Online detection method and device for laser welding seam and laser welding device

Abstract

The invention discloses a laser welding seam on-line detection method and device and a laser welding device.The detection method comprises the steps that after laser welding is completed, a laser plane is projected to the surface of a welding seam, modulated laser stripes are formed, and a laser stripe image is collected; a Gaussian fitting method is adopted to extract sub-pixel coordinates of the centers of all columns of stripes in the laser stripe image, and a center line of the section contour of the welding seam is generated; identifying feature endpoints on the center line; based on the feature endpoints and the center lines, the convexity and the concavity of the surface of the weld joint are calculated; and judging whether the weld quality is qualified or not according to the convexity and the concavity of the weld surface. The method detects the surface characteristics of the post-welding weld through a light vision method, can directly detect the size and shape of a molten pool on the surface of the weld, surface pores and other defects, and has the obvious advantages of rapidness, real-time performance, high detection precision and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of welding, in particular to a laser weld seam on-line detection method, device and laser welding device. BACKGROUND

[0002] In the field of antenna manufacturing, the manufacturing precision and reliability of the antenna panel, the core component, directly determine the overall electrical performance of the antenna. The surface of the antenna panel is usually provided with a complex and precise feed network groove. A key process in its production process is to accurately apply a special metal solder of a specific form (such as a sheet or a wire) into the groove, and to reliably fix it through a series of spot welding operations, laying a foundation for subsequent overall brazing or electroplating processes.

[0003] Currently, the welding of the antenna panel generally relies on manual operation mode. That is, the operator holds a welding gun (usually a micro resistance welding gun or a laser welding gun) and performs spot welding one by one at the key positions of the solder sheet through visual observation. This traditional manual operation mode has the problems of low efficiency and unguaranteed welding quality.

[0004] In recent years, although attempts have been made to replace part of the manual work with numerical control machine tools or simple automated equipment, most of them focus on simple path repetition. In terms of weld seam quality detection, complete reliance is placed on offline sampling inspection or manual re-inspection after welding, and automatic quality judgment for each weld cannot be achieved. This makes it impossible to fundamentally guarantee the reliability of automated welding, and it is difficult to be widely applied in antenna production which requires high quality. SUMMARY

[0005] In view of the above problems in the prior art, the present application provides a laser weld seam on-line detection method which can automatically detect the quality of the weld seam, improve the reliability of automated welding, and meet the production requirements of the antenna panel.

[0006] In addition, the present application also provides a device for implementing the laser weld seam on-line detection method, and a laser welding device for automatically detecting the quality of the weld seam using the laser weld seam on-line detection method.

[0007] To solve the above technical problems, the present application adopts the following scheme: A laser weld seam on-line detection method, comprising the following steps: After laser welding is completed, a laser plane is projected onto the weld seam surface to form a modulated laser stripe, and a laser stripe image containing the laser stripe is collected; The sub-pixel coordinates of the center of each column of stripes in the laser stripe image are extracted by using a Gaussian fitting method to generate a center line of the cross section profile of the weld seam; Identify the feature end points on the center line; Based on the feature end point and the center line, convexity and concavity of the weld surface are calculated; According to the convexity and concavity of the weld surface, whether the weld quality is qualified is determined.

[0008] Preferably, the Gaussian fitting method is used to extract the sub-pixel coordinates of the center of the laser stripe image, and the center line of the weld cross-section profile is generated, including: The gray maximum value is searched along the normal direction of the laser stripe image column by column, and the gray maximum value is used as the threshold for binary processing to obtain a binary image; The two edge points of the stripe are extracted from the binary image column by column, the midpoint of the two edge points is taken as the initial center point of the stripe of the column, and a single-pixel-wide center line image is generated; For each column of the single-pixel-wide center line image, the pixel row and its gray value within the preset width range are selected with the initial center point of the stripe of the column as the center, and Gaussian curve fitting is performed, and the peak point horizontal coordinate of the fitted Gaussian curve is taken as the sub-pixel coordinates of the center of the stripe of the column.

[0009] Preferably, the two edge points of the stripe are extracted from the binary image column by column, the midpoint of the two edge points is taken as the initial center point of the stripe of the column, and a single-pixel-wide center line image is generated, including: For each column, the upper edge row coordinate and the lower edge row coordinate of the binary image are located, the arithmetic mean of the upper edge row coordinate and the lower edge row coordinate is calculated as the initial center point of the stripe of the binary image in the column, the pixel gray value of the initial center point of the stripe is assigned as zero, and the pixel gray value of other points in the column is assigned as 1 to obtain a single-pixel-wide center line image.

[0010] Preferably, the feature end point on the center line is identified, including: The center line of the extracted weld cross-section profile is divided into a first straight line segment, a top curve segment and a second straight line segment; The straight line equations of the first straight line segment and the second straight line segment and the quadratic curve equation of the top curve segment are fitted by the least square method respectively; The coordinates of the first feature end point of the weld are calculated by simultaneously solving the straight line equation of the first straight line segment and the quadratic curve equation of the top curve segment, and the coordinates of the second feature end point of the weld are calculated by simultaneously solving the straight line equation of the second straight line segment and the quadratic curve equation of the top curve segment.

[0011] Preferably, the straight line equations of the first straight line segment and the second straight line segment and the quadratic curve equation of the top curve segment are fitted by the least square method respectively, including: The straight line equation of the first straight line segment is: a 1x + b 1 y + c 1 =0 ; The linear equation of the second straight line segment is: a 2 x + b 2 y + c 2 =0 ; The quadratic curve equation of the top curve segment is: y = a 3 x 2 +b 3 x + c 3 ; In the formula: a 1 、 b 1 、 c 1 、 a 2 、 b 2 、 c 2 、 a 3 、 b 3 、 c 3 are coefficients.

[0012] Preferably, based on the feature end point and the center line, the convexity and the concavity of the weld surface are calculated, comprising: connecting the first feature end point and the second feature end point to obtain a reference straight line; searching all pixel points between the first feature end point and the second feature end point on the center line; respectively calculating the vertical distance of each pixel point to the reference straight line, if the pixel point is above the reference straight line, it is recorded as a positive value, if the pixel point is below the reference straight line, it is recorded as a negative value; selecting the maximum value from all positive value distances as the convexity of the weld, and selecting the minimum value from all negative value distances as the concavity of the weld.

[0013] Preferably, based on the feature end point and the center line, the convexity and the concavity of the weld surface are calculated, comprising: The equation of the reference straight line is: ax + by + c = 0 ; The vertical distance of the pixel point ( x i ，y i ) to the reference straight line d(i) The calculation formula is:

[0014] In the formula: a , b , c are all coefficients.

[0015] Preferably, the convexity and concavity of the weld surface are determined to determine whether the weld quality is qualified, comprising: The calculated convexity and concavity values of the weld are converted into actual physical lengths through preset calibration parameters; The actual physical length is compared with the preset process requirement depth interval to determine whether the weld quality is qualified.

[0016] A laser weld online detection device for realizing the above-mentioned laser weld online detection method, comprising: A laser light source for projecting a laser plane to the weld surface to form a modulated laser stripe; An image collector for collecting a laser stripe image containing the laser stripe; A host computer, which is in communication connection with the laser light source and the image collector respectively, and is configured to: Receive the laser stripe image, extract the sub-pixel coordinates of the center of each column of stripes in the laser stripe image using Gaussian fitting method, and generate the center line of the weld cross-section profile; Identify the feature endpoints on the center line; Calculate the convexity and concavity of the weld surface based on the feature endpoints and the center line; Determine whether the weld quality is qualified according to the convexity and concavity of the weld surface.

[0017] A laser welding device, comprising: A reference platform; A workpiece fixing tool installed on the reference platform for fixing the workpiece to be processed; A three-axis motion module arranged above the reference platform; A motion controller for driving the three-axis motion module to move; A laser light path module installed on the three-axis motion module for emitting processing laser; A global camera is installed on the three-axis motion module to take a global image of the workpiece to be processed to identify the welding point; And the laser weld on-line detection device, the upper computer is respectively connected with the motion controller, laser light path module and global camera, the upper computer is also configured to: Based on the global image of the workpiece to be processed, the three-axis motion module drives the laser light path module to move to the welding point of the workpiece to be processed for welding through the motion controller.

[0018] Compared with the prior art, the present application has the following advantages: 1、After the laser welding is completed, the laser plane is projected to the weld surface to form the modulated laser stripe, and the laser stripe image containing the laser stripe is collected; Then the sub-pixel coordinates of the center of each column of the laser stripe image are extracted by using the Gaussian fitting method to generate the center line of the weld cross-section profile; The characteristic end points on the center line are identified; Based on the characteristic end points and the center line, the convexity and concavity of the weld surface are calculated; Finally, whether the weld quality is qualified is determined according to the convexity and concavity of the weld surface. Therefore, the whole process of weld quality determination of the present application is based on mathematical calculation and is completed automatically, completely eliminating the subjectivity, fatigue and inconsistency of manual visual inspection, making the determination result objective, reliable and traceable, improving the reliability of automatic welding, and meeting the production requirements of the antenna panel.

[0019] 2、The present application accurately positions the weld boundary by using the endpoint identification method based on the geometric model (intersection of straight line and quadratic curve fitting), and establishes a reliable reference for subsequent measurement. At the same time, by defining two core quantitative indicators of "convexity" and "concavity", the maximum positive and negative deviation of the actual profile to the ideal reference line is calculated, and the convex and concave defects of the weld are accurately and intuitively described, so that key appearance defects such as weld edge overlap, collapse and excessive excess height can be found in real time, and unqualified products can be avoided to flow into the next process.

[0020] 3、Laser welding has the characteristics of large weld depth-width ratio, narrow heat-affected zone, etc. due to small spot size and concentrated energy, and is prone to edge overlap defects. The present application detects the weld surface features after welding by optical vision method, which can directly detect the size and shape of the weld pool and surface porosity and other defects on the weld surface, and has obvious advantages of rapidity, real-time, high detection accuracy, etc. BRIEF DESCRIPTION OF DRAWINGS

[0021] FIG. 1 is a schematic diagram of the laser weld on-line detection device of the present application; Figure 1 FIG. 2 is a flow chart of the laser weld on-line detection method of the present application; FIG. 3 is a schematic diagram of the laser weld on-line detection device of the present application; Figure 2The flow chart of extracting sub-pixel coordinates of each column of stripe center in the laser stripe image by using Gaussian fitting method in the laser weld seam on-line detection method of the application, and generating the center line of the weld seam cross section profile; Figure 2 is a schematic diagram of the laser weld seam on-line detection method of the application. Figure 3 Figure 3 is a flow chart of identifying the feature end points on the center line in the laser weld seam on-line detection method of the application. Figure 4 is a schematic diagram of the laser weld seam on-line detection method of the application. Figure 4 Figure 5 is a profile feature map of the laser weld seam end face in the laser weld seam on-line detection method of the application. Figure 6 is a schematic diagram of the laser weld seam on-line detection method of the application. Figure 5 Figure 7 is a flow chart of calculating the convexity and concavity of the weld seam surface based on the feature end points and the center line in the laser weld seam on-line detection method of the application. Figure 8 is a schematic diagram of the laser weld seam on-line detection method of the application. Figure 6 Figure 9 is a structural schematic diagram of the laser welding device adopting the laser weld seam on-line detection method of the application. Figure 10 is a partial structural schematic diagram of the laser light path module in the laser welding device adopting the laser weld seam on-line detection method of the application. Figure 7 Figure 11 is a schematic diagram of the laser weld seam on-line detection method of the application.

[0022] Figure 12 is a schematic diagram of the laser weld seam on-line detection method of the application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some embodiments of the application, rather than all the embodiments. The components of the embodiments of the application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the application provided in the drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the application.

[0024] It is to be understood that the terms "first", "second", and the like used in the description and the claims of the present application as well as the above-described drawings do not necessarily have to describe particular sequences or order of steps, but can be used to distinguish similar objects from each other. It is to be understood that the data thus used can be interchanged, where appropriate, so that the embodiments of the application described herein can be carried out in other sequences than those illustrated or described herein. Furthermore, the terms "comprising", "having" and any of their derivatives, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a list of steps or units is not necessarily limited to those steps or units that are clearly listed, but can include other steps or units not clearly listed or inherent to such processes, methods, products, or apparatuses.

[0025] Furthermore, in describing representative embodiments, the specification can have presented the method and / or process as a particular sequence of steps. However, to the extent that the method or process depends on more than one step, the method or process should not be limited to the particular sequence of steps described, and as will be appreciated by the skilled artisan, other sequences of steps can be possible. The particular sequence of steps described in the specification is not an limitation on the order in which such steps are performed. Furthermore, the steps recited in the specification are not to be interpreted as being necessarily performed in the order in which they are recited.

[0026] First, in the present embodiment, a laser weld seam on-line detection method is provided, as shown in the accompanying drawings, comprising the following steps: Figure 1 The steps are as follows: Step S1) After laser welding is completed, a laser plane is projected to the weld seam surface to form a modulated laser stripe, and a laser stripe image containing the laser stripe is collected.

[0027] Specifically, the light plane projected by the laser light source is modulated to form a laser stripe on the weld seam surface, and then a laser stripe image containing the laser stripe is collected. By analyzing the characteristics of the laser stripe image, the welding quality of the weld seam can be determined.

[0028] In specific operation, in order to avoid the interference of background and strong light of laser welding, a light reduction lens can be installed in front of the image collector lens for collecting the laser stripe image. In order to complete the weld seam quality detection task, a visual detection model based on perspective projection transformation needs to be established and the related parameters need to be calibrated.

[0029] The weld quality of laser welding, such as concave-convex and undercut, is essentially the change of three-dimensional surface microscopic geometry. The traditional two-dimensional vision cannot directly measure the height information. The structured light technology is an active optical three-dimensional measurement method. A precise laser plane is like a "light ruler". When it irradiates to the surface of the object, the originally straight "light ruler" line will be distorted and deformed due to the ups and downs of the surface. When the laser is projected onto a flat surface, the stripe is straight; when projected onto a convex weld, the stripe at the corresponding position will be bent upwards; when projected onto a concave undercut, the stripe will be bent downwards. The deformation (the offset relative to the original straight line) directly encodes the height information of the point. The image collector captures the deformed stripe image which carries the three-dimensional profile information of the whole weld. Compared with the ordinary illumination image which directly captures the weld, the laser stripe in the image is a high-contrast, bright line with concentrated brightness, which is significantly different from the background, greatly facilitating the recognition and tracking of computer vision algorithms. At the same time, by using a specific wavelength of laser and a light reduction filter in front of the camera, the environmental light, metal reflection and residual heat radiation after welding in the welding site can be effectively suppressed, ensuring that the obtained image signal is clear and stable. In addition, this continuous stripe provides a perfect processing object for the center line extraction algorithm (such as Gaussian fitting method). The algorithm can systematically analyze along the stripe direction (usually column by column), thereby outputting an ordered and coordinate continuous weld cross-section profile line.

[0030] Step S2) The sub-pixel coordinates of the center of each column of the laser stripe image are extracted by using the Gaussian fitting method to generate the center line of the weld cross-section profile.

[0031] This step uses the Gaussian method to extract the center image of the laser stripe, which mainly uses the characteristic that the intensity distribution of the stripe approximately obeys the Gaussian distribution, fits the Gaussian curve in the normal direction of the stripe, and obtains the extreme value point position as the center of the stripe.

[0032] Specifically, as shown in the accompanying drawings, Figure 2 this step includes the following steps: Step S21) Search for the gray maximum value along the normal direction of the laser stripe image column by column, and perform binaryzation processing with the gray maximum value as the threshold to obtain a binary image.

[0033] Specifically, let the laser stripe image function be f(m, n) , where m , n are the row and column of the laser stripe image, respectively; f (m, n) is the gray value of the laser stripe image pixel at m, n , search for the gray maximum value along the normal direction of the stripe column by column fmax , this fmaxThis represents the brightest intensity that the laser stripe can reach at that column position. Since the laser stripe is much brighter than the background, the pixel grayscale value near the center area of ​​each column is theoretically the maximum value of that column. This step is to dynamically obtain the brightness threshold of each column.

[0034] Then, with fmax The image is obtained by binarizing the weld seam image with a threshold. fb In the new image fb The center pixel is marked as the foreground (e.g., white, grayscale value 255), and the other pixels in that column are marked as the background (black, grayscale value 0). The image at this point... fb It's a preliminary binary image; theoretically, the white areas correspond to laser stripes, but at this point... fb It is usually very noisy. fb In this image, the noise appears as isolated white pixels or tiny white branches that do not belong to the main stripe. At this point, filtering removes the false maxima to obtain a binary image. fc The image obtained after filtering is... fc It is a clean, continuous, and stable binary fringe image that removes most of the interference and retains only the pixel areas most likely to belong to the main laser stripe.

[0035] Step 22) Extract the two edge points of the stripes column by column from the binary image, take the midpoint of the two edge points as the initial center point of the stripes in that column, and generate a center line image with a width of one pixel.

[0036] Specifically, for each column, locate the upper edge row coordinates and the lower edge row coordinates of the binary image, calculate the arithmetic mean of the upper edge row coordinates and the lower edge row coordinates as the initial center point of the stripes in that column of the binary image, assign the pixel gray value of the initial center point of the stripes to zero, and assign the pixel gray value of other points in the column to 1, thus obtaining a center line image with a single pixel width.

[0037] For example, extracting an image fc No. j (j = 1, 2,..., n The row coordinates of the top edge of the two edges in column 2 x and lower edge row coordinates y For images fc The j Scan the column from top to bottom to find the row number where the first white pixel appears, and denote it as . x (Top edge row coordinates); then find the row number where the last white pixel appears, and record it as... y (Lower edge row coordinates). Because the image fc The white areas in the image represent the true stripe areas after noise has been filtered out. For a vertical stripe, the row coordinate of its upper edge in a given column is... xand lower edge row coordinates y This defines the vertical span of the stripes in that column. Then, the position ( (x + y) / 2, j) This serves as the initial center point of the stripes in that column. Since the cross-sectional intensity distribution of the laser stripes is symmetrical, its geometric center can be approximated as the center of the strongest intensity. Finally, the pixel grayscale value of this initial center point is assigned to zero, and the pixel grayscale values ​​of all other points in the same column are assigned to 1. After completing this operation for all columns... fd The image then shows a vertical line composed of black pixels. This line has only one black dot in each column, making it a single-pixel-wide center line, thus creating a single-pixel-wide center line image. fd .

[0038] Step 23) For each column of the single-pixel wide centerline image, taking the row containing the initial center point of that column's stripes as the center, select pixel rows and their grayscale values ​​within a preset width range, and perform Gaussian curve fitting. The abscissa of the peak point of the fitted Gaussian curve is used as the sub-pixel coordinate of the column's stripe center. This step utilizes the physical property that the laser stripe intensity distribution conforms to a Gaussian model, and through mathematical fitting, calculates the precise location of the intensity peak point. This location is the sub-pixel coordinate of the laser stripe center, and its accuracy can exceed that of a single pixel.

[0039] Specifically, for striped images j Column elements, in [ i - s / 2... i + s / 2) ]Behavior X Vector, with [ (i - s / 2, j)... (i + s / 2, j) The grayscale value of ] Y vector; i For image fd Single pixel center line j The row position corresponding to the column; s For vectors s No. j The corresponding elements in the column represent laser stripes. j The stripe width value corresponding to the column position is used to solve the equation system using Gaussian fitting curve to determine the sub-pixel position of the stripe center. xi, j ).

[0040] Step S3) Identify the feature end points on the centerline. Since the contour line of a weld bead can be very long, but we really need to evaluate only the part of the weld metal itself, not the base metal next to it. The computer cannot automatically know where the starting point and ending point of the weld bead from the image or the centerline directly. Therefore, two feature end points (A and B) need to be identified to accurately demarcate the theoretical demarcation point of the weld bead and the original base metal. The centerline segment between the two points is clearly defined as the weld bead area that needs to be analyzed and evaluated. The contour changes outside this area belong to the original state of the base metal and are irrelevant to the welding quality and should be excluded from the evaluation system.

[0041] Specifically, as shown in the accompanying drawings, Figure 3 this step further includes the following steps: Step 31) dividing the centerline of the extracted weld bead cross-sectional contour into a first straight line segment, a top curve segment, and a second straight line segment.

[0042] As shown in the accompanying drawings, Figure 4 the contour features of the laser welded weld bead end face are shown. The first straight line segment, i.e., line 1, and the second straight line segment, i.e., line 2, exhibit straight line characteristics, and the top curve segment, i.e., line 3, exhibits curve characteristics and is approximately a quadratic curve. This is determined by the typical cross-sectional contour of the weld bead: the two sides are the base metal surfaces that are not affected by heat and are approximately straight lines; the middle is the raised area formed after the molten metal solidifies, which is approximately a quadratic curve (parabola). By dividing the continuous centerline data points into three logical segments, the most suitable mathematical model can be used for description for different segments in the subsequent process, avoiding errors caused by using a single model to fit the entire contour.

[0043] Step S2) respectively fitting the straight line equations of the first straight line segment and the second straight line segment, and the quadratic curve equation of the top curve segment using the least squares method.

[0044] The straight line equation of the first straight line segment is: a 1 x + b 1 y + c 1 =0 ; The straight line equation of the second straight line segment is: a 2 x + b 2 y + c 2 =0 ; The quadratic curve equation of the top curve segment is: y = a 3 x 2 +b 3x + c 3 ; wherein: a 1 、 b 1 、 c 1 、 a 2 、 b 2 、 c 2 、 a 3 、 b 3 、 c 3 are coefficients.

[0045] Step S33) The coordinates of the first characteristic end point B of the weld are calculated by combining the straight line equation of the first straight line segment and the quadratic curve equation of the top curve segment; the coordinates of the second characteristic end point A of the weld are calculated by combining the straight line equation of the second straight line segment and the quadratic curve equation of the top curve segment. The first characteristic end point and the second characteristic end point strictly define the theoretical demarcation point between the weld convex portion and the original base material in geometry, and will also be used as the geometric reference for subsequent calculation of the convexity and concavity of the weld (an ideal reference straight line is obtained by connecting the two points).

[0046] Step S4) The convexity and concavity of the weld surface are calculated based on the characteristic end points and the center line.

[0047] Specifically, as shown in FIG. 4, this step further includes the following steps: Figure 5 Step S41) A reference straight line is obtained by connecting the first characteristic end point and the second characteristic end point. The line segment between the first characteristic end point B and the second characteristic end point A is the ideal weld end face profile, i.e., the reference straight line, and the equation of the reference straight line is: ax + by + The reference straight line represents a perfect weld surface in geometry, i.e., an ideal state in which the welding material exactly fills the groove and the surface is completely flush with the base material. This reference straight line is adaptive, and its position and angle will change with different weld widths and positions, thereby providing the most reasonable and objective evaluation standard for each weld. c=0

[0048] ​Step S42) Search for all pixels located between the first and second feature endpoints along the centerline. Since the extracted centerline may include portions of the base material on both sides, this step precisely extracts the contour line belonging only to the weld itself from the data using the first and second feature endpoints. This ensures that subsequent calculations only consider the morphological changes caused by the welding process, and are not affected by the original workpiece shape or background noise.

[0049] Step S43) Calculate the vertical distance from each pixel to the reference line. If the pixel is above the reference line, record it as a positive value; if the pixel is below the reference line, record it as a negative value.

[0050] Specifically, search for all pixels on the weld centerline image that lie between the first and second feature endpoints, and record the coordinates of the found pixels. x i ，y i ), Judgment point ( x i ，y i and reference line ax + by + c = 0 Positional relationship, if point ( x i ，y i Given a point located above a baseline line, calculate the distance from that point to the line and record it as positive, then store it in an array. d + (i) If point ( x i ，y i Located below the baseline line, calculate the distance from the point to the line and record it as negative, then store it in the array. d _( i) .

[0051] Among them, pixels ( x i ，y i perpendicular distance to the reference line d(i) The calculation formula is:

[0052] In the formula: a , b , c All are coefficients.

[0053] Step S44) Select the maximum value from all positive distances as the convexity of the weld, and select the minimum value from all negative distances as the concavity of the weld.

[0054]

[0055] The convexity is to find the maximum value max d+(i) from all positive deviations (convex), which represents how much the most convex place of the weld surface is higher than the ideal plane, and is an indicator for evaluating the excess height or bulge.

[0056] The concavity is to find the minimum value min d-(i) from all negative deviations (concave), which represents how much the most concave place of the weld surface is lower than the ideal plane, and is an indicator for evaluating the undercut, collapse or pore depth. d _( i) (i.e. the value that is most negative, with the largest absolute value). This represents how much the most concave place of the weld surface is lower than the ideal plane, and is an indicator for evaluating the undercut, collapse or pore depth.

[0057] Thus, the complex profile curve is finally condensed into two numbers with clear engineering significance: convexity and concavity For example, convexity = +0.05mm, indicating that the maximum bulge height is 0.05mm. Concavity = -0.02mm, indicating that the maximum concave depth is 0.02mm.

[0058] Step S5) Determine whether the weld quality is qualified according to the convexity and concavity of the weld surface.

[0059] Specifically, the calculated convexity and concavity values of the weld are converted to actual physical lengths through preset calibration parameters; The actual physical length is compared with the preset process requirement depth interval to determine whether the weld quality is qualified.

[0060] The two indicators of convexity and concavity are intuitive, comparable and evaluable. In actual use, they can be directly compared with the process specification (such as "convexity <0.1mm and concavity >-0.03mm is qualified") to make an automatic "qualified / unqualified" determination for a single weld position.

[0061] Compared with the prior art, the present application projects a laser plane to the weld surface after laser welding is completed, forms a modulated laser stripe, and collects a laser stripe image containing the laser stripe; then a Gaussian fitting method is used to extract the sub-pixel coordinates of the center of each column of stripes in the laser stripe image, to generate a center line of the weld cross-section profile; feature end points on the center line are identified; based on the feature end points and the center line, the convexity and concavity of the weld surface are calculated; finally, the weld quality is determined according to the convexity and concavity of the weld surface. Therefore, the entire process of weld quality determination is based on mathematical calculation and is completed automatically, completely eliminating the subjectivity, fatigue and inconsistency of manual visual inspection, making the determination result objective, reliable and traceable, improving the reliability of automatic welding, and meeting the production requirements of the antenna panel.

[0062] The present application accurately positions the weld boundary through an endpoint identification method based on a geometric model (intersection of straight lines and quadratic curve fitting), establishes a reliable reference for subsequent measurement. At the same time, by defining two core quantitative indicators of "convexity" and "concavity", the maximum positive and negative deviation of the actual profile from the ideal reference line is calculated, accurately and intuitively describing the convex and concave defects of the weld, thereby enabling real-time detection of key appearance defects such as weld undercut, collapse, and excessive excess height, avoiding the flow of substandard products into the next process.

[0063] Laser welding has the characteristics of large weld depth-width ratio and narrow heat-affected zone due to small spot size and concentrated energy, and is prone to undercut defects. The present application detects the weld surface features after welding through optical vision method, which can directly detect the size and shape of the weld pool and surface porosity and other defects, with obvious advantages of rapidity, real-time, high detection accuracy, etc.

[0064] In addition, the present embodiment also provides a laser welding device, as shown in the accompanying drawings Figure 6 and Figure 7 , comprising: a reference platform 1; a workpiece fixing tool 2 mounted on the reference platform 1 for fixing a workpiece to be processed 3; a three-axis motion module 4 arranged above the reference platform 1; a motion controller for driving the three-axis motion module 4 to move; a laser light path module 5 mounted on the three-axis motion module 4 for emitting processing laser; a global camera 6 mounted on the three-axis motion module 4 for shooting a global image of the workpiece to be processed to identify the welding point; And the laser weld online detection device, the laser weld online detection device includes laser light source, is used to the laser plane is projected to the weld surface, to form the modulated laser stripe;Image collector 7, for collecting laser stripe image containing laser stripe;Host computer, host computer is connected with laser light source and image collector respectively, simultaneously, host computer is connected with motion controller, laser light path module 5 and global camera 6 respectively, host computer is configured as: Based on the global image of the workpiece to be processed 3, the three-axis motion module 4 is driven by the motion controller to move the laser light path module 5 to the welding point of the workpiece to be processed 3 for welding; Receiving laser stripe image, using Gaussian fitting method to extract sub-pixel coordinates of each column stripe center in laser stripe image, generating center line of weld cross section profile; Identify the feature endpoints on the center line; Based on the feature endpoints and the center line, the convexity and concavity of the weld surface are calculated; According to the convexity and concavity of the weld surface, whether the weld quality is qualified is determined.

[0065] The working process of the laser welding device is as follows: the workpiece to be processed is installed on the workpiece fixing tool, the global camera takes a picture of the workpiece to be processed and sends it to the host computer, the host computer identifies the best welding point, the host computer sends the coordinates of all welding points to the motion controller, the motion controller drives the three-axis motion module to move the laser light path module to the welding point above the X and Y axes and performs laser welding, the image collector takes a picture of the weld, the host computer analyzes the welding quality in real time, if the welding is weak, it is welded, if the welding point is normal, it is moved to the next welding point for welding operation.

[0066] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and not to limit the technical solutions, those skilled in the art should understand that those who modify or replace the technical solutions of the present application without departing from the purpose and scope of the present application, should be covered in the scope of claims of the present application.

Claims

1. A method for online inspection of laser welds, characterized in that, Includes the following steps: After laser welding is completed, a laser plane is projected onto the weld surface to form modulated laser stripes, and an image of the laser stripes containing these stripes is acquired. The sub-pixel coordinates of the center of each column of stripes in the laser stripe image are extracted using Gaussian fitting to generate the center line of the weld cross-section profile. Identify the characteristic endpoints on the centerline; Based on the feature endpoints and the centerline, the convexity and concavity of the weld surface are calculated; The quality of the weld is determined based on the convexity and concavity of the weld surface.

2. The online laser weld inspection method according to claim 1, characterized in that, The sub-pixel coordinates of the stripe centers in the laser stripe image are extracted using Gaussian fitting to generate the centerline of the weld cross-section profile, including: The maximum gray value is searched column by column along the normal direction of the laser stripe image, and the maximum gray value is used as a threshold for binarization to obtain a binary image. The two edge points of the stripes are extracted column by column from the binary image. The midpoint of the two edge points is taken as the initial center point of the stripes in that column, and a center line image with a width of one pixel is generated. For each column of the single-pixel wide centerline image, with the row containing the initial center point of the stripe in that column as the center, select a pixel row and its gray value within a preset width range, perform Gaussian curve fitting, and use the abscissa of the peak point of the fitted Gaussian curve as the subpixel coordinate of the center of the stripe in that column.

3. The online laser weld inspection method according to claim 2, characterized in that, Extracting the two edge points of the stripes column by column from the binary image, taking the midpoint of the two edge points as the initial center point of the stripes in that column, and generating a single-pixel-wide centerline image, including: For each column, locate the upper and lower row coordinates of the binary image, calculate the arithmetic mean of the upper and lower row coordinates as the initial center point of the stripes in that column, assign the pixel gray value of the initial center point of the stripes to zero, and assign the pixel gray value of the other points in the column to 1, thus obtaining a center line image with a single pixel width.

4. The online laser weld inspection method according to claim 1, characterized in that, Identifying the characteristic endpoints on the centerline includes: The centerline of the extracted weld section profile is divided into a first straight line segment, a top curved segment, and a second straight line segment. The equations of the first and second line segments, as well as the quadratic curve equation of the top curve segment, are fitted using the least squares method, respectively. By combining the linear equation of the first straight line segment with the quadratic curve equation of the top curve segment, the coordinates of the first characteristic endpoint of the weld are calculated; by combining the linear equation of the second straight line segment with the quadratic curve equation of the top curve segment, the coordinates of the second characteristic endpoint of the weld are calculated.

5. The online laser weld inspection method according to claim 4, characterized in that, The equations of the first and second line segments, and the quadratic curve equation of the top curve segment, are fitted using the least squares method, respectively, including: The equation of the first line segment is: a 1 x+b 1 y+c 1 =0 ; The equation of the second line segment is: a 2 x+b 2 y+c 2 =0 ; The equation of the quadratic curve in the top curve segment is: y=a 3 x 2 +b 3 x+c 3 ; In the formula: a 1 , b 1 , c 1 , a 2 , b 2 , c 2 , a 3 , b 3 , c 3 All are coefficients.

6. The online laser weld inspection method according to claim 5, characterized in that, Based on the characteristic endpoints and the centerline, the convexity and concavity of the weld surface are calculated, including: A reference straight line is obtained by connecting the first feature endpoint and the second feature endpoint; Search the center line for all pixels located between the first feature endpoint and the second feature endpoint; Calculate the vertical distance from each pixel to the reference line. If the pixel is above the reference line, it is recorded as a positive value; if the pixel is below the reference line, it is recorded as a negative value. The maximum value among all positive distances is selected as the weld convexity, and the minimum value among all negative distances is selected as the weld concavity.

7. The online laser weld inspection method according to claim 6, characterized in that, Based on the characteristic endpoints and the centerline, the convexity and concavity of the weld surface are calculated, including: The equation of the reference line is: ax + by + c = 0 ; The pixel ( x i ，y i The perpendicular distance from the reference line d(i) The calculation formula is: ； In the formula: a , b , c All are coefficients.

8. The online laser weld inspection method according to claim 7, characterized in that, Determining whether the weld quality is acceptable based on the convexity and concavity of the weld surface includes: The calculated convexity and concavity values ​​of the weld are converted into actual physical lengths using preset calibration parameters; The actual physical length is compared with the preset depth range required by the process to determine whether the weld quality is qualified.

9. A laser weld seam online inspection device, characterized in that, The method for implementing the online laser weld inspection method as described in claim 1 includes: A laser source is used to project a laser plane onto the weld surface to form modulated laser stripes; An image acquisition device used to acquire images of laser stripes, including laser stripes; The host computer is communicatively connected to both the laser source and the image acquisition device, and is configured as follows: The laser stripe image is received, and the sub-pixel coordinates of the center of each column of stripes in the laser stripe image are extracted using Gaussian fitting method to generate the center line of the weld cross-section profile. Identify the characteristic endpoints on the centerline; Based on the feature endpoints and the centerline, the convexity and concavity of the weld surface are calculated; The quality of the weld is determined based on the convexity and concavity of the weld surface.

10. A laser welding apparatus, characterized in that, include: Benchmark platform; A workpiece fixing fixture is installed on the reference platform to fix the workpiece to be processed; A three-axis motion module is mounted above the reference platform; Motion controller, used to drive the movement of the three-axis motion module; A laser optical path module is mounted on the three-axis motion module and is used to emit processing lasers; A global camera, mounted on the three-axis motion module, is used to capture a global image of the workpiece to be processed in order to identify welding points; And the online laser weld inspection device as described in claim 9, wherein the host computer is communicatively connected to the motion controller, the laser optical path module, and the global camera, and the host computer is further configured to: Based on the global image of the workpiece to be processed, the motion controller controls the three-axis motion module to move the laser optical path module to the welding point of the workpiece to be processed for welding.

Images