JP5200469B2 - Welding spark detection method, spark detection device, and welded product manufacturing method - Google Patents

Description

  The present invention relates to a method for detecting sparks generated when performing high-frequency heating welding or high-frequency resistance welding, a detection device for sparks, and a method for manufacturing a welded product.

  For example, in high-frequency heating welding and high-frequency resistance welding used for the manufacture of welded products such as electric-welded steel pipes, the heat generation and light emission of the welded portion are almost constant and stable under normal conditions. On the other hand, when abnormalities occur due to various welding conditions, fluctuations occur in the current and voltage of the heating coil and sparks are generated.

  Spark is a phenomenon in which sparks violently scatter around the welded part, and it is known that the part where the spark is generated often results in poor welding. Accordingly, it is necessary to accurately detect whether or not a spark has occurred during welding, and when the spark is detected, it is necessary to cut and remove the part so as not to be shipped to the consumer.

  Patent Document 1 discloses a technique for detecting a spark by monitoring the output current and voltage of a high-frequency heating device. This technology monitors the current and voltage of the coil for high-frequency heating, and captures that because the current and voltage change suddenly when a spark occurs, compared with normal welding. It is a method to try to detect.

Further, Patent Document 2 discloses a technique for detecting by utilizing the characteristic that a welded portion becomes brighter when a spark is generated as compared with normal welding. This technique is a method of observing a welded part with a camera and monitoring the amount of light emitted from the welded part.
JP 2001-198682 A JP 10-34354 A

  However, during welding, spatter, which is a light emission phenomenon due to sparks similar to sparks, also occurs. In general, spatter occurs more frequently than sparks, but there is a low probability that the sputtered portion will be welded poorly. Therefore, it is important to discriminate from the spatter when detecting the spark.

  However, in the technique of Patent Document 1 that detects from fluctuations in the current and voltage of the high-frequency heating coil, both spark and spatter have characteristics that cause fluctuations in the current and voltage of the high-frequency heating coil. It is difficult to discriminate.

  Further, even in the technique of monitoring the light emission intensity of Patent Document 2, since both the spark and the spatter are similar phenomena in that the light emission intensity of the weld changes, it is extremely difficult to distinguish between the spark and the spatter. is there.

  As described above, in the conventional technology in which it is difficult to discriminate between spark and spatter, since spatter is excessively detected as spark, there are many re-inspection work in the focused quality inspection before shipment using ultrasonic flaw detection. As a result, there is a problem that production efficiency is hindered.

  An object of the present invention is to provide a spark detection method, a spark detection apparatus, and a welding product for high-frequency resistance welding that can accurately and reliably detect sparks generated during high-frequency resistance welding in order to solve the above-described problems. A manufacturing method is provided.

  First, in order to examine the discrimination method between sparks and spatters, the inventors observed the welding situation when manufacturing an ERW welded steel pipe by visual observation. As a result, it was qualitative that sparks showed very short flashes and sparks, while spatters showed bright sparks for a relatively long time, but no paleness. Gained insights.

  Next, in order to grasp the knowledge in visual observation quantitatively, the welding situation at the time of manufacture of an ERW welded steel pipe was imaged using the experimental apparatus shown in FIG.

  The welded portion 2 of the ERW steel pipe 3 being manufactured by the high-frequency heating device 1 was imaged at 30 frames (frames) per second using the imaging device 4 (here, a color camera). Then, images that are thought to have sparked or sputtered are extracted, and these images are associated with the results of confirming the quality of the welded part of the product by visual inspection, ultrasonic flaw detection, etc. in the subsequent process, and poor welding The computer 6 analyzed and compared the difference between the image of the spark accompanied with the spatter and the image of the spatter that did not result in poor welding.

  FIG. 2 corresponds to an image corresponding to a part determined to be a poorly welded portion (that is, a spark is generated at the time of welding) in a post-process confirmation, and corresponds to a part that was light-emitting but not a poorly welded (that is, spatter was generated) It is the graph which isolate | separated the image corresponding to a normal site | part without an image and light emission into the image of the color component of blue, green, and red, respectively, and compared the average luminance of those images.

  Looking at the average luminance of each color component of the image when a spark occurs, the increase in the blue component is particularly significant. On the other hand, for the sputtered image not accompanied by poor welding, the total amount of light becomes brighter, but the blue component has no particular change and rather tends to increase the red component.

  Furthermore, it has been found that there is a spark image in which the average luminance of the brightness hardly changes, and a change in which the luminance of the blue region only increases appears. This is, for example, a result suggesting the possibility of missing a spark in the method of observing only the light amount change without considering the color region as disclosed in Patent Document 2.

  Next, the results of examining the spark and spatter generation time will be described. 3 and 4 are conceptual diagrams of images of several frames before and after the time when both phenomena occur. The imaging time (exposure time) of one image frame was about 33 milliseconds, and each figure showed images of five consecutive frames every 33 milliseconds.

  In FIG. 3 showing the state at the time of the occurrence of spark, the other first, third, fourth, and fifth are during normal welding except that the spark can be confirmed only in the second frame (the image described as (2) in FIG. 3). It was an image.

On the other hand, in FIG. 4 showing the state at the time of occurrence of spatter, spatter can be confirmed in three frames of the second to fourth frames (images denoted as (2) to (3) in FIG. 4). these,
From FIG. 3 and FIG. 4, although the spark emission time is within 33 milliseconds, the sputter emission time lasts up to about 99 milliseconds corresponding to 3 frames, so that sputtering is more effective than spark. It can be determined that the light emission time has a long characteristic.

From the above, the following knowledge was obtained as the difference between spark and sputtering.
(1) As for the light emission of the spark, a change is seen in the blue component, and the light emission stops within 1/30 seconds.
(2) In the emission of spatter, no change is observed in the blue component, and the emission continues for more than 1/30 seconds.
The present invention has been made based on these findings,
According to a first aspect of the present invention, there is provided an imaging step of imaging a light emission state generated in a welded part during welding, an extraction step of extracting a wavelength component intensity of 200 nm to 500 nm from the captured image, and a wavelength component intensity of 200 nm to 500 nm. A welding spark detection method comprising: a determination step of determining a spark.

According to a second invention, in the spark according to the first invention, the determination step determines that a spark has occurred when a change amount of the wavelength component intensity of 200 nm to 500 nm is a predetermined value or more. It is a detection method.

According to a third aspect of the present invention, the determination step determines that a spark has occurred when an image in which the change amount of the wavelength component intensity from 200 nm to 500 nm is a predetermined value or more is continuously a predetermined number or less. A spark detection method according to the second aspect of the invention.

The fourth aspect of the invention is an imaging device that images a light emission state generated in a welded part during welding, an image is input from the imaging device, a wavelength component intensity from 200 nm to 500 nm is calculated from the image, and the 200 nm to 500 nm A spark detection device comprising an image processing device for determining whether or not a spark is generated based on a wavelength component intensity.

According to a fifth aspect of the present invention, an image pickup device that picks up an image of a light-emitting state generated in a welded portion during welding through a filter that transmits a wavelength of 200 nm to 500 nm, and an image is input from the image pickup device to obtain an intensity of the image. A spark detection device comprising an image processing device that determines whether or not a spark has occurred based on the image processing device.

  6th invention uses the welding process which welds with respect to a metallic material, and the spark detection method of the welding in any one of Claims 1 thru | or 3 WHEREIN: The quality of the welding part welded by the said welding process is determined. It is a manufacturing method of a welding product including the inspection process which judges.

  Since the spark detection method and the spark detection device of the present invention can reliably detect a spark with poor welding, excessive detection due to sputtering is suppressed, and the re-inspection time is greatly shortened. In addition, since it has been possible to reliably cut and remove the defective welded portion, an improvement in quality assurance level can be expected when manufacturing the welded product.

  An example in which the embodiment of the present invention is applied to a production line for an electric resistance welded steel pipe will be described below with reference to FIGS. 5 and 6.

  In the manufacturing process of the electric resistance welded steel pipe, the electric resistance welded steel pipe 3 is continuously manufactured by electro-welding the steel strip by the high frequency heating device 1. A number of unillustrated rolls exist around the welded portion 2, and the ERW steel pipe 3 is conveyed in the direction of arrow 11 by these rolls.

  In such a manufacturing process, for example, the imaging device 4 such as a monochrome CCD camera or a CMOS camera is obliquely upward so that light-emitting phenomena such as sparks and spatter in the welded portion 2 where welding is performed can be monitored. Install. The field of view (field size) may be appropriately set so as to be able to capture an image including the periphery illuminated by the spark 15 when the spark 15 is generated. For example, the field of view may be about 300 mm square around the welded portion. .

  Note that the position of the imaging device 4 is not necessarily as described above. However, since a dense plume is generated from the welded portion 2 and high-temperature molten iron powder is scattered around by the spark 15, the welded portion. It is preferable to avoid the position just above 2 and to dispose at least about 500 mm apart.

  Note that the imaging apparatus 4 has an imaging rate of 30 frames (frames) per second and the exposure time is maximum. This is because if the exposure time is shorter than the time at the image interval of the imaging rate, there is a possibility that a short-time emission spark may be missed. Therefore, if the imaging rate is 30 frames / second, the exposure time is 1/30 seconds (33 milliseconds).

  Blue filter having a characteristic that the transmittance is maximum in the wavelength range of about 300 nm or more and about 500 nm or less, and the light intensity of the wavelength component of about 300 nm to about 500 nm is more transmitted than the other wavelengths on the front surface of the camera lens. Wear 5 As a result, even a monochrome imaging device can extract and receive the blue component of the emitted light, and can detect sparks by distinguishing it from sputtering by changing the signal.

  FIG. 7 shows an example of the transmission characteristics of the blue filter 5. In the example of FIG. 7, the transmittance is maximum at about 400 nm, and the light transmitted through the blue filter is mostly occupied by the blue component, which is a wavelength component in the range of 200 nm to 500 nm. In the present invention, the transmission characteristic of the blue filter 5 is an important factor, and it is preferable to select a filter having the highest spark detection accuracy from a plurality of filters having different characteristics by experiments or the like. It may be determined as appropriate in consideration of the above.

In this way, the image signal of the imaged device 4 that has been imaged is input to the signal processing device 6.
In addition, the signal processing device 6 receives a signal of the rotary encoder 8 for measuring the manufacturing distance (the manufactured steel pipe length) of the ERW steel pipe 3 and tracking its position.

  In the signal processing device 6, the average luminance is calculated for each frame of the image input from the imaging device. Here, the average luminance is calculated by averaging the luminance of pixels in a preset partial area of the field of view of the camera or the luminance of pixels in the entire screen area. Then, a spark generation determination logic, which will be described later, is executed based on the calculated temporal change of the average luminance (change for each frame).

  The brightness in this case is not the average brightness of the entire image, but a partial range centering on the welded part in the image (for example, all pixels are 512 pixels wide and 480 pixels wide, 300 pixels wide and 200 pixels high) It is also possible to select and calculate only the range of degree. This is because a change in average luminance is detected more sensitively by selecting only a portion where fluctuations are likely to occur.

  Further, the manufacturing distance (manufacturing length) of the ERW steel pipe is calculated from the signal of the rotary encoder 8, and when it is determined that the spark 15 has occurred, the manufacturing distance (manufacturing length) is stored in the storage device. By doing so, it is possible to manage at which position of the ERW steel pipe 3 the spark 15 is generated. In addition, by outputting the position where the spark 15 is detected to the automatic marker 12 installed in the subsequent process of this inspection process, the position is tracked, and the presence of a defect is detected at the spark occurrence position on the outer surface of the ERW steel pipe. Defect marking to be displayed.

  In addition, instead of the average luminance of the entire region or the partial region, a luminance level that can be determined as a spark occurrence is set as a threshold value, and for each image, a pixel (area) that exceeds the threshold value is obtained, and it exceeds a predetermined value. In such a case, the spark may be determined. This determination is white (bright) when a spark occurs in the frame (2) in FIG. 3 and when only normal welds such as the frame (1) and the frames (3) to (5) are seen. This is because the area of the visible portion is different, and it is used.

  FIG. 8 shows an example in which the average luminance of 8 consecutive frames of images is calculated. The second spark generation frame from the left in FIG. 8 is the average luminance at the time when the spark is generated, and the other frames are the average luminance at the time of normal welding. Similarly, FIG. 9 shows the state of average brightness of sputtering.

  As apparent from FIGS. 8 and 9, the blue filter is used to increase the luminance of the spark and decrease the luminance at the time of sputtering and normal. Therefore, a threshold value may be set for the luminance level between them. In the case of FIG. 8, for example, when the threshold value of the average luminance is 200, the average luminance exceeds 200, and it is only an image of one frame, it is determined as a spark.

  An example of a flowchart of this determination process is shown in FIG. First, an image is input from the imaging device every 1/30 seconds (step S1). Next, for the input image, the average luminance of the entire region or a partially selected region is calculated (step S2). Then, it is determined whether the average luminance calculated in step S2 exceeds a preset threshold value (step S3). Then, it is determined whether the average luminance of the image of the next frame is less than the threshold value B (light emission time is within 1/30 second) (step 4). As a result, in the case of Yes, it is determined that it is a spark (step 5).

  In the above-described embodiment, it has been described that a spark is always within an exposure time of 1/30 seconds, that is, fits in one image. However, this is not always the case, and the beginning and end of a spark may span two images. In that case, it can be overlooked (undetected), so such a situation must also be considered.

  Therefore, an average value of luminances of two adjacent images (before and after) is calculated, and determination based on the change in the value is also performed. The concept of this determination will be described below with reference to FIG.

  In FIG. 11, the horizontal axis indicates the passage of time, that is, the image number, and the vertical axis indicates the relative average luminance of each image when the average image luminance during normal welding is 100. The white bar graph in FIG. 11 is the average luminance for each frame of the image, and has the same meaning as the graph shown in FIG. Furthermore, the addition average value of two adjacent images (in FIG. 11, expressed as two-frame addition average luminance) is the average value of the two-frame addition average luminance before a predetermined number of frames (for example, 20 frames) with respect to the current image. The average value of the two images currently being processed is displayed as a relative value. In addition, in FIG. 11, in the ultrasonic flaw detection test before shipment, locations that were confirmed to be poor welds were around 2 to 3, near 9 to 10, and around 15 to 17 when associated with the frame numbers on the horizontal axis. .

  On the other hand, in the process of FIG. 10, if the threshold value B (a value that is a spark candidate when exceeding B for each frame, which is the same as the threshold value B of FIG. 10) is 200, for example, it is determined as a spark. Are image numbers 2 and 16 on the horizontal axis in FIG. Therefore, the vicinity of 9 to 10 will be overlooked (undetected). The reason for this is considered to be that it spans two screens, and a decision is made not to detect undetected using the addition average value of two images.

  Here, since the addition average value of two adjacent images is always an intermediate value between the images, the threshold A when detecting a spark based on the addition average value of the two images is set for each image shown in FIG. It must be set lower than the threshold used for the determination process. In FIG. 11, the threshold A for detecting the screen number 10 is set to 150, for example.

A flowchart of this determination process will be described below with reference to FIG.
First, an image is input from the imaging device every 1/30 seconds (step S11). Next, with respect to the input image, the average luminance of the entire region or a partially selected region is calculated, and an addition that is an average value of the average luminance calculated in the previous image and the average luminance of the image input this time Average luminance is obtained (step S12). Although not shown, when the first frame at the start of measurement is input, since there is no previous frame, only the average luminance data of the first frame is calculated, and the addition averaging process of the two screens is not performed. Return to step S11.

  Next, the value of the addition average luminance calculated in step S12 is compared with a threshold A for detecting a spark (step S13). Here, as described above, the threshold A is set lower than the threshold B used for the determination process for each image shown in FIG. 10 because the addition average luminance is always an intermediate value between the images.

When the addition average luminance is equal to or less than the threshold value A (No in step S13), it is determined that no spark is generated, and the process returns to step S11.
When the addition average luminance exceeds the threshold A (Yes in step S13), it is determined whether the addition average value calculated at the timing of the next frame is less than the threshold A (step S14). It determines that it is a spark in the case where the previous addition average luminance exceeds the threshold value A and the subsequent addition average luminance is less than the threshold value A among the addition average luminance values for every two consecutive screens (step S15). In the example of FIG. 11, the image numbers 3, 10, and 17 are determined to be sparks by the processing of the flowchart of FIG. 12, and it is possible to prevent oversight.

  In addition, a very large-scale light emission (for example, image number 15 in FIG. 11) is determined to be any abnormality regardless of whether it is spark or sputtering. Therefore, regardless of the light emission time, that is, the number of images, when the average luminance is very large even in an image of one frame, for example, when a high threshold is exceeded as shown by threshold C in FIG.

  This determination flowchart is shown in FIG. In this flowchart, regardless of the light emission time, it is determined that a spark has occurred when the average luminance of a certain image exceeds the threshold value C (Yes in step S23) (step S24). In the example of FIG. 11, the number 15 is determined to be a spark by the processing in the flowchart of FIG. 13.

  As described above, in this embodiment, there are three cases of A, B, and C shown below as a spark determination.

  (A) When the average brightness of an image exceeds the threshold B and the average brightness of the next image is less than the threshold B.

  (B) A case where the addition average value of the average luminance of two consecutive images exceeds the threshold A and the next addition average value is less than the threshold A.

  (C) When the average luminance of the image exceeds the threshold C regardless of the light emission time.

  However, it is assumed that threshold A <threshold B <threshold C.

  In the implementation results of the present invention, 90% or more of the cases recognized as sparks were cases of (A), but there were also examples that met the conditions of (B). It was possible to avoid the place where the luminance was supposed to be undetected because the luminance was below the threshold value B. In addition, regardless of the light emission time, large-scale light emission that seems to be an obvious welding abnormality could be detected under the condition (C).

  As a method for adjusting the threshold value, for example, the threshold value A and the threshold value B are initially set to be extremely low. In this way, a small-scale spark that does not involve a fatal welding failure is also detected, so that an appropriate threshold value may be determined by gradually increasing the set value while taking the relationship between the detection result and the welding failure.

  Since the threshold value also depends on manufacturing conditions, that is, the material size such as the diameter and thickness of the steel pipe, the welding speed, the product grade, and the like, the threshold value can be set for each manufacturing condition.

  Also, instead of calculating the average luminance directly from the image, binarization processing is performed for each image, and the area of the light emitting portion including the welded portion (the number of pixels exceeding the binarization threshold) is obtained. A similar detection result can be obtained by a method of determining a spark when the value is exceeded. Which to select may be determined by the specifications of the image processing apparatus.

  In the above description, the blue filter is installed before the monochrome camera. However, only a blue component signal may be extracted and processed using a color camera.

  Furthermore, in this embodiment, a camera capable of imaging at an image rate of 30 frames per second is used. However, if real-time image processing is in time, for example, a camera capable of high-speed imaging of 60 frames per second or more is used. Measurement may be performed with increased resolution. On the other hand, in order to obtain and process an image similar to 30 frames per second, a method of calculating and monitoring an addition value or an addition average value of average luminance values of a plurality of adjacent frames may be used. In addition, although the time reference is set to 1/30 second, the present invention is not limited to this, and the value may be appropriately changed according to measurement conditions such as a measurement target and an imaging cycle.

  Even if normal welding is performed, the average luminance varies little by little depending on the manufacturing conditions. Therefore, when the change in average luminance is observed as an absolute value, the average luminance during normal welding is about 100. Therefore, it is desirable to automatically control the sensitivity of the CCD camera. Alternatively, the average luminance for each image to be calculated may be controlled, for example, by averaging the luminance of about 10 images going back from the present time and making the average value about 100.

  As another way to deal with fluctuations in the average brightness of normal welding, calculate the temporal moving average value of the average brightness, calculate the change rate of the average brightness for each image, and if a specific change rate is exceeded You may decide that it is a spark.

  Then, as shown in FIG. 6, the position where the spark is detected is tracked by the rotary encoder 8, and marking is performed at the spark occurrence location by the automatic marker. The marking range may be a length that can be surely recognized by visual confirmation or the like in the post-reinspection process, and may be, for example, about 500 mm in length with the spark spot interposed therebetween. The marked part is re-inspected by ultrasonic flaw detection or the like, and if there is a defect, the part is cut off and shipped.

In addition, although the example applied to the manufacturing method of ERW welded steel pipe was explained,
It goes without saying that the spark detection method of the present invention can also be applied to a method for manufacturing a welded product manufactured by welding other than an electric resistance welded steel pipe, for example, a use for monitoring welding of a structure or the like.

It is a figure which shows the method of imaging the welding part of an electric resistance steel pipe. It is a figure which shows the average brightness | luminance in the color area division of each image. It is a sketch which shows the image at the time of spark generation. It is a sketch which shows the image at the time of spatter generation. It is a figure which shows the apparatus structure of a spark detection apparatus. It is a figure which shows the Example of invention. It is a figure which shows the permeation | transmission characteristic of a blue filter. It is a figure which shows the image average brightness | luminance before and after spark generation. It is a figure which shows the image average brightness | luminance before and behind sputter | spatter generation | occurrence | production. It is a spark detection flow based on average luminance and light emission time. It is a figure explaining the relationship between a relative average brightness | luminance and a threshold value. It is a spark detection flow using the addition average luminance. It is a large-scale light emission abnormality detection flow.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 High-frequency heating device 2 Welding part 3 ERW steel pipe 4 Imaging device 5 Blue filter 6 Computer 7 Conveyance roll 8 Rotary encoder 9 Direction of passage of steel pipe 11 Direction of conveyance of steel pipe product 12 Automatic marker 13 Steel pipe including poorly welded part 14 Product 15 Spark

Claims (6)

An imaging process for imaging a light emission state generated in a welded part during welding, an extraction process for extracting a wavelength component intensity of 200 nm to 500 nm from the captured image, and a determination process for determining a spark from the wavelength component intensity of 200 nm to 500 nm And a method for detecting a spark of welding. 2. The spark detection method according to claim 1, wherein the determination step determines that a spark has occurred when a change amount of the wavelength component intensity from 200 nm to 500 nm is a predetermined value or more. The determination step is characterized in that it is determined that a spark has occurred when an image in which a change amount of the wavelength component intensity from 200 nm to 500 nm is equal to or greater than a predetermined value is continuously equal to or less than a predetermined number. 2. The spark detection method according to 2. An imaging device that captures a light emission state generated in a welded part during welding, an image is input from the imaging device, a wavelength component intensity of 200 nm to 500 nm is calculated from the image, and based on the wavelength component intensity of 200 nm to 500 nm A spark detection device comprising an image processing device for determining whether or not a spark has occurred. The light emitting state generated in the welded part at the time of welding is imaged through a filter that transmits a wavelength of 200 nm to 500 nm, and an image is input from the imaging device, and whether or not a spark is generated is determined based on the intensity of the image. A spark detection device comprising an image processing device for determination.   A welding process for welding the metal material, and an inspection process for determining the quality of the welded portion welded in the welding process using the welding spark detection method according to any one of claims 1 to 3. A method of manufacturing a welded product including:

Images