or captured images. The weld area includes at least a molten pool. The weld area may further include areas other than the molten pool, and may more preferably include welding wire and arc.

[0044] The data processing device 800 extracts information related to welding defects based on the welding image. At this time, since information on porosity defects such as pits and blowholes is of high importance for welding, it is advisable to extract at least information related to porosity defects from among the welding defects. Information related to porosity defects includes information on bubbles that occur on the molten pool. Information related to porosity defects may also include information indicating porosity defects other than bubbles.

[0045] In this embodiment, the data processing device 800 classifies information related to porosity defects into two or more types of porosity defects. The type of porosity defect can be selected from, for example, at least one of three types: porosity defects due to poor shielding gas, porosity defects originating from plating such as galvanized steel sheets, and porosity defects caused by moisture or oil adhering to the surface of the steel sheet. For example, in this embodiment, the types of porosity defects are classified into two types: porosity defects due to poor shielding gas and porosity defects due to other reasons (porosity defects resulting from all causes other than porosity defects due to poor shielding gas). Note that other types of porosity defects besides the three types described above may also be provided.

[0046] Pore ​​defect types may be classified based on at least one of the following: the location of occurrence on the molten pool and the brightness of the bubbles. The location of occurrence on the molten pool may be classified into ranges based on the brightness of the molten pool, and the bubbles that occur in each range may be classified as different pore defect types.

[0047] The data processing device 800 extracts at least one type of porosity defect and calculates quantitative data related to the porosity defect. Quantitative data refers to information that can be counted and expressed numerically. Quantitative data related to porosity defects includes, for example, the amount of bubbles, as described later. Furthermore, if shielding gas defects are present, the amount of sputtering also changes, and therefore the amount of sputtering can also be included as quantitative data related to porosity defects. Here, the amount of bubbles includes the area of ​​the bubbles, the ratio of the bubble area to the molten pool area, and the frequency. Other quantitative data related to porosity defects may include other values ​​that correlate with various values ​​such as the amount of bubbles or the amount of sputtering.

[0048] The value relating to the amount of bubbles may be calculated based on at least one of the following: number, area, and time of occurrence. For example, the value relating to the amount of bubbles may be the number of bubbles, the total area occupied by bubbles, or the time of time of bubble occurrence. The value relating to the amount of bubbles may also be a combination of two or more of these values. For example, a value obtained by multiplying the area occupied by bubbles by the frequency of bubble occurrence may be a value relating to the amount of bubbles. The value relating to the amount of bubbles may be calculated for each bead position or time.

[0049] The welding system 50 may implement the above-described welding defect detection process using a trained model. Figure 5 is a diagram illustrating the welding defect detection process according to this embodiment using a trained model.

[0050] The welding system 50 calculates quantitative data related to porosity defects of any given type by inputting image-based information into a learning device that includes a trained model 815B, which takes image-based information as input and outputs information that can derive quantitative data related to porosity defects of each occurrence type. In this example, the learning device corresponds to the main unit 810 included in the data processing device 800 shown in Figure 4.

[0051] The processing example shown in Figure 5 extracts information related to porosity defects, including bubble information that occurs on the molten pool, among welding defects. In this example, there are two types of porosity defects that are classified based on the brightness of the bubbles: white bubbles with brightness above a predetermined threshold, and black bubbles with brightness below a predetermined threshold. White bubbles correspond to porosity defects due to poor shielding gas. Black bubbles correspond to porosity defects due to factors other than poor shielding gas. The quantitative data in this example is the amount of bubbles, which is the total number of pixels of bubbles captured in the image.

[0052] Furthermore, slag is also present in the weld bead, so it is important whether or not the bubbles captured in the image can be successfully extracted. Here, bubbles burst and disappear immediately after they are formed, while non-bubble substances such as slag do not disappear immediately. Therefore, it is acceptable to distinguish between bubbles and slag from the captured image based on the changes over time in the welding image captured as a video, and then extract the bubbles.

[0053] The imaging device 700 captures welding images, which are images of the welded area, in real time. The data processing device 800 acquires the captured welding images (St101).

[0054] The data processing device 800 inputs the welding image to the learning device (main unit 810) (St102). The data processing device 800 calculates and outputs the amount of white bubbles and the amount of black bubbles from the image of the welding area (St103).

[0055] The imaging device 700 and the data processing device 800 repeat the process in steps St101 to St103 until welding is completed.

[0056] Figure 6 shows an example of bubble quantity output according to this embodiment. As described above, the data processing device 800 outputs the bubble quantity for each bead position. In Figure 6, the horizontal axis shows the bead position, and the vertical axis shows the bubble quantity. As shown in the figure, the bubble quantity is output for each type of porosity defect.

[0057] (Evaluation of welding defects based on detection results) As described above, after quantitative data relating to porosity defects of any type has been calculated, the data processing device 800 predicts the information to be predicted based on the quantitative data. The information to be predicted may be, for example, the presence or absence of welding defects during welding, welding defect information indicating the content of welding defects after welding, the amount of any element contained in the molten metal, or the amount of deposits on the workpiece surface. The amount of any element may be, for example, the amount of nitrogen, but may also be the amount of other elements. The deposits on the workpiece surface may include plating deposits. The data processing device 800 evaluates the presence or absence of welding defects based on the predicted values ​​of the information to be predicted. The data processing device 800 may evaluate the presence or absence of porosity defects based on the predicted values ​​of the information to be predicted.

[0058] For example, when determining the presence or absence of welding defects during welding as prediction information, a threshold can be set in advance for the quantity of bubbles, which is quantitative data, and the presence or absence of welding defects can be determined based on that threshold. Also, when predicting the amount of nitrogen contained in the molten metal as prediction information, the correlation between the quantity of bubbles, which is quantitative data for evaluating welding defects, and the amount of nitrogen in the weld metal can be analyzed in advance, and a relational equation can be constructed. For example, this relational equation can be stored in the data processing device 800.

[0059] The data processing device 800 calculates the amount of nitrogen in the weld metal in real time based on the amount of bubbles output as described above and the stored relational expression. Based on the calculated amount of nitrogen, the data processing device 800 predicts in real time whether welding defects will occur. For example, if there is a positive correlation such as an increase in the amount of nitrogen in the weld metal leading to an increase in the amount of bubbles, the data processing device 800 stores a relational expression that outputs a predicted value for the amount of nitrogen from the input amount of bubbles. Here, for example, if the mechanical properties deteriorate when the amount of nitrogen in the weld metal is within a predetermined first range, and blowout occurs when the amount of nitrogen in the weld metal is within a predetermined second range, the data processing device 800 predicts the amount of nitrogen based on the amount of bubbles output as described above and the aforementioned relational expression, and evaluates whether welding defects are present based on the predicted value of the amount of nitrogen.

[0060] Furthermore, if deterioration of mechanical properties or porosity defects are detected, additional control measures may be taken. For example, the operation of various devices included in the welding system 50 may be stopped, or the setting values ​​for the operation of the various devices may be changed.

[0061] (Example display) Figure 7 is an example of the display screen provided by the display unit according to this embodiment. The display screen provided by the display unit 830 shows the bubble region extraction result D1, the color map D2, the time series data D3, and the control box D4.

[0062] As the bubble region extraction result D1, the welding image captured by the imaging device 700 is displayed as a video.

[0063] Colormap D2 is a colormap that classifies and displays the numerical range of quantitative data into arbitrary colors. In this example, a bubble amount colormap is displayed that classifies the numerical range of bubble amounts for each bead position into arbitrary colors.

[0064] Time-series data D3 is a line graph showing quantitative data for each bead position or for each count of the captured image taken by the imaging device 700. In this example, a line graph showing the amount of bubbles for each count of the captured image is displayed.

[0065] Control box D4 includes buttons for playing, pausing, or stopping the video displayed in bubble region extraction result D1.

[0066] (modified version) In the above embodiment, welding images are captured by the imaging device 700, and welding defects are detected using the welding images. As a modified example, welding sounds may be captured using a microphone, for example, and welding defects may be detected by combining the audio data and the welding images.

[0067] Furthermore, as explained in Figure 3, multiple imaging devices may be provided, and welding defects may be detected from images obtained by imaging device 700, and welding conditions may be controlled from images obtained by imaging device 701. The trained model 815B may be equipped with a trained model that, when a welding image obtained by imaging device 701 is input, outputs feature points related to welding or geometric quantities obtained from feature points. Here, feature points related to welding include the tip of the welding wire, the center point of the arc, the positions of the left and right (however, up and down in the case of a sideways orientation) tips of the molten pool with respect to the welding direction, and the positions of the left and right (however, up and down in the case of a sideways orientation) ends of the molten pool with respect to the welding direction. Examples of geometric quantities obtained from feature points include the distance between the welding wire and the tip of the molten pool in the welding direction, the width of the molten pool, or the area of ​​the molten pool. Based on such geometric quantity data, welding conditions such as welding speed may be controlled.

[0068] The robot control device 600 and the data processing device 800 may be integrated. Within the bounds of consistency, the data processing device 800 may perform the processing that the robot control device 600 is supposed to perform, and the robot control device 600 may perform the processing that the data processing device 800 is supposed to perform. Other devices may perform the processing that the robot control device 600 or the data processing device 800 is supposed to perform.

[0069] The present invention is not limited to the embodiments described above. It is also intended and within the scope of protection to be provided for the combination of each configuration of the embodiments, as well as for modifications and applications by those skilled in the art based on the description in the specification and well-known art.

[0070] As described above, the following matters are disclosed in this specification:

[0071] (1) A method for detecting welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, The method is characterized by comprising a fourth step of extracting at least one type of pore defect occurrence and calculating quantitative data related to the pore defect, Methods for detecting welding defects.

[0072] (2) In the second step, the information relating to the pore defects includes information about bubbles generated on the molten pool, In the third step described above, the occurrence of two or more types of pore defects is at least two or more types of bubble formation. The detection method according to (1), characterized in that, in the fourth step, for at least one of the bubble generation types, the quantitative data relating to the pore defect is a value relating to the amount of bubbles.

[0073] (3) The detection method according to (2), characterized in that the value relating to the amount of bubbles is a value calculated based on at least one of the number, area, and generation time.

[0074] (4) The detection method according to (2), characterized in that the type of porosity defect is classified based on at least one of the location of occurrence on the molten pool and the brightness of the bubbles.

[0075] (5) The generation location on the melt pool is classified into ranges based on the brightness of the melt pool, The detection method according to (4), characterized in that bubbles generated in each range are classified as the respective types of pore defects.

[0076] (6) The detection method according to (1), characterized in that it includes the step of inputting the image-based information into a learning device which includes a trained model that takes image-based information as input and outputs information capable of deriving quantitative data of the occurrence type of porosity defects, thereby calculating quantitative data of an arbitrary occurrence type of porosity defect.

[0077] (7) A method for evaluating welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, A fourth step involves extracting at least one type of pore defect and calculating quantitative data related to the pore defect, A fifth step in which, based on the aforementioned quantitative data, information on welding defects after welding, the amount of any element contained in the molten metal, or the amount of deposits on the workpiece surface are predicted, The invention is characterized by comprising a sixth step of evaluating the presence or absence of porosity defects based on the welding defect information and predicted values ​​of arbitrary elemental amounts or deposit amounts. Methods for evaluating welding defects.

[0078] (8) A control device for detecting welding defects, comprising a control unit, The control unit, during welding, At least obtain images of the welded area, Based on the aforementioned image, information relating to at least porosity defects among the welding defects is extracted. Based on the information relating to the aforementioned pore defects, classify them into two or more types of pore defect occurrences. The method is characterized by extracting at least one type of pore defect and calculating quantitative data related to the pore defect. Control device.

[0079] (9) Further comprising a display unit, The control device according to (8), wherein the display unit is characterized in that it classifies and displays the numerical range of the quantitative data into arbitrary colors.

[0080] (10) A welding system for detecting welding defects, Equipped with a control device, The control device during welding, At least obtain images of the welded area, Based on the aforementioned image, information relating to at least porosity defects among the welding defects is extracted. Based on the information relating to the aforementioned pore defects, classify them into two or more types of pore defect occurrences. The method is characterized by extracting at least one type of pore defect and calculating quantitative data related to the pore defect. Welding system.

[0081] (11) Further equipped with a display unit, The welding system according to (10), wherein the display unit is characterized in that it classifies and displays the numerical range of the quantitative data into arbitrary colors.

[0082] (12) A detection program for detecting welding defects, During welding, the control device A function to at least acquire images of the welded area, Based on the aforementioned image, a function is provided to extract information relating to at least porosity defects among the welding defects, The function of classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, A function to extract at least one type of pore defect and to calculate quantitative data related to the pore defect, Features that realize Detection program.

[0083] (13) A method for detecting welding defects, During welding, The first step is to obtain at least an image of the welded area, Based on the aforementioned image, A third step involves classifying the information relating to the welding defects into two or more types of welding defects, The method is characterized by comprising a fourth step of extracting at least one type of welding defect and calculating quantitative data related to the welding defect, Methods for detecting welding defects. [Explanation of Symbols]

[0084] 50 Welding Systems 100 Portable Welding Robots 200 Torches 210 nozzles 211 Welding wire 300 Feeding device 400 Welding Power Supply 410 Power Cable 420 Conduit Tube 430 Power Cable 500 Shielding gas supply source 510 Gas Tube 600 Robot Control Device 601 Data storage unit 602 Bevel shape information calculation unit 603 Welding Condition Acquisition Section 604 Control Unit 610 Robot Control Cable 620 Power control cable 700, 701 Imaging device 800 Data Processing Devices 810 Main Unit 815 Non-volatile memory 815A Learning Program 815B Pre-trained Model 815C Quantitative Data Generation Program 815D Image Data 816 Input / Output Interfaces 817 Communication Interface 818 Video output interface 819 Calculation Department 820 Input Section 830 Display section

Claims

1. A method for detecting welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, The method is characterized by comprising a fourth step of extracting at least one type of pore defect occurrence and calculating quantitative data related to the pore defect, Methods for detecting welding defects.

2. In the second step described above, the information relating to the pore defects includes information about bubbles generated on the molten pool, In the third step described above, the two or more types of pore defects are at least two or more types of bubble formation. The detection method according to claim 1, characterized in that, in the fourth step, for at least one of the bubble generation types, the quantitative data relating to the pore defect is a value relating to the amount of bubbles.

3. The detection method according to claim 2, characterized in that the value relating to the amount of bubbles is a value calculated based on at least one of the number, area, and generation time.

4. The detection method according to claim 2, characterized in that the type of porosity defect is classified based on at least one of the location of occurrence on the molten pool and the brightness of the bubbles.

5. The generation location on the melt pool is classified into ranges based on the brightness of the melt pool. The detection method according to claim 4, characterized in that bubbles generated in each range are classified as corresponding types of pore defects.

6. The detection method according to claim 1, characterized in that it includes the step of inputting the image-based information into a learning device that includes a trained model that takes image-based information as input and outputs information capable of deriving quantitative data of the occurrence type of porosity defects, thereby calculating quantitative data of an arbitrary occurrence type of porosity defect.

7. A method for evaluating welding defects, During welding, The first step is to obtain at least an image of the welded area, A second step involves extracting information related to at least porosity defects from the welding defects based on the aforementioned image, A third step involves classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, A fourth step involves extracting at least one type of pore defect and calculating quantitative data related to the pore defect, A fifth step in which, based on the aforementioned quantitative data, information on welding defects after welding, the amount of any element contained in the molten metal, or the amount of deposits on the workpiece surface are predicted, The invention is characterized by comprising a sixth step of evaluating the presence or absence of porosity defects based on the welding defect information and predicted values ​​of arbitrary elemental amounts or deposit amounts. Methods for evaluating welding defects.

8. A control device for detecting welding defects, comprising a control unit, The control unit, during welding, At least obtain images of the welded area, Based on the aforementioned image, information relating to at least porosity defects among the welding defects is extracted. Based on the information relating to the aforementioned porosity defects, classify them into two or more types of porosity defect occurrences. The method is characterized by extracting at least one type of pore defect and calculating quantitative data related to the pore defect. Control device.

9. It also includes a display unit, The control device according to claim 8, wherein the display unit is characterized in that it classifies and displays the numerical range of the quantitative data into arbitrary colors.

10. A welding system for detecting welding defects, Equipped with a control device, The control device during welding, At least obtain images of the welded area, Based on the aforementioned image, information relating to at least porosity defects among the welding defects is extracted. Based on the information relating to the aforementioned porosity defects, classify them into two or more types of porosity defect occurrences. The method is characterized by extracting at least one type of pore defect and calculating quantitative data related to the pore defect. Welding system.

11. It also includes a display unit, The welding system according to claim 10, wherein the display unit is characterized in that it classifies and displays the numerical range of the quantitative data into arbitrary colors.

12. A detection program for detecting welding defects, During welding, the control device A function to at least acquire images of the welded area, Based on the aforementioned image, a function is provided to extract information relating to at least porosity defects among the welding defects, The function of classifying the information relating to the aforementioned porosity defects into two or more types of porosity defect occurrences, A function to extract at least one type of pore defect and to calculate quantitative data related to the pore defect, Features that realize Detection program.

13. A method for detecting welding defects, During welding, The first step is to obtain at least an image of the welded area, Based on the aforementioned image, A third step involves classifying the information relating to the welding defects into two or more types of welding defects, The method is characterized by comprising a fourth step of extracting at least one type of welding defect and calculating quantitative data related to the welding defect, Methods for detecting welding defects.

Images