Closed-loop control method for laser-arc hybrid welding process of thin plate long weld

Abstract

The application provides a closed-loop control method for a laser-arc hybrid welding process of a long weld seam of a thin plate, comprising the following steps: step one, welding groove image processing; step two, welding gap and misalignment positioning; step three, welding process matching; step four, welding process closed-loop strategy generation: based on assembly quality and welding process, corresponding welding processes are matched for different gaps and misalignment sizes, and a welding process closed-loop strategy is generated; and step five, welding path and process generation: the generated welding path and process are issued to a robot and a welding machine through a control system, and a welding process is performed; through the closed-loop control method for the laser-arc hybrid welding process of the long weld seam of the thin plate, the setting of the control method can be optimized, online compensation of welding process parameters in the welding process is realized, the long weld seam welding structure welding quality and mechanical property requirements can be met, labor can be saved, and welding cost can be greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of automated welding technology in laser-arc hybrid welding, and more specifically, to a closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates. Background Technology

[0002] With the improvement of arc welding equipment and laser performance, the development of laser-arc hybrid welding technology has accelerated, becoming a hot research area in welding. It significantly improves welding efficiency and reduces welding deformation by combining the high energy density of lasers with the filling advantage of electric arcs. Application research in laser-arc hybrid welding involves industries such as automotive, shipbuilding, aerospace, and oil pipelines, primarily targeting thin and medium-thick plate welding. However, this technology has not yet been well applied in the welding of long weld seams in thin plates. This is because the assembly quality at local locations in long weld seams exceeds the assembly precision requirements, and deformation during welding may further increase the bevel assembly gap and misalignment. Currently, most robotic welding operates on a "teach-and-reproduce" welding model, unable to autonomously adjust process parameters based on assembly conditions, lacking the ability to perceive and provide feedback on assembly precision, and failing to achieve closed-loop quality control of the welding process.

[0003] Domestic and international researchers have conducted extensive research and practical work on closed-loop quality control of the welding process. Foreign countries are leading in multi-physics coupling modeling for welding. The welding simulation system developed by the team at the University of Tokyo in Japan can accurately predict weld reinforcement deviation. The Fraunhofer Institute in Germany has achieved precise control of process parameters by establishing a heat input-porosity regression model. Furthermore, the Fraunhofer Institute has also developed a composite welding system based on the CMT+P (Cold Metal Transfer Pulse) arc mode, achieving stable welding of 45mm high-strength steel by dynamically adjusting the pulse frequency to match laser power fluctuations. RWTH Aachen University in Germany optimizes the filament spacing through plasma dynamics simulation and combines artificial intelligence (such as genetic algorithms and neural networks) to predict the molten pool morphology in real time, reducing porosity and spatter. US research institutions have introduced magnetic field control to regulate molten pool flow (such as Lorentz force controlling element distribution) to improve the uniformity of thick plate welding. In addition, Rockwell Automation's ADAPT system significantly improves welding speed and penetration stability by learning weld penetration data online.

[0004] Domestically, Southwest Jiaotong University, through high-speed imaging and spectral analysis, discovered that oscillating laser parameters can stabilize the electric arc and reduce keyhole fluctuations, lowering the porosity of thick aluminum plates to below 0.5%. China Nuclear Industry Huaxing Construction further integrated beveling and welding functions, achieving a closed-loop integrated production process for nuclear power plant steel linings, combining milling and welding. Southwest Jiaotong University confirmed that oscillating lasers can enlarge the keyhole opening, suppress the risk of collapse, and improve the consistency of weld depth in thick plates. Shanghai University of Engineering Science developed a fusion system of high-speed cameras and Hall effect sensors to extract features such as droplet transition frequency and arc deflection angle, combining them with machine learning algorithms to identify defects such as collapse and weld misalignment.

[0005] In summary, although there has been some research on closed-loop quality control in the welding process, it has largely focused on dynamic monitoring of welding process quality and the construction of welding defect models, and has not yet achieved online compensation of welding parameters. Therefore, research is needed on how to achieve online compensation of welding process parameters during the welding process. This would enable the welding quality and mechanical performance requirements of long weld seams to be met, thus offering advantages such as labor savings and low welding costs, which is of great significance for improving the efficiency and quality of shipbuilding.

[0006] Patent CN116060772A discloses a laser-arc hybrid welding method and equipment, including a laser-hybrid welding robot, an active laser vision system, a laser light source, an image processing system, and a control system. The active laser vision system and the laser light source are both mounted on the laser-hybrid welding robot, and the image processing system is located in the control system. The laser-hybrid welding robot, the active laser vision system, the positioning cross slide, the laser light source, and the image processing system are all connected to the control system. The aim is to solve the problem that the arc light and spatter interference in existing laser-arc hybrid welding cause excessive image noise in the weld seam tracking system, ultimately affecting the welding quality, welding accuracy, and efficiency. However, since this equipment and method are applied to Y-shaped bevels, they cannot effectively achieve online compensation of welding process parameters during the welding of long weld seams in thin plates. Summary of the Invention

[0007] In view of this, the present invention aims to propose a closed-loop control method for laser-arc hybrid welding of thin plate long welds, in order to solve the problems in the existing technology where the welding technology cannot autonomously adjust process parameters according to the assembly situation and lacks the ability to perceive and provide feedback on assembly accuracy, thus failing to achieve closed-loop quality control of the thin plate long weld process. Even though current research on closed-loop quality control of the welding process is underway, it cannot achieve online compensation of welding parameters, which easily leads to high welding costs and low efficiency. The present invention aims to optimize the setting of the control method and achieve online compensation of welding process parameters during the welding process, so as to meet the welding quality and mechanical performance requirements of thin plate long weld structures, thereby saving labor and greatly reducing welding costs.

[0008] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0009] This invention relates to a closed-loop control method for laser-arc hybrid welding of long weld seams in thin plates, the method comprising:

[0010] Step 1: Welding bevel image processing: Before welding, the welding bevel of the entire thin plate long weld structure is scanned and processed.

[0011] Step 2, Welding gap and misalignment positioning: Based on the results of image scanning and processing, determine the assembly dimensions of different parts of the long weld, and locate the welding gap and misalignment in different positions;

[0012] Step 3: Welding process matching: Use welding process database software to determine the welding process specifications for different assembly dimensions;

[0013] Step 4: Welding process closed-loop strategy generation: Based on assembly quality and welding process, match the corresponding welding process for different gaps and misalignment dimensions to generate a welding process closed-loop strategy;

[0014] Step 5: Welding path and process generation: The generated welding path and process are sent to the robot and welding machine through the control system to execute the welding process.

[0015] Furthermore, before processing the welding bevel image in step one, it is necessary to assemble an arc-starting plate at the arc-starting position of the welding bevel for the welded structural component.

[0016] Furthermore, step one includes:

[0017] Step S11: Image preprocessing: Before welding, the welding groove of the entire thin plate long weld structure is scanned, and then the scanned image is preprocessed using an image processing algorithm;

[0018] Step S12: Centerline extraction: The centerline of the laser stripes is quickly extracted using the required thinning algorithm; and boundary pixels that meet specific conditions are iteratively deleted to finally obtain the skeleton representation of the image.

[0019] Step S13: Denoising algorithm after centerline extraction: Based on the physical continuity constraint of laser stripes, isolated noise points are eliminated through column scanning and neighborhood correlation analysis;

[0020] Step S14: Obtain the gap variation area;

[0021] Step S15: Obtain feature points of the gap change area: Extract feature points from the weld bevel to obtain image information of the weld bevel.

[0022] Furthermore, the image processing algorithm in step S11 can be any one of the following: grayscale processing algorithm, Gaussian filtering algorithm, or adaptive binarization method.

[0023] Furthermore, step S13 includes:

[0024] Step S131: Input the skeletonized image obtained by line laser scanning;

[0025] Step S132: Perform column scanning and determine whether there are pixels in the current column; if yes, proceed to step S133; if no, skip the column, repeat step S132, and perform scanning of the next column.

[0026] Step S133: Determine whether the current column pixel is a single point; if yes, proceed to step S134; if no, proceed to step S135.

[0027] Step S134: Directly save the pixel data of the current column and proceed to step S136;

[0028] Step S135: After minimizing the distance with the previous column of points using the nearest neighbor selection strategy, proceed to step S136;

[0029] Step S136: Output the filtered point set.

[0030] Furthermore, the column scan in step S132 is a column vector scan.

[0031] Furthermore, step S135 includes:

[0032] Nearest neighbor decision stage: When there are two or more candidate pixels in a column, the minimum distance criterion is used for filtering; First column processing: Select the first valid pixel in the column as the initial reference point; Subsequent column processing: Calculate the vertical distance between all candidate pixels in the current column and the reference point of the previous column, and select the point with the smallest distance as the new reference point.

[0033] Furthermore, step S135 also includes two physical constraints: a vertical jump constraint and a trend maintenance constraint.

[0034] Furthermore, in step two, the three-dimensional point cloud obtained by scanning the welding groove is processed into a three-dimensional point cloud image for assembly and gap positioning. Combining the three-dimensional point cloud image processing method, its specific position and length in the weld are determined, and coordinate annotation is performed.

[0035] Furthermore, in step four, for long weld structures, the welding process and corresponding welding distance for different welding positions are determined, and a closed-loop welding process strategy is generated.

[0036] Compared with existing technologies, the closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates described in this invention has the following advantages:

[0037] The closed-loop control method for laser-arc hybrid welding of thin plate long welds described above can optimize the control method settings and realize online compensation of welding process parameters during the welding process. This enables the method to meet the welding quality and mechanical performance requirements of thin plate long weld structures, thereby saving labor and greatly reducing welding costs. Attached Figure Description

[0038] The accompanying drawings, which constitute a part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0039] Figure 1 This is a schematic diagram of the overall flow of the control method;

[0040] Figure 2 This is a schematic diagram of a Gaussian filtered image;

[0041] Figure 3 This is a schematic diagram of a binarized image;

[0042] Figure 4 Schematic diagram of image extraction for centerline;

[0043] Figure 5 This is a flowchart illustrating the denoising algorithm.

[0044] Figure 6 This is a schematic diagram of the filtered image after centerline extraction (first moment).

[0045] Figure 7 This is a schematic diagram of the filtered image after centerline extraction (first moment).

[0046] Figure 8 Schematic diagram of image extraction for assembly gap feature points;

[0047] Figure 9 A schematic diagram of image extraction for misaligned edge feature points during assembly;

[0048] Figure 10 A schematic diagram of the point cloud image for welding bevel assembly;

[0049] Figure 11 This is a schematic diagram of the welding process database;

[0050] Figure 12 This is a schematic diagram of a closed-loop strategy for welding processes.

[0051] Figure 13 A schematic diagram of the actual structure of the butt welding test plate;

[0052] Figure 14 This is a schematic diagram of the actual structure of the fillet weld test plate. Detailed Implementation

[0053] The inventive concepts of this disclosure will be described below using terminology commonly used by those skilled in the art to communicate the essence of their work to others skilled in the art. However, these inventive concepts may be embodied in many different forms and should not be construed as limited to the embodiments described herein.

[0054] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0055] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0056] To address the shortcomings of existing welding technologies, such as the inability to autonomously adjust process parameters based on assembly conditions and the lack of perception and feedback capabilities regarding assembly accuracy, thus hindering closed-loop quality control in the long weld of thin plates, and despite current research on closed-loop quality control, online compensation of welding parameters is not yet possible, leading to high welding costs and low efficiency. This embodiment proposes a closed-loop control method for laser-arc hybrid welding of long welds in thin plates, specifically a closed-loop control method for long welds based on laser-arc hybrid welding. The method includes the following steps:

[0057] Step 1: Welding bevel image processing: Before welding, the welding bevel of the entire thin plate long weld structure is scanned and processed.

[0058] Step 2, Welding Gap and Misalignment Positioning: Based on the results of image scanning and processing, determine the assembly dimensions of different parts of the long weld, and locate the welding gap and misalignment in different positions; so as to determine the specific location and corresponding length of the gap or misalignment in the welded structure.

[0059] Step 3: Welding process matching: Use welding process database software to determine the welding process specifications for different assembly dimensions;

[0060] Step 4: Welding process closed-loop strategy generation: Based on assembly quality and welding process, corresponding welding processes are matched for different gaps and misalignment dimensions to generate a welding process closed-loop strategy; through the proposed welding process quality closed-loop control strategy, different welding process parameters are called for different welding positions during the welding process to achieve online compensation of welding process parameters.

[0061] Step 5: Welding path and process generation: The generated welding path and process are sent to the robot and welding machine through the control system to execute the welding process.

[0062] In step one, before processing the weld bevel image, an arc-starting plate needs to be installed at the arc-starting position of the weld bevel for the welded structural component. Additionally, when locating misalignments and gaps, the length of the weld arc-starting plate must be considered to ensure accurate positioning of any gaps and misalignments that may occur during assembly.

[0063] To address the issue of assembly dimensions exceeding accuracy requirements in laser-arc hybrid welding of long weld seams in thin steel plates for marine applications, this application proposes a closed-loop control method for laser-arc hybrid welding of long weld seams in thin plates, based on assembly quality and welding process data. This method optimizes the control settings and enables online compensation of welding process parameters during welding, thereby meeting the welding quality and mechanical performance requirements of long weld seam structures in thin plates. This also saves labor and significantly reduces welding costs.

[0064] Step one includes:

[0065] Step S11: Image preprocessing: Before welding, the welding groove of the entire thin plate long weld structure is scanned, and then the scanned image is preprocessed using an image processing algorithm;

[0066] Step S12: Centerline Extraction: The centerline of the laser stripes is quickly extracted using the required thinning algorithm; and boundary pixels that meet specific conditions are iteratively deleted to finally obtain the skeleton representation of the image. The required thinning algorithm is an improved ZhangSuen skeletonization algorithm.

[0067] Step S13: Denoising algorithm after centerline extraction: Based on the physical continuity constraint of laser stripes, isolated noise points are eliminated through column scanning and neighborhood correlation analysis;

[0068] Step S14: Obtain the gap variation area;

[0069] Step S15: Obtain feature points of the gap change area: Extract feature points from the weld bevel to obtain image information of the weld bevel.

[0070] By setting each step in the process, the area of ​​change in the gap can be effectively obtained, thereby enabling the effective acquisition of feature points.

[0071] The image processing algorithm in step S11 can be any one of the following: grayscale processing algorithm, Gaussian filtering algorithm, or adaptive binarization method.

[0072] Because scanning the weld bevel is affected by the reflective properties of the weld surface, process residues, and environmental noise, image processing of the weld bevel is required to stably extract effective geometric feature points representing the weld from complex light stripe images, obtain the two-dimensional cross-sectional profile of the weld bevel, and calculate the assembly gap and dimensions.

[0073] Image processing algorithms that use grayscale conversion convert RGB images to grayscale images and preserve human-eye-sensitive features through weighted summation.

[0074] I{gray}(x,y) refers to the color mixing function; R(x,y) refers to the red primary color function; G(x,y) refers to the green primary color function; and B(x,y) refers to the blue primary color function.

[0075] The advantage of using grayscale processing algorithm is that it can reduce the amount of computation by more than 60% compared with three-channel RGB images, and can eliminate color information interference, highlight the characteristics of laser stripes, and provide a unified brightness benchmark for subsequent binarization.

[0076] The principle behind the Gaussian filtering algorithm used in image processing is to use a two-dimensional Gaussian kernel for convolution to suppress high-frequency noise. G(x,y) refers to the Gaussian filter function; σ refers to the standard deviation; e refers to the base of the natural logarithm (approximately 2.71828); (x,y) are the point coordinates, which can be considered integers in image processing. In this embodiment, when the image processing algorithm of this application uses the Gaussian filter algorithm, the size of the Gaussian kernel is 3×3, and σ=0.8; it can improve the signal-to-noise ratio (SNR) by 2-3dB, so as to preserve sub-pixel level edge sharpness.

[0077] When the image processing algorithm employs an adaptive binarization algorithm, it first improves the local window-based Otsu algorithm using a first formula. Next, it adjusts the characteristic value k of metallic reflection and uses a second formula to achieve adaptive window calculation. In this embodiment, the value of the characteristic value k for metallic reflection is adjusted to k=0.5.

[0078] It should be noted that in step S11, the first formula is:

[0079] ,

[0080] Where T(x,y) refers to the function for calculating the grayscale value of a local region; μ local (x,y) refers to the average gray value of a local area, that is, the average gray value of a small area centered on pixel (x,y); k refers to the characteristic coefficient of metallic reflection; σ local (x,y) refers to the standard deviation of a local region, that is, the degree of dispersion of pixel gray values ​​within that region.

[0081] The second formula is: W(x,y) refers to the function used to calculate the grayscale values ​​of the selected region; w min This refers to the minimum grayscale value of the selected area; w base This refers to the maximum grayscale value of the selected area; I global It refers to the grayscale value of the global region; μ(x,y) refers to the average grayscale value of that region.

[0082] By adjusting the adaptive binarization algorithm, the problem of uneven reflection on the welding surface can be overcome, reducing the missegmentation rate by 42% compared to the traditional threshold, and making it more robust to regions with gradual changes in light intensity.

[0083] The denoising algorithm after centerline extraction in step S13 is designed for skeletonized images obtained by line laser scanning. It achieves noise filtering and centerline optimization through column scanning and dynamic nearest neighbor selection strategies.

[0084] Specifically, step S13 includes:

[0085] Step S131: Input the skeletonized image obtained by line laser scanning;

[0086] Step S132: Perform column scanning and determine whether there are pixels in the current column; if yes, proceed to step S133; if no, skip the column, repeat step S132, and perform scanning of the next column.

[0087] Step S133: Determine whether the current column pixel is a single point; if yes, proceed to step S134; if no, proceed to step S135.

[0088] Step S134: Directly save the pixel data of the current column and proceed to step S136;

[0089] Step S135: After minimizing the distance with the previous column of points using the nearest neighbor selection strategy, proceed to step S136;

[0090] Step S136: Output the filtered point set.

[0091] The application of the denoising algorithm in step S13 can facilitate the removal of noise interference, improve the accuracy of detection results, and thus ensure the precision and reliability of system control.

[0092] The column scan in step S132 is a column vector scan. Step S132 includes:

[0093] Step S1321: Perform column vector scanning. In this stage, the denoising algorithm scans each column of the image sequentially from left to right. All white pixels are detected in each column. White pixels refer to pixels with a grayscale value of 255.

[0094] Step S1322: For each column of pixels, determine whether a pixel is detected in the current detection column. If yes, proceed to step S133 to determine the number of valid pixels in the current column; otherwise, skip the processing of that column and return to step S1321 to scan the next column.

[0095] Step S133 includes: determining whether a single valid pixel is detected in the current column. If yes, directly use that pixel as the representative point of the current column and proceed to step S134. If no, and two or more valid pixels are detected in the current column, proceed to step S135 and enter the neighbor decision stage. The situation where two or more valid pixels are detected in a single column is commonly seen in cases of noise interference or branch residue.

[0096] By determining the existence and number of valid pixels in steps S132-133, a more suitable scanning adjustment method can be selected, thereby reducing the occurrence of test errors and improving the accuracy of the scanning results.

[0097] Step S135 includes:

[0098] Nearest neighbor decision stage: When there are two or more candidate pixels in a column, the minimum distance criterion is used for selection. First column processing: Select the first valid pixel in the column as the initial reference point; Subsequent column processing: Calculate the vertical distance between all candidate pixels in the current column and the reference point of the previous column, and select the point with the smallest distance as the new reference point.

[0099] Step S135 further includes two physical constraints: a vertical jump constraint and a trend maintenance constraint. The vertical jump constraint, by default, limits the vertical offset of adjacent column reference points to no more than 5 pixels (adjustable parameter τ). The trend maintenance constraint implicitly constructs a motion trend by recording historical point positions, preventing "Z"-shaped jitter in multiple consecutive columns.

[0100] By implicitly implementing these two physical constraints through the denoising algorithm, the smoothness of the centerline can be effectively improved and the noise of spikes can be suppressed, thereby improving the integrity and accuracy of the image scanning results.

[0101] Step S136: Output optimization stage: The final output point set satisfies the following: Topological continuity: The horizontal distance between any two adjacent points is strictly 1 pixel (column adjacent), and the vertical distance is ≤ τ pixels; Positional optimality: The distance between each selected point and the previous reference point in its column is minimized; Completeness: All column information of the original skeleton in the effective area is retained without omission.

[0102] Step S14 includes:

[0103] Step S141: Obtain the gap variation region: First, the filtered laser center line pixels are continuously segmented. By setting the maximum allowable interval, the point set is divided into several continuous segments to ensure the difference between the x coordinates of adjacent points in each segment.

[0104] Step S142: Next, through statistical analysis, traverse all adjacent point pairs within the segment, count the number of times the x value increases and decreases, and calculate the allowable upper limit of fluctuation based on the fault tolerance ratio, and output the characteristic segment that meets the requirements.

[0105] Step S15 includes: acquiring feature points of the gap variation region: For the detected effective gap variation region, its start point and end point are taken as key positions characterizing the weld geometry. These two feature points completely define the core geometric features of the weld groove gap region. The weld groove assembly gap can be calculated using the feature point image coordinates.

[0106] It should be noted that, similar to the steps in step one, this closed-loop control method can also extract the assembly misalignment feature points.

[0107] By configuring steps S14-S15, feature points within the gap variation area can be obtained, and the welding groove assembly gap can be calculated. This lays the foundation for improving the accuracy of welding gap and misalignment positioning based on the scanning results.

[0108] In step two, the three-dimensional point cloud obtained by scanning the welding groove is processed into a three-dimensional point cloud image for assembly and gap positioning. The specific position and length of the point cloud in the weld are determined by combining the three-dimensional point cloud image processing method, and coordinate annotation is performed.

[0109] By setting up 3D point clouds, welding efficiency and quality can be significantly improved.

[0110] In step three, since different assembly gaps and misalignments correspond to different welding processes, the welding process database is called up based on the positioning of welding gaps and misalignments to match welding processes for different assembly dimensions.

[0111] In step four, for long weld structures, the welding process and corresponding welding distance for different welding positions are determined, and a closed-loop strategy for the welding process is generated.

[0112] Step five includes: Welding path and process generation: The welding process closed-loop strategy is parsed and converted into robot path code and welding machine execution commands. The generated welding path and process are transmitted to the welding robot and welding machine through the control port to execute the welding process.

[0113] By coordinating step five with steps one through four, the control method of this application possesses advantages such as simple principle, wide applicability, and high engineering application value. Furthermore, it overcomes the problem that gaps and misalignments occurring during the assembly of long weld seams prevent a single welding process from guaranteeing welding quality. It achieves online compensation of welding process parameters matching different assembly dimensions during the welding process, completing closed-loop control of the welding process. This method can significantly improve welding efficiency and quality.

[0114] Example 1:

[0115] Using the control method described in this application, welds were performed on butt weld plates and fillet weld plates of different specifications. The butt weld plate was 4mm thick, with a corresponding weld length of 6m; the fillet weld plate was 6mm thick, with a corresponding weld length of 5.2m. After welding, the results showed that the flaw detection of both the butt weld and fillet weld met the requirements, and the welding quality was good.

[0116] Example 2:

[0117] The difference from Example 1 lies in the specifications of the butt weld plate, fillet weld plate, and weld seam. Specifically, the butt weld plate is 6mm thick, corresponding to a weld length of 5.2m; the fillet weld plate is also 6mm thick, with a corresponding weld length of 5.2m. After welding, the results showed that the flaw detection of both the butt weld and fillet weld seam met the requirements, indicating good welding quality.

[0118] Example 3:

[0119] The difference from Example 1 lies in the specifications of the butt weld plate, fillet weld plate, and weld seam. Specifically, the butt weld plate is 8mm thick, corresponding to a weld seam length of 6m; the fillet weld plate is also 8mm thick, corresponding to a weld seam length of 6m. After welding, the results showed that the flaw detection of both the butt weld and fillet weld seam met the requirements, indicating good welding quality.

[0120] By setting different specifications for welding plates and welds of different sizes in Examples 1-3, and then obtaining the welding conditions in different groups, it is easy to see that the control method in this application can meet the requirements of weld flaw detection and effectively ensure good welding quality.

[0121] It should be noted that the control method in this application is applicable to (1) thick plate long weld seam structure unclear and root pass welding.

[0122] (2) For thick plate root cleaning welding, in addition to being applicable to root pass welding, this application can also be used for the first pass welding after root cleaning and grinding, where the weld groove width is uneven;

[0123] (3) Applicable to butt welding and fillet welding in one-time welding of thin plate long weld seam structure.

[0124] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates, characterized in that, include: Step 1: Welding bevel image processing: Before welding, the welding bevel of the entire thin plate long weld structure is scanned and processed. Step 2, Welding gap and misalignment positioning: Based on the results of image scanning and processing, determine the assembly dimensions of different parts of the long weld, and locate the welding gap and misalignment in different positions; Step 3: Welding process matching: Use welding process database software to determine the welding process specifications for different assembly dimensions; Step 4: Welding process closed-loop strategy generation: Based on assembly quality and welding process, match the corresponding welding process for different gaps and misalignment dimensions to generate a welding process closed-loop strategy; Step 5: Welding path and process generation: The generated welding path and process are sent to the robot and welding machine through the control system to execute the welding process; Step one includes: Step S11: Image preprocessing: Before welding, the welding groove of the entire thin plate long weld structure is scanned, and then the scanned image is preprocessed using an image processing algorithm; Step S12: Centerline extraction: The centerline of the laser stripes is quickly extracted using the required thinning algorithm; and boundary pixels that meet specific conditions are iteratively deleted to finally obtain the skeleton representation of the image. Step S13: Denoising algorithm after centerline extraction: Based on the physical continuity constraint of laser stripes, isolated noise points are eliminated through column scanning and neighborhood correlation analysis; Step S14: Obtain the gap variation area; Step S15: Obtain feature points of the gap change area: Extract feature points from the weld bevel to obtain image information of the weld bevel; Specifically, step S13 includes: Step S131: Input the skeletonized image obtained by line laser scanning; Step S132: Perform column scanning and determine whether there are pixels in the current column; if yes, proceed to step S133; if no, skip the column, repeat step S132, and perform scanning of the next column. Step S133: Determine whether the current column pixel is a single point; if yes, proceed to step S134; if no, proceed to step S135. Step S134: Directly save the pixel data of the current column and proceed to step S136; Step S135: After minimizing the distance with the previous column of points using the nearest neighbor selection strategy, proceed to step S136; Step S136: Output the filtered point set.

2. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, Before processing the welding bevel image in step one, it is necessary to assemble an arc-starting plate at the arc-starting position of the welding bevel for the welded structural component.

3. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, The image processing algorithm in step S11 can be any one of the following: grayscale processing algorithm, Gaussian filtering algorithm, or adaptive binarization method.

4. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, The column scan in step S132 is a column vector scan.

5. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, Step S135 includes: Nearest neighbor decision stage: When there are two or more candidate pixels in a column, the minimum distance criterion is used for filtering; First column processing: Select the first valid pixel in the column as the initial reference point; Subsequent column processing: Calculate the vertical distance between all candidate pixels in the current column and the reference point of the previous column, and select the point with the smallest distance as the new reference point.

6. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, Step S135 also includes two physical constraints: a vertical jump constraint and a trend maintenance constraint.

7. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, In step two, the three-dimensional point cloud obtained by scanning the welding groove is processed into a three-dimensional point cloud image, and then assembled and gap positioned. Combining the three-dimensional point cloud image processing method, its specific position and length in the weld are determined, and coordinates are marked.

8. The closed-loop control method for laser-arc hybrid welding process of long weld seams in thin plates according to claim 1, characterized in that, In step four, for long weld seam structures, the welding process and corresponding welding distance for different welding positions are determined, and a closed-loop welding process strategy is generated.

Images