JP2004074264A - Weld visualizing device, and welding control unit and welding method using the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a weld visualizing device for visualizing a welding condition in a vicinity of a molten pool during the welding with high resolution, and a welding control unit and a welding control method for analyzing the welding condition by using an image from the weld visualizing device on the real time basis, and controlling the welding according to the result of analysis. <P>SOLUTION: The weld visualizing device 10 comprises a high pixel CCD camera 2 having an electronic shutter and a high brightness flash lamp 4 and synchronizes the illumination of a weld by the flash lamp with the image pickup by the CCD camera. The welding control unit 20 comprises an image take-in unit 12 to take in image data outputted by the weld visualizing device 10, an image processor 14 to process image data and determine the welding condition, and an output device 16 to output the welding condition. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention is a welding control device that analyzes a welding situation in real time using a weld visualization device and an image obtained thereby, and controls a welding position and a welding power according to the analysis result to improve welding quality. About the method.
[0002]
[Prior art]
As techniques for monitoring a welding state during welding, Japanese Patent Application Laid-Open No. 9-225666, "Laser Welding Monitoring Device", Japanese Patent Application Laid-Open No. 2000-196923, "Imaging Device for Light Emitting Element Using CCD Camera and Laser Lighting", 2000-246441, a "welding status remote monitoring device" and the like have been proposed.
Further, as an automatic welding device, Japanese Patent Laid-Open No. 2000-210773 has proposed "A welding position automatic copying control device and a method for welding a groove of a member to be welded".
[0003]
[Problems to be solved by the invention]
The conventional welding status monitoring technology described above can visually monitor the welded portion during welding, but the image information is small, the accuracy of estimating the welding status is low, and automation using this is difficult. . Therefore, the conventional welding automatic control device utilizes the above-described conventional welding condition monitoring technology and other contact-type sensors, but has a problem that the accuracy is low and the device is complicated and the versatility is low. .
[0004]
In other words, the conventional welding condition monitoring technology and welding automatic control device have a problem that the image of the vicinity of the molten pool, which originally has the most information, cannot be sufficiently utilized, and a highly accurate welding condition analysis cannot be performed. Was.
[0005]
The present invention has been made to solve the above problems. That is, an object of the present invention is to provide a welding part visualization device capable of visualizing a welding state in the vicinity of a molten pool at a high resolution, which could not be seen conventionally during welding, and a welding state in real time using an image obtained thereby. It is an object of the present invention to provide a welding control apparatus and method for analyzing and controlling a welding position and a welding power in accordance with the analysis result to improve welding quality.
[0006]
[Means for Solving the Problems]
According to the present invention, a high-pixel CCD camera (2) having an electronic shutter and a high-brightness flash lamp (4) are provided, and illumination of a welding portion by the high-brightness flash lamp and photographing by the high-pixel CCD camera are synchronized. A weld visualization device is provided.
[0007]
With the welding part visualization device of this configuration, the welding part can be photographed as a high-contrast high-resolution image, and the welding situation near the molten pool, which could not be seen at the time of welding, can be visualized with high resolution. .
[0008]
According to a preferred embodiment of the present invention, the CCD camera (2) continuously captures images at intervals exceeding a transfer time Δt required for transferring one frame of images to the memory table and outputs image data. I do.
With this configuration, it is possible to photograph the welded portion at regular intervals without omission and output image data.
[0009]
In the butt welding or the lap welding, an optical system (6) for guiding an image of a weld portion including a molten pool to a CCD camera (2) is provided.
With this configuration, it is possible to photograph the weld including the molten pool and the periphery thereof with high contrast, and to keep the photographing range within the depth of field.
[0010]
Further, according to the present invention, there is provided an image capturing device (12) for capturing image data output from the weld visualization device (10), and an image processing device (14) for performing image processing on the image data to determine a welding state. And an output device (16) for outputting the welding status.
[0011]
According to the welding control device having this configuration, the image capturing device (12) captures the image data output by the welding portion visualization device (10), and the image processing device (14) processes the image data to perform image processing. Is determined, and the welding status is output by the output device (16). Therefore, the welding position and the welding power can be controlled in accordance with the analysis result of the welding situation to improve the quality of welding.
[0012]
According to a preferred embodiment of the present invention, a PC for image processing and a frame grabber board for capturing image data are provided.
With this configuration, image data can be captured by a frame grabber board and image processing can be performed by a PC (personal computer).
[0013]
A plurality of memory tables are provided, one frame of image is sequentially transferred to each memory table, and all images are taken in real time as a whole.
With this configuration, even if it takes time to transfer data to a single memory table, all images can be captured in real time.
[0014]
Further, according to the present invention, the molten pool, groove, and bead are photographed as high-resolution image data, and from this image data, the groove position, groove width, molten pool position, molten pool state, bead position, or bead. A welding control method is provided, wherein at least one of the states is detected, and a welding state is analyzed in real time from the detected content.
[0015]
According to a preferred embodiment of the present invention, the analysis result is fed back to the welding device in real time, and the welding device is automatically controlled.
In addition, at least one of groove profiling, groove width detection, blow hole detection, and welding condition abnormality determination is analyzed in real time.
[0016]
In the groove tracing, the image is analyzed to determine the straight line vector of the groove, the current weld pool position, and the future weld pool position. The welding head is moved to improve the quality of welding when the positional deviation is predicted.
[0017]
In the groove width detection, the groove width is detected, and when the groove is opened, the welding head is moved to one edge to prevent welding failure.
In the blow hole detection, the state of the bead is determined, and if a blow appears, the welding power is adjusted.
[0018]
In the welding condition abnormality determination, a normal welding condition is grasped, and the normal welding condition is always compared with the real-time welding condition to thereby grasp the welding trouble and notify the trouble.
[0019]
According to the welding control method of the present invention, a groove position, a groove width, a weld pool position, a weld pool state, a bead position or a bead state are detected from high-resolution image data including a weld pool, a groove, and a bead. Since the welding condition is analyzed in real time, the welding position and welding power can be controlled in accordance with the analysis result to improve the quality of welding.
[0020]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. In each of the drawings, common portions are denoted by the same reference numerals, and redundant description will be omitted.
[0021]
1. The welding portion visualization device of the present invention is a device that visualizes a welding situation that could not be seen conventionally by controlling a powerful strobe light source and a high-speed electronic shutter.
A welding control device that extracts features of the welding state by image processing using the image of welding obtained by the welding portion visualization device and provides feedback based on the result is studied. By realizing such a welding control device, it is possible to propose a high value-added welding line such as automatic welding correction and defective product detection. FIG. 1 shows a weld visualization device 10 of the present invention and a welding control device 20 using the same.
[0022]
2. Examination using conventional visualization device
Prior to the development of the weld visualization device of the present invention, YAG laser welding was performed using an analog camera + pulse laser illumination type weld visualization device (hereinafter simply referred to as “conventional visualization device”) proposed in JP-A-2000-196923. The video during butt welding was collected. In addition, seam tracking and defect detection were examined based on this video.
[0023]
2.1 Image collection
The welding condition of YAG laser welding was observed using a conventional visualization device, and the state was recorded on a DV tape. In addition, a still image was randomly cut from the captured video, and the brightness of the image was numerically output to confirm and examine the tendency of the image. Further, the image was cut vertically or horizontally, and the luminance distribution on the line was examined.
[0024]
2.2 Examination of images
(1) Observation direction
When performing butt welding, consider how the groove appears when the observation direction is changed.
FIG. 2 shows a luminance change in a certain line on the image of the video in which the state of welding is observed from the side so that the groove (welding line) is horizontal.
A portion surrounded by a circle in FIG. 2 corresponds to a groove. As shown in this figure, the groove appears darker than the surroundings. In addition, in the case of side observation, it was found that the butted edge strongly reflected the laser illumination, and the groove position could be estimated using this.
On the other hand, in the front observation in which the state of welding was observed from the front so that the groove was vertical, it was not possible to obtain sufficient contrast to perform stable image processing. From this, it can be seen that it is necessary to provide illumination and a CCD so that sufficient contrast can be obtained for image processing, and it is necessary to design the entire observation area so that it falls within the depth of field. .
[0025]
(2) Detection of groove gap
It was examined whether a gap generated when two flat plates were welded without being temporarily fixed can be captured. A brightness graph across the groove gap is shown in FIG.
From this figure, it can be seen that when a gap occurs, the luminance of that portion basically drops. Therefore, it can be determined that there is a groove gap if the dark portion is wider than the surroundings.
[0026]
(3) Detection of welding position deviation
Welding was started at a position of -2 mm in the direction perpendicular to the groove, and the welding robot was moved to a position of +2 mm when the welding was completed, thereby virtually causing a welding displacement.
In order to capture such a displacement, it is necessary to recognize the straight line by capturing the groove position on the screen at two or more points, and to check whether a molten pool exists on the straight line. However, basically, since the molten pool exists at a fixed position on the screen, it is sufficient to focus on capturing the groove as a straight line.
FIG. 4 shows a luminance graph when the welding line is cut vertically at two places. By connecting the groove positions at the two cutting positions, the groove can be grasped as a straight line. When this straight line does not pass through a specific point (melt pool) on the image, it indicates that a welding position shift has occurred.
[0027]
(4) Blowhole detection
It is possible to recognize when a typical blowhole appears from an image capturing the state of occurrence of a blowhole. It is also possible to count how many blow holes have appeared in one welding. However, since a wider area must be searched as compared with groove detection, the data processing amount is greatly increased. Therefore, it is necessary to consider an efficient extraction algorithm.
(5) About exception images
When determining a feature amount from an input image, an image that is completely different from the assumed input image may come in.
When seam tracking is performed on such an input image, erroneous recognition of a groove may occur, and the welding head may move to an unexpected position. Therefore, it is necessary to determine what kind of image is input when image processing is started, and whether to actually perform seam tracking or count the number of blow holes.
For example, it is preferable to start the image processing when the robot starts the automatic operation, and to start the seam tracking when the molten pool and the groove are detected with a certain degree of certainty or more. Further, when a frame having a very large amount of disturbance comes in during welding, if the certainty of the processing result is low, the frame is preferably ignored.
[0028]
3. Configuration of welding part visualization device and welding control device of the present invention
FIG. 5 shows a device configuration of the weld visualization device of the present invention and a welding control device using the same.
The weld visualization device 10 according to the present invention includes a combination of a high pixel digital camera 2 and a flash lamp 4. The welding part visualization device 10 of the present invention realizes low cost and improved operability by using the flash lamp 4 instead of the pulse laser as the strobe light source. In addition, the welding part visualizing device 10 of the present invention employs the camera 2 having a high pixel CCD (for example, 1.4 million pixels). The camera 2 has a digital output (for example, 1320 × 1040 Pixel, 25 Hz uncompressed) in addition to the analog output.
The welding control device 20 performs image processing using the digital output. The welding control device 20 needs to have a function of taking in a high-speed and large-capacity digital output output from the camera 2 without omission and reproducing it without omission. In addition, it is necessary to have a function of executing data processing such as image processing at high speed. The following is used as the welding control device 20 for realizing such a function at low cost.
[0029]
(1) Image processing apparatus (PC): OS: Windows (registered trademark) NT 4.0, CPU: Pentium (registered trademark) 4 (1.5 GHz), main memory: RDRAM (768 MB)
(2) Frame grabber board: IPM8560D manufactured by Graphin (DMA transfer function, pixel clock Max 60 MHz)
Hereinafter, a description will be given of a framework for realizing an image capturing and recording function without frame loss using the above-described device.
[0030]
3.1 Realization of real-time image capture function
In the welding control device 20 of the present invention, the welding status is captured as an image from the welding portion visualization device 10, and the welding status is determined based on the image. First, consider the software configuration for capturing images in real time. Note that real-time image capture means capturing the video sent from the camera 2 into the main memory without any frame loss and no frame delay.
[0031]
(1) Import processing by IPM-8560 sample program
FIG. 6 shows a timing chart for capturing a basic image using the IPM-8560D. Hereinafter, the contents of the timing chart will be described along the processing flow of the image capturing software.
(S1) Capture start instruction
Using the IPM-8560D library “mgGrabberStart”, an image for one frame is transferred to a memory table reserved in advance. The mgGrabberStart returns immediately without waiting for the end of loading.
(S2) Wait for synchronization object
The preset synchronization object is in a non-signal state until the capture of one frame is completed. WaitForSignalObject is a WIN32 API that waits for the synchronization object to become signaled.
(S3) Window display processing
The contents (video) of the memory table are displayed on the screen of the PC.
(S4) Copy the contents of the memory table to the main memory
Prior to image processing, the contents of the memory table are copied to the main memory.
(S5) Image processing
Image processing is performed on the contents of the main memory.
[0032]
(2) Image capture processing using a ring buffer
The image capturing process described above is a basic capturing process shown in the sample program of IPM8560D, but can capture only one frame out of two frames. Therefore, by using two or more memory tables together, it is possible to perform capture without frame loss.
In FIG. 7, the number of captures is counted using two memory tables. If the number is odd, the data is transferred to the memory table 1; if the number is even, the data is transferred to the memory table 2. As a result, real-time capturing becomes possible, and a relatively long time (approximately 40 msec) for image processing can be secured.
[0033]
(3) Countermeasures for missing frames
An application for realizing a timing chart using two memory tables was implemented, and operation was confirmed. This application made sure that all frames were captured.
When an experiment was performed using this application, frames were lost once every 10 seconds (250 times) on average (a loss occurrence rate of 0.400%). This is probably because the margin between the frames is very short, and the interrupt indicating that the synchronization object has changed to the signal state could not be received during the short timing.
Therefore, an experiment was performed by transferring only 1392 × 1000 Pixels out of 1392 × 1040 Pixels to extend the margin time between frames. When transfer was performed twice for one hour (90000 times), frame loss occurred only once in both cases (missing rate 0.001%). As described above, although a part of the image is not taken in, it is possible to improve the certainty of real-time taking in while taking into account a necessary range of the image.
[0034]
3.2 Real-time recording function
It is indispensable for the welding control device 20 to have a function of recording an observation image without deterioration or omission and to be able to confirm the image repeatedly. Here, the recording function of the welding control device 20 will be considered.
[0035]
(1) Memory allocation processing
In order to record high-speed, large-capacity image data without deterioration or loss, it is necessary to reliably transfer the image data to the main memory. However, simply securing the memory makes it difficult to record in real time due to paging processing that occurs when a large amount of memory is used.
Therefore, for the process of capturing the video, the working size of the process is increased, and the HD is not accessed at the time of recording by locking the recording area in the memory. This makes it possible to record the captured image in a memory in real time.
Lockable memory is limited by physical main storage capacity. Since the current apparatus configuration uses a 768 MB main memory, the upper limit is 768 MB. The upper limit of the capacity of the memory table that can be managed by the frame grabber board IPM8560D is 512 MB. Therefore, the upper limit of the lockable area is 512 MB.
[0036]
(2) Operation check
The 512 MB memory corresponds to a video of about 14 seconds (1392 × 1040 (Pixel) × 25 (Hz) × 14 (second) = 506,688,000 Bytes). With this device configuration, recording was performed for 14 seconds, and it was confirmed that images could be recorded in the memory without frame loss.
However, it is necessary to copy the contents of the memory to another medium (hard disk, MO disk, CD-R, etc.) before starting the next recording and clear the memory.
[0037]
4. Image sampling test by welding control device of the present invention
The state of YAG laser welding was observed and collected using a welding control device equipped with the above-described image capturing and recording function. Welding tests were performed under various conditions. The main welding conditions are as follows.
(1) Welding method: butt welding, lap welding (for blowhole detection)
(2) Device configuration: coaxial type, single type
Here, the coaxial type has a device configuration in which a jig combined with a mirror or the like is used to collect an image of a welded portion as viewed from directly above. In addition, the stand-alone type is a device configuration in which an independent camera 2 is attached to a welding head, and an image that looks into the welded portion from an oblique direction is collected.
(3) Welding direction: horizontal, vertical, diagonal 45 ° on the screen
Based on these images, an automatic seam tracking system and an automatic defect detection system of the welding control device will be examined.
[0038]
5. Considering automatic seam tracking
The automatic seam tracking function by the welding control device of the present invention will be described below.
5.1 Processing delay in welding control device
In the welding control device of the present invention, as described above, it takes some time to capture the event A that has actually occurred as an image and recognize the event A by image processing. Furthermore, it takes a certain amount of time to realize a drive system process such as driving the stage based on the recognition result of the event A.
This delay is expected to take about 3 frames (120 msec) (assuming 40 msec for image processing and 40 msec for stage driving). FIG. 8 shows this state.
In order to absorb this delay, as shown in FIG. 9, it is necessary to predict an event A that occurs three frames later from the input image and apply a correction process in advance.
[0039]
5.2 1-3 dimensional seam tracking
The functions to be implemented by seam tracking differ depending on whether the direction in which the robot arm is driven by one welding operation is one-dimensional (linear), two-dimensional (plane), or three-dimensional (three-dimensional). FIG. 10 is an image diagram of one-dimensional to three-dimensional seam tracking. Hereinafter, one-dimensional to three-dimensional seam tracking is defined as follows.
(A) One-dimensional seam tracking
The position of the molten pool is always constant, the welding direction is always constant, and the groove appears basically at a fixed position. In this case, since the position of the molten pool several frames ahead can be predicted, a deviation between this position and the groove position is detected, and a process of moving the working head in advance to correct this deviation is performed.
(B) Two-dimensional seam tracking
The position of the molten pool is always constant, but the welding direction is undefined, and the direction in which the groove appears is also undefined. In this case, since it is difficult to predict a few frames ahead, a process is performed to calculate how much the groove is displaced from the current molten pool position, and to move the working head to correct the amount of the displacement.
(C) 3D seam tracking
When observing the periphery of the molten pool from an oblique direction, when a deviation occurs in the height between the processing head and the steel plate, in addition to the above (B), the position of the molten pool is also undefined. In this case, in addition to the two-dimensional processing, it is necessary to calculate the height deviation amount of the processing head from the position of the molten pool and move the processing head also in the height direction.
[0040]
5.3 Examination of groove detection processing flow
The procedure for implementing one-dimensional seam tracking can be roughly divided into the following two procedures.
(A1) Prediction of weld pool position after 3 frames
(A2) Detection of groove straight line by image processing
(A1) The weld pool position after three frames is always fixed during welding. Therefore, the process may be performed once at the start of welding.
The detection of the groove straight line by the image processing of (A2) must be performed for all input images. Therefore, the image processing on the image A needs to be completed within 40 msec until the image B of the next frame is input.
[0041]
The seam tracking processing procedure is as follows.
(S1) Calculate the molten pool position after three frames
(S2) A groove straight line is detected from an image input for each frame.
(S3) Calculate a vector from the molten pool position after three frames to the groove straight line.
(S4) This vector represents the amount of deviation between the molten pool and the groove. If the deviation is large, the jig is driven to correct the deviation. Further, a processing result image is created and displayed on a display.
FIG. 11 shows a groove straight line detection flowchart corresponding to (S2) in the above procedure.
[0042]
5.4 Results of groove detection test
FIG. 12 shows a screen example of the groove detection application in which the above-described seam tracking processing flow is implemented.
Using this application, a groove detection processing test was performed on observation video using a coaxial weld visualization device (a weld visualization device using a jig that can capture an image viewed from directly above). As a result, as long as the groove was not wide open, no erroneous recognition occurred, and the recognition rate of the groove was about 90%, and almost no problem was obtained. When the groove was wide open, the groove recognition rate was as low as 60%. In such a case, it is preferable to make an error display, for example, a judgment such as “a gap has occurred in the groove and welding has failed”.
[0043]
Next, the image processing speed will be discussed.
In the welding control device of the present invention, the image frame interval is 40 msec, and it is necessary to complete the image processing within this time (40 msec) in order to add image processing to all frames without omission.
The average image processing speed by the program developed this time was about 20 msec for one frame of the image (Pentium (registered trademark) 4, 1.5 GHz). This means that image processing can be basically applied to all frames without omission.
From these results, in the YAG laser welding observation using the coaxial weld visualization device of the present invention, it is possible to detect a groove with a probability of about 90% without frame loss.
[0044]
6. Examination of automatic defect detection
Hereinafter, the results of studies on the realization of the automatic defect detection function using the weld visualization device of the present invention will be summarized.
6.1 Blowhole detection processing
An example of a welding defect is a blowhole generated during lap welding. If blowholes occur frequently, the welding strength will be insufficient, and it will be treated as a defective product.
Here, the processing procedure of the defect detection system that detects blow holes generated during welding and counts the number is as follows.
(S1) Calculate blowhole observation area
(S2) The presence or absence of blow holes in the observation area is checked for each image of each frame. (S3) The number of blows in one welding is counted and presented to the operator.
The blowhole observation area is an area determined by the weld pool position, the welding vector, the welding speed, the visual field range, and the like, where there is no overlap for each frame.
FIG. 13 shows a flowchart of the blowhole detection process.
[0045]
A blowhole detection test was performed using a blowhole detection application that implements the blowhole detection processing flow. There were four blowholes in the target image, and all of them were recognized as blowholes. However, one other place and a place other than a blowhole were erroneously recognized as blowholes.
From this, it is considered that the blowhole has a feature that can be captured by image processing and can be detected with a certain degree of accuracy.
[0046]
6.2 Consideration of groove gap detection processing flow
As an example of a welding defect, a case where a gap is generated in a groove at the time of butt welding and welding fails is assumed.
Here, the groove width was detected and output using the groove detection processing application described above. This is shown in FIG. The abscissa represents the passage of time (the number of frames) and the ordinate represents the width of the groove (Pixel). From this figure, it can be seen that the groove width gradually increases from around the 140th frame. As described above, if the image can detect the groove, the transition of the groove width can be grasped.
[0047]
It should be noted that the present invention is not limited to the above-described embodiment, and can be variously changed without departing from the gist of the present invention.
[0048]
【The invention's effect】
As described above, the welding control device and method according to the present invention can be used to obtain a groove position, a groove width, a weld pool position, a weld pool state, a bead position, a high-quality, high-resolution image obtained by the weld visualization device 10. An apparatus and method for detecting a bead state and similar welding features and analyzing a welding situation in real time from the feature amount. The analysis result is fed back to the welding machine in real time, and the welding position and power are automatically corrected. Thereby, it is possible to prevent a defect that is expected to occur in the future.
[0049]
Therefore, the welding part visualization device of the present invention can visualize the welding situation near the molten pool at a high resolution, which could not be seen conventionally at the time of welding, and the welding control device and method of the present invention provide the obtained image. The welding condition can be analyzed in real time by using, and the welding position and welding power can be controlled in accordance with the analysis result to improve welding quality.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a weld visualization device of the present invention and a welding control device using the same.
FIG. 2 is a diagram illustrating a luminance change in a certain line on an image in side observation.
FIG. 3 is a diagram showing a luminance graph traversing a groove gap;
FIG. 4 is a diagram showing a luminance graph when a welding line is cut vertically at two places.
FIG. 5 is a view showing a device configuration of a weld visualization device of the present invention and a welding control device using the same.
FIG. 6 is a diagram showing a basic image capture timing chart.
FIG. 7 is a diagram showing a timing chart for capturing an image using two memory tables.
FIG. 8 is a diagram showing a delay in correction processing for event A.
FIG. 9 is a diagram showing a prediction / correction process of event A.
FIG. 10 is an image diagram of one-dimensional to three-dimensional seam tracking.
FIG. 11 is a flowchart of groove straight line detection in seam tracking processing.
FIG. 12 is a halftone image on a CRT showing an example of a screen for groove detection by the seam tracking processing flow.
FIG. 13 is a flowchart of a blow hole detection process.
FIG. 14 is a diagram showing a detection state of a groove gap.
[Explanation of symbols]
2 High pixel CCD camera,
4 High brightness flash lamp,
6 coaxial optical system,
10 weld visualization device,
12 image capture device,
14 image processing devices,
16 output devices,
20 Welding control device

Claims (13)

Welding comprising a high-pixel CCD camera (2) with an electronic shutter and a high-brightness flash lamp (4), wherein illumination of the welding portion by the high-brightness flash lamp and photographing by the high-pixel CCD camera are synchronized Part visualization device. 2. The CCD camera according to claim 1, wherein the CCD camera continuously captures images at intervals exceeding a transfer time Δt required for transferring one frame of images to the memory table and outputs image data. The visualization device for a welded part according to item 1. The welding part visualization device according to claim 1, further comprising: an optical system (6) for guiding an image of the weld including the molten pool to the CCD camera (2) in butt welding or lap welding. An image capturing device (12) that captures image data output by the welding portion visualization device (10); an image processing device (14) that performs image processing on the image data to determine a welding status; and an output that outputs the welding status. A welding control device, comprising: a welding device (16). The welding control device according to claim 4, further comprising a PC for image processing and a frame grabber board for capturing image data. 5. The welding control apparatus according to claim 4, comprising a plurality of memory tables, sequentially transferring one frame of image to each memory table, and taking in all the images in real time as a whole. Weld pool, groove, and bead are photographed as high-resolution image data, and from this image data, at least one of a groove position, a groove width, a molten pool position, a molten pool state, a bead position, or a bead state is detected. And a welding condition is analyzed in real time from the detected contents. The welding control method according to claim 7, wherein the analysis result is fed back to the welding device in real time to automatically control the welding device. The welding control method according to claim 7, wherein at least one of groove profiling, groove width detection, blow hole detection, and welding condition abnormality determination is analyzed in real time. In the groove tracing, the image is analyzed to determine the straight line vector of the groove, the current weld pool position, and the future weld pool position. The welding control method according to claim 7, wherein it is determined whether or not the welding occurs, and when a misalignment is predicted, the welding head is moved to improve welding quality. The welding control according to claim 7, wherein in the groove width detection, a groove width is detected, and when the groove is opened, a welding head is moved to one edge to prevent welding failure. Method. The welding control method according to claim 7, wherein in the blow hole detection, the state of the bead is determined, and when a blow appears, the welding power is adjusted. In the welding condition abnormality determination, a normal welding condition is grasped, and a normal welding condition is compared with a real-time welding condition at all times, thereby grasping a welding abnormality and notifying the abnormality. Item 8. The welding control method according to Item 7.

Images