CN112548321A - Coaxial monitoring-based vacuum laser welding seam defect identification method - Google Patents
Coaxial monitoring-based vacuum laser welding seam defect identification method Download PDFInfo
- Publication number
- CN112548321A CN112548321A CN202011400726.6A CN202011400726A CN112548321A CN 112548321 A CN112548321 A CN 112548321A CN 202011400726 A CN202011400726 A CN 202011400726A CN 112548321 A CN112548321 A CN 112548321A
- Authority
- CN
- China
- Prior art keywords
- welding
- molten pool
- vacuum
- laser welding
- monitoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003466 welding Methods 0.000 title claims abstract description 123
- 238000000034 method Methods 0.000 title claims abstract description 90
- 230000007547 defect Effects 0.000 title claims abstract description 56
- 238000012544 monitoring process Methods 0.000 title claims abstract description 37
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims description 50
- 238000013528 artificial neural network Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 10
- 210000001503 joint Anatomy 0.000 claims description 10
- 239000010953 base metal Substances 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 8
- 229910000838 Al alloy Inorganic materials 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000002945 steepest descent method Methods 0.000 claims description 4
- 239000002184 metal Substances 0.000 claims description 3
- 229910052751 metal Inorganic materials 0.000 claims description 3
- 150000002739 metals Chemical class 0.000 claims description 3
- HEMHJVSKTPXQMS-UHFFFAOYSA-M Sodium hydroxide Chemical compound [OH-].[Na+] HEMHJVSKTPXQMS-UHFFFAOYSA-M 0.000 claims 3
- 238000005406 washing Methods 0.000 claims 3
- CSCPPACGZOOCGX-UHFFFAOYSA-N Acetone Chemical compound CC(C)=O CSCPPACGZOOCGX-UHFFFAOYSA-N 0.000 claims 2
- 239000007864 aqueous solution Substances 0.000 claims 2
- 230000007797 corrosion Effects 0.000 claims 2
- 238000005260 corrosion Methods 0.000 claims 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims 2
- GRYLNZFGIOXLOG-UHFFFAOYSA-N Nitric acid Chemical compound O[N+]([O-])=O GRYLNZFGIOXLOG-UHFFFAOYSA-N 0.000 claims 1
- 239000003513 alkali Substances 0.000 claims 1
- 238000005238 degreasing Methods 0.000 claims 1
- 229910017604 nitric acid Inorganic materials 0.000 claims 1
- 230000003647 oxidation Effects 0.000 claims 1
- 238000007254 oxidation reaction Methods 0.000 claims 1
- 238000001914 filtration Methods 0.000 abstract description 7
- 238000011161 development Methods 0.000 abstract description 4
- 238000012545 processing Methods 0.000 description 25
- 230000003287 optical effect Effects 0.000 description 13
- 238000012360 testing method Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- PXHVJJICTQNCMI-UHFFFAOYSA-N Nickel Chemical compound [Ni] PXHVJJICTQNCMI-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000006185 dispersion Substances 0.000 description 2
- 238000004020 luminiscence type Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000000155 melt Substances 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000006641 stabilisation Effects 0.000 description 2
- 238000011105 stabilization Methods 0.000 description 2
- 229910000881 Cu alloy Inorganic materials 0.000 description 1
- 229910001069 Ti alloy Inorganic materials 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 239000000956 alloy Substances 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 229910052759 nickel Inorganic materials 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K26/00—Working by laser beam, e.g. welding, cutting or boring
- B23K26/02—Positioning or observing the workpiece, e.g. with respect to the point of impact; Aligning, aiming or focusing the laser beam
- B23K26/03—Observing, e.g. monitoring, the workpiece
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K26/00—Working by laser beam, e.g. welding, cutting or boring
- B23K26/02—Positioning or observing the workpiece, e.g. with respect to the point of impact; Aligning, aiming or focusing the laser beam
- B23K26/03—Observing, e.g. monitoring, the workpiece
- B23K26/032—Observing, e.g. monitoring, the workpiece using optical means
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K26/00—Working by laser beam, e.g. welding, cutting or boring
- B23K26/12—Working by laser beam, e.g. welding, cutting or boring in a special atmosphere, e.g. in an enclosure
- B23K26/1224—Working by laser beam, e.g. welding, cutting or boring in a special atmosphere, e.g. in an enclosure in vacuum
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B23—MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR
- B23K—SOLDERING OR UNSOLDERING; WELDING; CLADDING OR PLATING BY SOLDERING OR WELDING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; WORKING BY LASER BEAM
- B23K26/00—Working by laser beam, e.g. welding, cutting or boring
- B23K26/70—Auxiliary operations or equipment
Abstract
A vacuum laser welding seam defect identification method based on coaxial monitoring relates to a vacuum laser welding seam defect identification method. The invention aims to solve the technical problem that the monitoring of laser welding pool and surface characteristic information under the existing vacuum environment is limited by the volume of a vacuum bin. The invention relates to a real-time welding surface molten pool dynamic monitoring system in a vacuum chamber, which takes a high-dynamic industrial CMOS camera as a main sensing device, integrates a light filtering system, a computer and other devices, and provides a basic development platform for realizing the application in the industry in the future based on the coaxial monitoring of the molten pool, the extraction of image characteristic information and the defect identification and judgment.
Description
Technical Field
The invention relates to a method for identifying weld defects of vacuum laser welding.
Background
The vacuum laser welding method presents huge technical advantages after more than 30 years of development and has wide market development prospect. The most obvious characteristics of the method are shown by summarizing the current research situation at home and abroad: large fusion depth, small brilliance and high quality. Meanwhile, the vacuum laser welding method is suitable for various common engineering materials, and particularly has excellent welding effect on difficult-to-weld materials such as copper alloy, titanium alloy, nickel-based high-temperature alloy and the like.
In order to ensure a vacuum environment, welding is generally carried out in a vacuum chamber, so that the size of a workpiece is limited, and the detection of dynamic change of molten pool and surface characteristic information in the welding process is greatly limited. And the stability of the molten pool is closely related to the welding quality in the welding process. Therefore, the development of a detection technology of the vacuum laser welding process based on the visual sensing of the weld pool image becomes a key for ensuring the quality of the welding joint.
In various monitoring modes of the welding process, the visual sensing detection area based on optical signals is larger, the obtained image information is more visual and rich, the information amount is larger, and the attention of researchers at home and abroad is received. The laser welding visual monitoring mode can be divided into coaxial acquisition and paraxial acquisition according to the acquisition angle of an optical signal. Because the volume in the vacuum bin is limited, the paraxial acquisition image is very easily limited by the space size, and compared with the paraxial acquisition, the coaxial acquisition is more comprehensive and accurate in information such as the shape and size of the molten pool because the image is directly acquired from the right upper part of the molten pool. Because the overlooking images of the molten pool and the keyhole are obtained, no deformation occurs, and the method is favorable for image processing and characteristic information extraction after the images are collected. Meanwhile, the coaxial monitoring system has the advantages of compact structure, easier installation and small occupied space, so that the coaxial monitoring system has very high industrial applicability.
Disclosure of Invention
The invention provides a vacuum laser welding seam defect identification method based on coaxial monitoring, aiming at solving the technical problem that the laser welding pool and surface characteristic information monitoring in the existing vacuum environment are limited by the volume of a vacuum bin.
The method for identifying the weld defects of the vacuum laser welding based on the coaxial monitoring is carried out according to the following steps:
placing two base metals to be welded in a vacuum bin, wherein the joints are in butt joint, starting a laser head to weld, transmitting a dynamic image of a molten pool to a lens of a camera along light rays in the welding process, and connecting a data output end of the camera with an input end of a computer; the light transmitted by the dynamic image of the molten pool is coaxial with the laser; the computer extracts characteristic information values from a molten pool image shot by the camera through the image acquisition module, the low-pass Gaussian filter module and the contour extraction module to obtain characteristic information for defect identification, wherein the characteristic information is molten pool width, molten pool area, keyhole width and keyhole area, and the defects in the dynamic welding process are accurately identified through a forward neural network;
the method for judging the defects of the welding process through the forward neural network comprises the following steps: an output node is set, a four-dimensional vector form is selected and used for expression according to four corresponding welding states, and four typical welding states are firstly determined: after forward information transmission, solving weight change and error backward propagation by using a gradient descent method, carrying out system programming based on Labview software, directly calling a BP algorithm function in a database of the Labview, carrying out BP algorithm operation on the four characteristic values in combination with weight and a threshold value, outputting an actual output value, and comparing the actual output value with the ideal output value; the actual output value output after the BP algorithm is not always an ideal value, the judgment criterion is that the value with the largest component in the vector is regarded as the value closest to the corresponding welding state, and the welding state is the welding state corresponding to the output quantity; the online identification function of the monitoring system is realized by selecting proper sample data;
the method of learning patterns through the forward neural network is as follows: collecting welding state data in a welding process; and continuously adjusting the weight and the threshold of the network by using a steepest descent method through back propagation to minimize the sum of squares of errors of the network, and accurately judging the welding defects according to the data acquired in real time after the weight and the threshold reach the minimum.
The working principle of the invention is as follows: due to the large dynamic range of the welding area, the strong plasma luminescence at the center of the area can cause the problem of overexposure of the chip of the image sensor, and the image quality is affected. Therefore, the industrial camera selected by the invention needs to have a larger dynamic response range, and the optical filter 4 and the dimmer 5 are arranged to filter light waves in the wavelength range of plasma, so that the camera is positioned in a narrower wave band where interference light is weaker than molten pool radiation to acquire a molten pool image;
the image acquisition module can transmit dynamic image information of a molten pool in the welding process acquired by the CMOS camera 1 to a software system in real time, and the molten pool image is processed frame by frame through embedding of image processing technologies such as low-pass filtering, binarization processing and the like, wherein the processing includes extraction of a molten pool edge profile and identification of a splashing phenomenon outside the molten pool;
the image is denoised by using the low-pass Gaussian filter module, so that sharp noise and some interference signals can be well avoided, a clearer and smoother image can be obtained, and great help is brought to the processing of the subsequent image; the low-pass filtering processing can remove some small spots in the image, and can also make the salient points become fuzzy and smooth, thereby being more beneficial to the observation processing of the image;
the contour extraction module adopts an image binarization processing algorithm to extract the contours of the molten pool and the keyhole after filtering and denoising, so that the characteristic information of the molten pool and the keyhole can be conveniently extracted; the binarization of the image refers to the binarization processing of the pixel value of the collected image, and by setting a pixel threshold, all pixel points higher than the pixel threshold are assigned with a value of 255, all pixel points lower than the pixel threshold are assigned with another value of 0, and the image after the threshold processing has only two pixel values of 0 and 255, namely the image after the binarization processing is a black-and-white image; the gray level image acquired by the industrial camera is transformed in such a way, so that a binary black-and-white image which can still reflect the overall and local characteristics of the image is obtained; in order to more intuitively see the outlines of the molten pool and the keyhole, an edge extraction operation can be carried out on the binary image; after the image is subjected to binarization processing and the edge contour is extracted, the characteristic information can be extracted; the edge of a full-bright pixel point can be identified through an IMAQ Clamp Horizontal Max function, the maximum value of the transverse width is extracted, and the extraction effect can be displayed in a binary image;
after a molten pool image shot by a coaxial CMOS camera is subjected to characteristic information numerical value extraction by a software system, the occurrence of the collapse defect in the welding process can be judged through the change of the width of the molten pool; judging a threshold value of ten continuous pictures, and when the value of six or more pictures is smaller than the threshold value, considering that a collapse defect is generated at the position; the judgment basis avoids information mutation caused by unstable molten pool fluctuation and noise signal interference in the welding process, and the judgment result is more reliable;
after a molten pool image shot by a coaxial CMOS camera is subjected to characteristic information value extraction by a software system, the stability of the welding process can be judged through the variance change of the width of a keyhole; variance is the average of the sum of the squares of the mean and the difference of each value in the data, and may reflect the degree of dispersion between the data; meanwhile, if only a few data in a group of data have large deviation degrees, the deviation degrees can be well reflected through the variance, so that the data can be utilized as a judgment standard for sudden unstable conditions in the welding process; meanwhile, whether splash is generated or not can be judged by identifying the image area, the generation time is recorded in time, and the splash generation position can be recorded and stored by combining the welding process parameters and the time for calculation.
Compared with the prior art, the invention has the following beneficial effects:
firstly, the invention solves the problems of limited space of a laser welding vacuum bin in a vacuum environment, difficult monitoring in the welding process and the like by integrating a coaxial CMOS industrial camera (the light 10 for collecting images is coaxial with laser), realizes the visualization of a molten pool and surface information in the welding process, has simple and easy realization and stable and reliable work, and can be widely applied to a vacuum welding system;
secondly, the invention establishes a set of coaxial vacuum laser welding process monitoring system based on the molten pool coaxial visual sensing technology, judges the collapse defect in the welding process by analyzing the width fluctuation condition of the molten pool in the welding process, and identifies the generation position of the splash defect.
Drawings
FIG. 1 is a schematic diagram of an apparatus for a vacuum laser welding seam defect identification method based on coaxial monitoring according to a first embodiment;
FIG. 2 is a first photograph of a mid-low pass Gaussian filter module after processing;
FIG. 3 is a second photograph of a mid-low pass Gaussian filter module after processing in an experiment;
FIG. 4 is a photograph of an image after binarization processing in experiment one;
FIG. 5 is a graph showing the portion of the molten pool other than the keyhole extracted in trial one as white, and the middle black portion as the keyhole area;
fig. 6 is a photograph of the image after the binarization process and the edge contour extraction.
Detailed Description
The first embodiment is as follows: the embodiment is a coaxial monitoring-based vacuum laser welding seam defect identification method, which is specifically carried out according to the following steps:
placing two base metals to be welded in a vacuum bin, wherein the joints are in butt joint, starting a laser head to weld, transmitting a dynamic image of a molten pool to a lens of a camera along light rays in the welding process, and connecting a data output end of the camera with an input end of a computer; the light transmitted by the dynamic image of the molten pool is coaxial with the laser; the computer extracts characteristic information values from a molten pool image shot by the camera through the image acquisition module, the low-pass Gaussian filter module and the contour extraction module to obtain characteristic information for defect identification, wherein the characteristic information is molten pool width, molten pool area, keyhole width and keyhole area, and the defects in the dynamic welding process are accurately identified through a forward neural network;
the method for judging the defects of the welding process through the forward neural network comprises the following steps: an output node is set, a four-dimensional vector form is selected and used for expression according to four corresponding welding states, and four typical welding states are firstly determined: after forward information transmission, solving weight change and error backward propagation by using a gradient descent method, carrying out system programming based on Labview software, directly calling a BP algorithm function in a database of the Labview, carrying out BP algorithm operation on the four characteristic values in combination with weight and a threshold value, outputting an actual output value, and comparing the actual output value with the ideal output value; the actual output value output after the BP algorithm is not always an ideal value, the judgment criterion is that the value with the largest component in the vector is regarded as the value closest to the corresponding welding state, and the welding state is the welding state corresponding to the output quantity; the online identification function of the monitoring system is realized by selecting proper sample data;
the method of learning patterns through the forward neural network is as follows: collecting welding state data in a welding process; and continuously adjusting the weight and the threshold of the network by using a steepest descent method through back propagation to minimize the sum of squares of errors of the network, and accurately judging the welding defects according to the data acquired in real time after the weight and the threshold reach the minimum.
The second embodiment is as follows: the first difference between the present embodiment and the specific embodiment is: the optical filter 4 is a narrow-band optical filter with the wavelength of 500 +/-10 nm. The rest is the same as the first embodiment.
The third concrete implementation mode: the present embodiment differs from the first or second embodiment in that: the L-shaped pipeline 8 is a plastic pipe. The others are the same as in the first or second embodiment.
The fourth concrete implementation mode: the difference between this embodiment mode and one of the first to third embodiment modes is: the dimmer 5 is an attenuation sheet with an attenuation rate of 10%. The rest is the same as one of the first to third embodiments.
The fifth concrete implementation mode: the fourth difference between this embodiment and the specific embodiment is that: the through hole arranged at the center of the baffle 6 is a round hole. The rest is the same as the fourth embodiment.
The sixth specific implementation mode: the fifth embodiment is different from the fifth embodiment in that: the baffle 6 is circular. The rest is the same as the fifth embodiment.
The seventh embodiment: the first difference between the present embodiment and the specific embodiment is: and the two parent metals to be welded are A5083 aluminum alloy. The rest is the same as the first embodiment.
The specific implementation mode is eight: the first difference between the present embodiment and the specific embodiment is: two base metals to be welded are placed in a vacuum bin, and the joints are in butt joint and form an I-shaped groove. The rest is the same as the first embodiment.
The specific implementation method nine: the eighth embodiment is different from the eighth embodiment in that: the sizes of the two base metals to be welded are both 6mm multiplied by 100mm multiplied by 200 mm. The rest is the same as the embodiment eight.
The detailed implementation mode is ten: the present embodiment differs from the ninth embodiment in that: the welding process parameters are as follows: the laser power is 3kW, the welding speed is 1m/min, and the defocusing amount is-2 m. The rest is the same as in the ninth embodiment.
The invention was verified with the following tests:
test one: the test is a vacuum laser welding seam defect identification method based on coaxial monitoring, as shown in fig. 1, specifically as follows: placing two base metals 11 to be welded in a vacuum bin, wherein the base metals are A5083 aluminum alloy, the plate size is 6mm multiplied by 100mm multiplied by 200mm, the joint form is butt joint and is an I-shaped groove, selecting a YLR-10000 high-power optical fiber laser, starting a laser head 7 to weld, and the technological parameters are laser power 3kW, welding speed 1m/min and defocusing amount-2 m; in the welding process, a dynamic image of a molten pool is transmitted to a spectroscope 2 along a light ray 10, then is refracted to be a horizontal light ray to a reflector 3, then is vertically reflected upwards through the reflector 3, sequentially passes through a central through hole of a baffle 6, a dimmer 5 and a light filter 4 and is reflected to a lens of a CMOS camera 1, so that overlooking images of the molten pool and a key hole are obtained, and the light ray 10 is coaxial with laser; a barrier 6 is installed on the other end of the dimmer sheet 5. The baffle 6 with the holes can protect the CMOS camera 1, reduce the influence caused by secondary images in a molten pool image, improve the image acquisition quality, and simultaneously utilize the optical filter 4 and the light reduction sheet 5 to filter and reduce light so as to improve the definition of a shot molten pool image;
the computer 9 extracts characteristic information values from the molten pool image shot by the CMOS camera 1 through an image acquisition module, a low-pass Gaussian filter module and a contour extraction module to obtain characteristic information for defect identification, wherein the characteristic information comprises molten pool width, molten pool area, keyhole width and keyhole area, and the defects in the dynamic welding process are accurately identified through a forward neural network;
the method for judging the defects of the welding process through the forward neural network comprises the following steps: setting an output node, selecting and using a four-dimensional vector form to express according to four corresponding welding states, firstly determining ideal output quantities of four typical welding states (stable, couch setting, welding deviation and unstable process), solving weight variation and error back propagation by using a gradient descent method after forward information transfer, programming a system based on Labview software, directly calling a BP algorithm function in a database of the Labview, carrying out BP algorithm operation on the four characteristic values by combining weights and thresholds, outputting an actual output value, and comparing the actual output value with the ideal output value. The actual output value output after the BP algorithm is not always an ideal value, and the judgment criterion is that the value with the largest component in the vector is regarded as the value closest to the corresponding welding state, so that the welding state is the welding state corresponding to the output quantity. And the online identification function of the monitoring system is realized by selecting proper sample data.
The method of learning patterns through the forward neural network is as follows: collecting welding state data in a welding process; and continuously adjusting the weight and the threshold of the network by using a steepest descent method through back propagation to minimize the sum of squares of errors of the network, and accurately judging the welding defects according to the data acquired in real time after the weight and the threshold reach the minimum.
The method for judging the collapse defect in the welding process through the forward neural network comprises the following steps: firstly determining an ideal output quantity corresponding to the generation of the yield defect, then carrying out BP (back propagation) operation on four kinds of characteristic information including the width of a molten pool, the area of the molten pool, the width of a key hole and the area of the key hole, and considering that the yield defect is generated at the position when the output value is consistent with the ideal output value or the position of a corresponding value with the maximum corresponding component in the output quantity is consistent with the position of the maximum value in the yield ideal value;
the method for judging the welding deviation defect in the welding process through the forward neural network comprises the following steps: firstly determining an ideal output quantity corresponding to the generation of the welding deviation defect, then carrying out BP (back propagation) operation on four kinds of characteristic information including a molten pool width, a molten pool area, a keyhole width and a keyhole area, and considering that the welding deviation defect is generated at the position when the output value is consistent with the ideal output value or the position of a corresponding value with the maximum corresponding component in the output quantity is consistent with the position of the maximum value in the welding deviation ideal value;
the method for judging the unstable defect of the welding process through the forward neural network comprises the following steps: firstly determining an ideal output quantity corresponding to the generation of the unstable defect, then carrying out BP (back propagation) operation on four kinds of characteristic information including a molten pool width, a molten pool area, a keyhole width and a keyhole area, and considering that the unstable defect is generated at the position when the output value is consistent with the ideal output value or the position of a corresponding value with the maximum corresponding component in the output quantity is consistent with the position of the maximum value in the unstable ideal value;
the method for judging the stability of the welding process through the forward neural network comprises the following steps: firstly determining the corresponding ideal output quantity during stabilization, then carrying out BP operation on four kinds of characteristic information of the width of a molten pool, the area of the molten pool, the width of a keyhole and the area of the keyhole, and considering that the stabilization defect is generated at the position when the output value is consistent with the ideal output value or the position of the corresponding value with the maximum corresponding component in the output quantity is consistent with the position of the maximum value in the stable ideal value. Table 1 shows ideal output data for typical weld conditions.
TABLE 1
The test is a device used in the vacuum laser welding seam defect identification method based on coaxial monitoring, and as shown in figure 1, the test specifically comprises a CMOS camera 1, a spectroscope 2, a reflector 3, an optical filter 4, a light reduction sheet 5, a baffle 6, a laser head 7, an L-shaped pipeline 8 and a computer 9; the optical filter 4 is a narrow-band optical filter with the wavelength of 500 +/-10 nm; the dimmer 5 is an attenuation sheet with an attenuation rate of 10 percent; the through hole arranged at the center of the baffle 6 is a round hole; the baffle 6 is circular; the spectroscope 2 is fixed in the laser head 7 through a bracket; the CMOS camera 1 is an XVC-1000 model CMOS camera produced by XIRIS company of Canada;
the spectroscope 2 is arranged in the laser head 7, the spectroscope 2 is arranged on a laser transmission line, and the spectroscope 2 and the horizontal plane form an angle of 45 degrees; a through hole is formed in the side wall of the laser head 7, the through hole is equal in height to the spectroscope 2, the reflecting surface of the spectroscope 2 faces the through hole, an opening at one end of the L-shaped pipeline 8 is horizontally arranged and fixed on the through hole in the side wall of the laser head 7, an opening at the other end of the L-shaped pipeline 8 is vertically arranged upwards, the reflector 3 is fixed on the inner wall of the corner of the L-shaped pipeline 8, and the reflector 3 and the horizontal surface form an angle of 45 degrees;
a through hole is formed in the center of the baffle 6, the baffle 6 is horizontally fixed on an upward vertical opening of the L-shaped pipeline 8, the dimmer 5 is horizontally fixed on the baffle 6, the optical filter 4 is horizontally fixed on the dimmer 5, and the lens of the CMOS camera 1 faces the upper surface of the optical filter 4; the axes of the baffle 6, the optical filter 4 and the light reduction sheet 5 are superposed; the data output end of the CMOS camera 1 is connected with the data input end of the computer 9; and the computer 9 is provided with an image acquisition module, a low-pass Gaussian filter module and a contour extraction module.
The working principle of the invention is as follows: due to the large dynamic range of the welding area, the strong plasma luminescence at the center of the area can cause the problem of overexposure of the chip of the image sensor, and the image quality is affected. Therefore, the industrial camera selected by the invention needs to have a larger dynamic response range, and the optical filter 4 and the dimmer 5 are arranged to filter light waves in the wavelength range of plasma, so that the camera is positioned in a narrower wave band where interference light is weaker than molten pool radiation to acquire a molten pool image;
the image acquisition module can transmit dynamic image information of a molten pool in the welding process acquired by the CMOS camera 1 to a software system in real time, and the molten pool image is processed frame by frame through embedding of image processing technologies such as low-pass filtering, binarization processing and the like, wherein the processing includes extraction of a molten pool edge profile and identification of a splashing phenomenon outside the molten pool;
the image is denoised by using the low-pass Gaussian filter module, so that sharp noise and some interference signals can be well avoided, a clearer and smoother image can be obtained, and great help is brought to the processing of the subsequent image; the low-pass filtering process can remove some small spots in the image, and can also blur and smooth the salient points, which is more helpful for the image observation process, as shown in fig. 2 and 3;
the contour extraction module adopts an image binarization processing algorithm to extract the contours of the molten pool and the keyhole after filtering and denoising, so that the characteristic information of the molten pool and the keyhole can be conveniently extracted; the binarization of the image refers to the binarization processing of the pixel value of the collected image, and by setting a pixel threshold, all pixel points higher than the pixel threshold are assigned with a value of 255, all pixel points lower than the pixel threshold are assigned with another value of 0, and the image after the threshold processing has only two pixel values of 0 and 255, namely the image after the binarization processing is a black-and-white image; the gray level image acquired by the industrial camera is transformed in such a way, so that a binary black-and-white image which can still reflect the overall and local characteristics of the image is obtained; thus, referring to FIGS. 4 and 5, the experiment sets two thresholds for simultaneously extracting the melt pool and keyhole profiles, with the portions of the melt pool other than the keyhole extracted to appear white and the middle black portion to appear as the keyhole area. As shown in fig. 6, after binarization processing, in order to more intuitively see the outlines of the molten pool and the keyhole, an edge extraction operation can be performed on the binarized image; after the image is subjected to binarization processing and the edge contour is extracted, the characteristic information can be extracted; the edge of a full-bright pixel point can be identified through an IMAQ Clamp Horizontal Max function, the maximum value of the transverse width is extracted, and the extraction effect can be displayed in a binary image;
after a molten pool image shot by a coaxial CMOS camera is subjected to characteristic information numerical value extraction by a software system, the occurrence of the collapse defect in the welding process can be judged through the change of the width of the molten pool; judging a threshold value of ten continuous pictures, and when the value of six or more pictures is smaller than the threshold value, considering that a collapse defect is generated at the position; the judgment basis avoids information mutation caused by unstable molten pool fluctuation and noise signal interference in the welding process, and the judgment result is more reliable;
after a molten pool image shot by a coaxial CMOS camera is subjected to characteristic information value extraction by a software system, the stability of the welding process can be judged through the variance change of the width of a keyhole; variance is the average of the sum of the squares of the mean and the difference of each value in the data, and may reflect the degree of dispersion between the data; meanwhile, if only a few data in a group of data have large deviation degrees, the deviation degrees can be well reflected through the variance, so that the data can be utilized as a judgment standard for sudden unstable conditions in the welding process; meanwhile, whether splash is generated or not can be judged by identifying the image area, the generation time is recorded in time, and the splash generation position can be recorded and stored by combining the welding process parameters and the time for calculation.
In summary, the invention provides a vacuum laser welding process monitoring system based on coaxial image information of a molten pool, which is combined with an industrial camera and a software system, can perform online real-time monitoring on welding parameters, analyze and process monitoring data, extract and analyze geometric characteristic information of a molten pool image, and compare the geometric characteristic information with data of a defect database to judge the quality of a weld joint. The invention can quickly and effectively monitor the laser welding process and evaluate the welding quality.
Claims (10)
1. A vacuum laser welding seam defect identification method based on coaxial monitoring is characterized in that the vacuum laser welding seam defect identification method based on coaxial monitoring is as follows: placing two base metals to be welded in a vacuum bin, wherein the joints are in butt joint, starting a laser head to weld, transmitting a dynamic image of a molten pool to a lens of a camera along light rays in the welding process, and connecting a data output end of the camera with an input end of a computer; the light transmitted by the dynamic image of the molten pool is coaxial with the laser; the computer extracts characteristic information values from a molten pool image shot by the camera through the image acquisition module, the low-pass Gaussian filter module and the contour extraction module to obtain characteristic information for defect identification, wherein the characteristic information is molten pool width, molten pool area, keyhole width and keyhole area, and the defects in the dynamic welding process are accurately identified through a forward neural network;
the method for judging the defects of the welding process through the forward neural network comprises the following steps: an output node is set, a four-dimensional vector form is selected and used for expression according to four corresponding welding states, and four typical welding states are firstly determined: after forward information transmission, solving weight change and error backward propagation by using a gradient descent method, carrying out system programming based on Labview software, directly calling a BP algorithm function in a database of the Labview, carrying out BP algorithm operation on the four characteristic values in combination with weight and a threshold value, outputting an actual output value, and comparing the actual output value with the ideal output value; outputting an actual output value after the BP algorithm, and regarding a judgment criterion that the vector with the largest component is the value closest to the corresponding welding state, wherein the welding state is the welding state corresponding to the output quantity; the online identification function of the monitoring system is realized by selecting proper sample data;
the method of learning patterns through the forward neural network is as follows: collecting welding state data in a welding process; and continuously adjusting the weight and the threshold of the network by using a steepest descent method through back propagation to minimize the sum of squares of errors of the network, and accurately judging the welding defects according to the data acquired in real time after the weight and the threshold reach the minimum.
2. The method for identifying the defect of the weld joint in the vacuum laser welding based on the coaxial monitoring as claimed in claim 1, wherein the two parent metals to be welded are both A5083 aluminum alloy.
3. The method for identifying the weld defect of the vacuum laser welding based on the coaxial monitoring is characterized in that two base metals to be welded are placed in a vacuum chamber, and the joint is in a butt joint mode and is an I-shaped groove.
4. The method for identifying the weld defect in the vacuum laser welding based on the coaxial monitoring as claimed in claim 1, wherein the camera is a CMOS camera.
5. The method for identifying the weld defect in the vacuum laser welding based on the coaxial monitoring as claimed in claim 1, wherein the two parent metals to be welded have the size of 6mm x 100mm x 200 mm.
6. The method for identifying the weld defect in vacuum laser welding based on coaxial monitoring as claimed in claim 5, wherein the two base metals to be welded are pretreated before welding, specifically, degreasing and oxide film removing are performed, and welding is performed within 24 hours after pretreatment.
7. The method for identifying the defects of the vacuum laser welding seam based on the coaxial monitoring as claimed in claim 6, wherein the specific method for removing the greasy dirt and the oxidation film is as follows: firstly, placing an aluminum alloy plate into a sodium hydroxide aqueous solution with the mass fraction of 8% -10% and the temperature of 40-60 ℃ for corrosion, keeping the corrosion for 5-10 min, then taking out the aluminum alloy plate, then placing the plate into cold water for washing for 2-3 min, then placing the plate into a nitric acid aqueous solution with the mass fraction of 30% for photochemical treatment to neutralize residual alkali, finally washing for 2-3 min by flowing water, storing the plate in a baking box at 100 ℃ for baking after washing, taking the plate out of the baking box at any time in 10min before welding, and wiping the part to be welded by acetone.
8. The method for identifying the weld defect of the vacuum laser welding based on the coaxial monitoring as claimed in claim 1, wherein the welding process parameters are as follows: the laser power is 3kW, and the welding speed is 1 m/min.
9. The method for identifying the weld defect in the vacuum laser welding based on the coaxial monitoring as claimed in claim 4, wherein the CMOS camera is an XVC-1000 model CMOS camera manufactured by XIRIS, Canada.
10. The method for identifying the weld defect in the vacuum laser welding based on the coaxial monitoring as claimed in claim 1, wherein the defocusing amount of the welding is-2 m.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011400726.6A CN112548321A (en) | 2020-12-04 | 2020-12-04 | Coaxial monitoring-based vacuum laser welding seam defect identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011400726.6A CN112548321A (en) | 2020-12-04 | 2020-12-04 | Coaxial monitoring-based vacuum laser welding seam defect identification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112548321A true CN112548321A (en) | 2021-03-26 |
Family
ID=75048101
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011400726.6A Pending CN112548321A (en) | 2020-12-04 | 2020-12-04 | Coaxial monitoring-based vacuum laser welding seam defect identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112548321A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113210852A (en) * | 2021-04-19 | 2021-08-06 | 北京工业大学 | Laser welding process monitoring method based on small-orifice diameter distortion rate in-situ measurement |
| CN113566711A (en) * | 2021-09-23 | 2021-10-29 | 四川国芯通智能科技有限公司 | Method for determining repair welding position |
| CN113838050A (en) * | 2021-11-23 | 2021-12-24 | 昆山宝锦激光拼焊有限公司 | Weld forming prediction method based on complementary two-channel convolution neural network |
| CN114147376A (en) * | 2021-12-13 | 2022-03-08 | 上海工程技术大学 | Laser welding weldment fixing device with infrared ranging |
| DE102022107886A1 (en) | 2022-04-01 | 2023-10-05 | Gehring Technologies Gmbh + Co. Kg | Method and device for evaluating the welding quality of a weld seam between end sections of two conductor elements |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102189331A (en) * | 2010-03-05 | 2011-09-21 | 奥林巴斯株式会社 | Defect correction device and defect tracking method |
| CN102424971A (en) * | 2011-12-21 | 2012-04-25 | 西安铂力特激光成形技术有限公司 | Rapid laser repair method and device for defect of aluminum alloy guide blade |
| CN103506756A (en) * | 2013-09-11 | 2014-01-15 | 上海交通大学 | Laser lap welding gap detecting system and laser lap welding gap detecting method based on molten pool image visual sensing |
| CN103692119A (en) * | 2013-12-17 | 2014-04-02 | 南京理工大学 | Dynamic monitoring device of electronic beam deep penetration welding molten pool based on vision sensing |
| CN105458490A (en) * | 2015-12-30 | 2016-04-06 | 哈尔滨工业大学 | Real-time monitoring system and monitoring method for judging welding type of laser deep penetration welding by using high-speed camera |
| CN109175608A (en) * | 2018-09-30 | 2019-01-11 | 华南理工大学 | Weld bead feature points position On-line Measuring Method and seam track automatic measurement system |
| WO2019014937A1 (en) * | 2017-07-21 | 2019-01-24 | 易思维 (天津) 科技有限公司 | Method and apparatus for detecting defect of laser welding protection lens on automobile production line |
| CN109843497A (en) * | 2016-10-14 | 2019-06-04 | 浜松光子学株式会社 | Laser processing device and method for confirming operation |
| CN111230299A (en) * | 2020-03-12 | 2020-06-05 | 广东工业大学 | Laser welding online detection device and method |
| CN111775154A (en) * | 2020-07-20 | 2020-10-16 | 广东拓斯达科技股份有限公司 | Robot vision system |
-
2020
- 2020-12-04 CN CN202011400726.6A patent/CN112548321A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102189331A (en) * | 2010-03-05 | 2011-09-21 | 奥林巴斯株式会社 | Defect correction device and defect tracking method |
| CN102424971A (en) * | 2011-12-21 | 2012-04-25 | 西安铂力特激光成形技术有限公司 | Rapid laser repair method and device for defect of aluminum alloy guide blade |
| CN103506756A (en) * | 2013-09-11 | 2014-01-15 | 上海交通大学 | Laser lap welding gap detecting system and laser lap welding gap detecting method based on molten pool image visual sensing |
| CN103692119A (en) * | 2013-12-17 | 2014-04-02 | 南京理工大学 | Dynamic monitoring device of electronic beam deep penetration welding molten pool based on vision sensing |
| CN105458490A (en) * | 2015-12-30 | 2016-04-06 | 哈尔滨工业大学 | Real-time monitoring system and monitoring method for judging welding type of laser deep penetration welding by using high-speed camera |
| CN109843497A (en) * | 2016-10-14 | 2019-06-04 | 浜松光子学株式会社 | Laser processing device and method for confirming operation |
| WO2019014937A1 (en) * | 2017-07-21 | 2019-01-24 | 易思维 (天津) 科技有限公司 | Method and apparatus for detecting defect of laser welding protection lens on automobile production line |
| CN109175608A (en) * | 2018-09-30 | 2019-01-11 | 华南理工大学 | Weld bead feature points position On-line Measuring Method and seam track automatic measurement system |
| CN111230299A (en) * | 2020-03-12 | 2020-06-05 | 广东工业大学 | Laser welding online detection device and method |
| CN111775154A (en) * | 2020-07-20 | 2020-10-16 | 广东拓斯达科技股份有限公司 | Robot vision system |
Non-Patent Citations (5)
| Title |
|---|
| 宫建锋等: "基于同轴图像传感的激光焊接过程质量监测技术", 《焊接学报》 * |
| 李亚江等: "《特种连接技术》", 31 January 2020 * |
| 蔡创等: "真空激光焊接技术研究现状与展望", 《光电工程》 * |
| 辽宁省安全科学研究院: "《特种设备基础知识》", 30 April 2017 * |
| 雷仕湛等: "《激光智能制造技术》", 29 June 2018 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113210852A (en) * | 2021-04-19 | 2021-08-06 | 北京工业大学 | Laser welding process monitoring method based on small-orifice diameter distortion rate in-situ measurement |
| CN113566711A (en) * | 2021-09-23 | 2021-10-29 | 四川国芯通智能科技有限公司 | Method for determining repair welding position |
| CN113566711B (en) * | 2021-09-23 | 2021-12-07 | 四川国芯通智能科技有限公司 | Method for determining repair welding position |
| CN113838050A (en) * | 2021-11-23 | 2021-12-24 | 昆山宝锦激光拼焊有限公司 | Weld forming prediction method based on complementary two-channel convolution neural network |
| CN113838050B (en) * | 2021-11-23 | 2022-06-03 | 昆山宝锦激光拼焊有限公司 | Weld forming prediction method based on complementary two-channel convolution neural network |
| CN114147376A (en) * | 2021-12-13 | 2022-03-08 | 上海工程技术大学 | Laser welding weldment fixing device with infrared ranging |
| DE102022107886A1 (en) | 2022-04-01 | 2023-10-05 | Gehring Technologies Gmbh + Co. Kg | Method and device for evaluating the welding quality of a weld seam between end sections of two conductor elements |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112548321A (en) | Coaxial monitoring-based vacuum laser welding seam defect identification method | |
| CN111650204B (en) | Power transmission line hardware defect detection method and system based on cascade target detection | |
| CN107402220A (en) | A kind of selective laser fusing shaping powdering quality vision online test method and system | |
| CN106442543A (en) | Detection method for identifying surface quality of continuous laser seam of metal workpiece online | |
| CN108931535A (en) | A kind of laser gain material manufacture gas hole defect on-line monitoring method | |
| Chu et al. | A study on welding quality inspection system for shell-tube heat exchanger based on machine vision | |
| CN108067714B (en) | Online monitoring and defect positioning system and method for end connection quality of thin-wall circular seam | |
| CN110405384A (en) | Method and system for real-time detection insufficient (incomplete) penetration during robot welding | |
| Jiang et al. | Image processing of aluminum alloy weld pool for robotic VPPAW based on visual sensing | |
| CN112730432A (en) | Laser welding defect detection equipment and detection method for lithium battery of mobile phone | |
| CN110717909A (en) | Metal surface scratch detection method and device | |
| CN114119595A (en) | GMAW welding quality on-line monitoring and evaluating method based on integrated deep learning | |
| CN108152291B (en) | End seam welding unfused defect real-time detection method based on dynamic characteristics of molten pool | |
| CN114627122A (en) | Defect detection method and device | |
| CN112529103A (en) | Fusion penetration identification method based on bidirectional molten pool geometric and textural feature fusion | |
| CN116664508A (en) | Weld surface quality detection method and computer readable storage medium | |
| CN111307817A (en) | Online detection method and system for PCB production process of intelligent production line | |
| CN115266743B (en) | System and method for evaluating chip quality under nondestructive testing | |
| CN114789307B (en) | Plate welding quality real-time monitoring method based on digital twinning | |
| CN110405383A (en) | Weld blowhole real-time detection method and system for robot welding technique | |
| CN114862777A (en) | Connecting sheet welding detection method and system | |
| CN116109587A (en) | Sealing nail welding seam quality detection method in complex environment | |
| CN112129774A (en) | Online detection method for welding unfused defect | |
| Fei et al. | Machine Vision Analysis of Welding Region and Its Application to Seam Tracking in Arc Welding | |
| CN113112458A (en) | Metal surface defect detection method based on support vector machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210326 |
|
| RJ01 | Rejection of invention patent application after publication |