Magnetic conformal positioning calibration block, system, and method for welding path planning

Abstract

This invention discloses a magnetically adhering conformal positioning calibration block, system, and method for welding path planning, belonging to the field of welding robot teaching technology. The calibration block includes a first substrate and a second substrate. Both substrates have a checkerboard calibration surface with diffuse reflection characteristics on their front sides and a magnetic adhering surface on their back sides. One end of the first and second substrates is hinged together by a movable hinge. An encoding area is provided on the front side of either the first or second substrate. The encoding area consists of at least two color blocks with high contrast colors, used to encode the calibration type attribute. This calibration block achieves adaptive conformal fitting with the bevel surface, exhibiting strong scene adaptability. The teaching method involves attaching the magnetically adhering conformal positioning calibration block to key weld positions. Combined with a binocular vision sensor, this allows for rapid identification and calculation of teaching points, significantly reducing weld teaching time and reliance on operator experience, thus achieving efficient generation of robot welding trajectories.

Description

Technical Field

[0001] This invention belongs to the field of welding robot teaching technology, specifically relating to a magnetically adhering conformal positioning calibration block, system, and method for welding path planning in unstructured scenarios. Background Technology

[0002] In high-end equipment manufacturing sectors such as heavy industry, shipbuilding, and aerospace, welding is a core manufacturing process, and its quality and efficiency directly determine product reliability and production cycle. Replacing manual welding with robots not only effectively avoids the health hazards to operators from harsh working conditions such as high temperatures, arc light, and fumes, but also ensures consistent weld formation through stable repeatability, making it an inevitable choice for the automation and intelligent transformation of welding. In the entire robotic welding process, welding trajectory planning is the core link determining operational accuracy and efficiency. Essentially, it involves acquiring the three-dimensional spatial information of key weld feature points to generate motion commands that the robot can execute, directly impacting welding quality and production efficiency.

[0003] Currently, robot welding trajectory planning mainly falls into three technical paths: teach-and-playback, offline programming, and intelligent teach-free path planning. Offline programming heavily relies on standard CAD models of the workpiece. However, in unstructured welding scenarios, workpieces commonly exhibit processing errors and assembly deviations, and often lack complete pre-set models, leading to a significant disconnect between the planned path and actual working conditions, requiring substantial manual correction. While intelligent teach-free path planning can autonomously identify weld seams through sensors, it lacks adaptability to complex bevel morphologies, strong arc light interference, and workpiece surface corrosion, thus hindering large-scale industrial application. Therefore, in unstructured welding scenarios, teach-and-playback remains the only stable trajectory planning method currently available. However, traditional teach-and-playback requires operators to calibrate the weld seam trajectory point-by-point using a teach pendant, resulting in high learning costs. Furthermore, in welding large structural components with discrete weld seams and multiple layers / passes, repeated adjustments to the robot's pose are necessary, with calibration times reaching several minutes, severely restricting production efficiency.

[0004] Against this backdrop, overcoming the efficiency bottleneck of traditional teaching and reproduction methods and achieving rapid calibration of key feature points in welds has become a core issue in improving the efficiency of robotic welding in unstructured scenarios. In existing technologies, weld key point calibration largely relies on manual teaching, which is cumbersome, time-consuming, and requires highly experienced operators, making it difficult to adapt to production needs. Therefore, there is an urgent need for a calibration device and method that can quickly and accurately acquire the three-dimensional coordinates of weld key points to achieve efficient generation of robotic welding trajectories. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a magnetically adhering conformal positioning calibration block, system, and method for welding path planning, enabling robot systems to capture the key points of weld seams more quickly, while reducing reliance on operator experience and achieving efficient generation of robot welding trajectories.

[0006] This solution achieves adaptive bonding with the workpiece surface through magnetically attached conformal positioning calibration blocks. Combined with binocular vision to calculate the three-dimensional spatial coordinates of the calibration blocks in real time, the calibration time for key points is reduced from minutes to seconds, significantly improving the efficiency of trajectory planning.

[0007] To achieve the above objectives, the technical solution adopted by this invention is as follows: a magnetically oriented conformal positioning calibration block for welding path planning, comprising a first substrate and a second substrate. Both the first and second substrates have a checkerboard calibration surface with diffuse reflection characteristics on their front sides, and magnetic surfaces for adsorbing metal workpieces on their back sides. One end of the first and second substrates is hinged together by a movable hinge. An encoding area is provided on the front side of either the first or second substrate. The encoding area consists of at least two color blocks with high contrast colors, used to encode the calibration type attribute. The two high-contrast colors are such as black and white, where one color represents binary 0 and the other represents binary 1.

[0008] The magnetically attached conformal positioning calibration block provided by this solution can conformally fit the surface of the workpiece, so that the feature teaching points calculated by the magnetically attached conformal positioning calibration block can directly correspond to the key feature points of the weld, reducing the manual calculation process.

[0009] In the above scheme, the lack of clear start and end markers can easily lead to sequence errors or ambiguities during decoding. Therefore, as a further improvement to the above scheme, the encoding area includes an encoding start end, an encoding end, and a custom encoding area; the encoding start end is fixedly set to a single color block, and the encoding end is fixedly set to a second color block; the custom encoding area is located between the encoding start end and the encoding end and consists of at least one configurable color block. The encoding start end and encoding end provide clear decoding direction indications, eliminate sequence ambiguity, and ensure decoding accuracy.

[0010] If the coding area can only identify the location and cannot express the process semantics (such as starting point, trajectory type, edge point, etc.), subsequent program generation will still require manual intervention, which is not conducive to the automatic generation of planning paths. To overcome this problem, as a further improvement to the above solution, the coding custom area consists of three color blocks, which form eight codes through the binary combination of two colors. Each code corresponds to the following meanings: welding starting point, linear interpolation trajectory point, circular interpolation trajectory point, spline interpolation trajectory point, left edge point of bevel, right edge point of bevel, welding ending point, and welding reserved position. By corresponding the three-bit binary code one-to-one with the welding process attributes, conditions are created for the automatic generation of programs with process logic.

[0011] Ordinary calibration surfaces are prone to specular reflection, making feature point extraction difficult under arc light illumination and affecting visual positioning accuracy. To overcome this problem, the checkerboard calibration surface is made of alumina with a black and white checkerboard pattern. The diffuse reflection properties of alumina improve the accuracy and robustness of checkerboard corner point extraction.

[0012] The first and second substrates are made of lightweight aluminum alloy profiles. While ensuring structural rigidity, the overall weight is light and easy to use.

[0013] This invention also provides a welding path planning system, including the aforementioned magnetic conformal positioning calibration block, binocular vision sensor, robot, and controller. The magnetic conformal positioning calibration block is attached to the key weld points of the workpiece to be taught. The binocular vision sensor is installed at the end of the robot to acquire images of the magnetic conformal positioning calibration block. The controller calculates the three-dimensional coordinates of the teaching point corresponding to each magnetic conformal positioning calibration block in the sensor coordinate system based on the acquired images. According to the encoding attributes of each magnetic conformal positioning calibration block and the coordinate transformation matrix between the binocular sensor and the world coordinate system, the coordinates of each teaching point are transformed to the world coordinate system, and a robot welding program containing welding trajectory and process instructions is generated. This system realizes automatic capture, calculation, and program generation of key weld points, requiring only manual operation when the magnetic conformal positioning calibration block is attached, thus improving automation and efficiency.

[0014] The present invention also provides a welding trajectory teaching method based on the above-mentioned magnetically oriented conformal positioning calibration block, comprising the following steps:

[0015] S1. Arrange calibration blocks: attach the corresponding type of magnetic suction conformal positioning calibration block at the corresponding position of the workpiece weld, and make the movable hinge of the magnetic suction conformal positioning calibration block adaptively bend according to the weld bevel angle, so that the two bases are respectively attached to the two sides of the bevel plane.

[0016] S2. Calculate the coordinates of the teaching point: Control the robot to move so that the binocular vision sensor fixed to the robot can collect images of the magnetic conformal positioning calibration block, and calculate the three-dimensional coordinates of the teaching point of each magnetic conformal positioning calibration block in the sensor coordinate system based on the images.

[0017] S3. Generate robot executable program: Transform the three-dimensional coordinates of the teaching points collected at different locations in the sensor coordinate system to the world coordinate system, and generate a robot executable welding program containing welding trajectory and process instructions based on the encoding attributes of each magnetic conformal positioning calibration block.

[0018] This method, by attaching a magnetic conformal positioning calibration block to the key position of the weld, combined with a binocular vision sensor, can quickly identify and calculate the teaching points. The calibration, calculation and program generation are all automated, which greatly reduces the teaching time of multi-layer, multi-pass, discrete welds in unstructured scenarios and improves production efficiency.

[0019] Furthermore, step S2 specifically includes:

[0020] S21. Target recognition: Using a deep learning model, identify the two checkerboard calibration surfaces and the coding area of ​​the magnetic conformal positioning calibration block in the binocular image;

[0021] S22. Decoding: Perform color segmentation on the encoded area, identify the start and end of the encoding, extract the binary code of the customized encoding area, and determine the type attribute of the current magnetic conformal positioning calibration block.

[0022] S23, checkerboard corner point recognition: Extract the corner point pixel coordinates of the checkerboard calibration surface of the first substrate and the second substrate, and calculate the three-dimensional coordinates of each corner point in the sensor coordinate system through the principle of binocular vision;

[0023] S24. Calculation of teaching point coordinates: Using the three-dimensional coordinates of the corner points of the checkerboard calibration surfaces of the first and second bases, fit the planes P1 and P2 of the two checkerboard calibration surfaces and their normal vectors respectively; translate P1 and P2 along the opposite direction of their normal vectors by the thickness d of the magnetic suction conformal positioning calibration block to obtain the planes P1' and P2' where the magnetic suction surface is located; calculate the intersection line of P1' and P2'; project the checkerboard corner point near the starting end of the encoding area onto P1' to obtain the projection point, and then find the foot of the perpendicular from the projection point to the intersection line. This foot of the perpendicular is the teaching point P.

[0024] This scheme achieves precise mapping from the pose of the calibration block to the position of the weld, and realizes high-precision calculation of the three-dimensional coordinates of the key points of the weld, providing a reliable spatial positioning reference for robot welding trajectory planning.

[0025] Furthermore, in S24, the checkerboard corner closest to the starting end of the encoding area is one of the three corner points in the first or second base that are closest to the end where the encoding area is located. By specifying the corner point closest to the starting end of the encoding area as a reference, the endpoint of the teaching point is uniquely determined, ensuring the consistency of the calculation.

[0026] Furthermore, in step S3, the encoded attributes of the magnetically attached conformal positioning calibration block include a welding start point, a welding end point, one or more trajectory interpolation points, and N bevel feature points. The trajectory interpolation points are linear interpolation points, circular interpolation points, or spline interpolation points; N is an even number equal to or greater than 0; the bevel feature points include the left edge point or the right edge point of the bevel. Arc initiation and extinguishing commands are automatically generated based on the welding start point and welding end point, and linear interpolation, circular interpolation, or spline interpolation commands are automatically generated according to the trajectory point type. By decoding the encoded attributes of the calibration block, the corresponding interpolation commands are automatically matched, facilitating the automatic generation of motion programs. In multi-layer, multi-pass application scenarios, combined with the bevel feature points collected by the magnetically attached conformal positioning block, the system can automatically generate multi-layer, multi-pass welding operation programs in batches.

[0027] Furthermore, in step S3, when transforming the teaching point coordinates from the sensor coordinate system to the world coordinate system, the robot end pose matrix T1 at the time of shooting and the hand-eye relationship transformation matrix T2 are used for transformation.

[0028] The beneficial effects of this invention are as follows: The magnetic conformal positioning calibration block disclosed in this invention, through the combination design of the magnetic surface and the movable hinge, can achieve adaptive conformal fitting to irregular bevel surfaces with processing and assembly errors, exhibiting extremely strong scene adaptability. Its magnetic layer on the back can quickly adhere to the surface of the metal workpiece using magnetic force, resulting in rapid fitting and shortening the calibration block placement time. The checkerboard calibration surface is made of alumina material, possessing diffuse reflection characteristics, and the pattern is a high-contrast black and white checkerboard, significantly improving the accuracy and robustness of the binocular vision sensor extraction. The calibration block integrates an encoding area composed of high-contrast color blocks, fixing the start end, end end, and intermediate custom encoding area to form a 3-bit binary code, which can represent various different process attributes. This maps the physical position of the calibration block to its semantic meaning in the welding process, allowing the system to decode and determine the motion command and welding action to be executed at that point, eliminating the need for manual marking or additional programming.

[0029] The welding trajectory teaching method disclosed in this invention eliminates the need for precisely moving the robot's end effector point by point to the welding trajectory position. By attaching a magnetic conformal positioning calibration block to the key locations of the weld, and in conjunction with a binocular vision sensor, the teaching points can be quickly identified and calculated. This significantly reduces the teaching time for multi-layer, multi-pass, and discrete welds in unstructured scenarios, while also reducing reliance on operator experience and achieving efficient generation of robot welding trajectories. A high-precision alumina checkerboard pattern with diffuse reflection properties is used as the visual feature carrier, combined with binocular vision technology, to achieve high-precision calculation of the three-dimensional coordinates of key weld points, providing a reliable spatial positioning reference for robot welding trajectory planning. Attached Figure Description

[0030] Figure 1 This is a front view of the magnetically oriented conformal positioning calibration block disclosed in this invention;

[0031] Figure 2 This is a schematic diagram of the back side of the magnetically oriented conformal positioning calibration block disclosed in this invention;

[0032] Figure 3 This is a schematic diagram of the coding area of ​​the magnetically adhering conformal positioning calibration block disclosed in this invention;

[0033] Figure 4 This is a schematic diagram of the teaching points of the magnetically adhering conformal positioning calibration block disclosed in this invention;

[0034] Figure 5 This is a schematic diagram of observation using a binocular vision sensor.

[0035] Figure 6 This is a diagram showing the corner point numbering of the magnetically attracted conformal positioning calibration block chessboard grid disclosed in this invention;

[0036] Figure 7 This is a schematic diagram illustrating the application of the magnetically pleasing conformal positioning calibration block disclosed in this invention in a multi-segment straight fillet weld scenario.

[0037] Figure 8 This is a schematic diagram illustrating the application of the magnetically pleasing conformal positioning calibration block disclosed in this invention in a straight V-groove welding scenario.

[0038] Figure 9 This is a schematic diagram illustrating the application of the magnetic conformal positioning calibration block disclosed in this invention in a curved V-groove welding scenario.

[0039] In the diagram, 1-Magnetic conformal positioning calibration block; 11-First base; 12-Second base; 13-Movable hinge; 14-Checkerboard calibration surface; 15-Magnetic surface; 16-Encoding area; 161-Encoding start end; 162-Encoding end end; 163-Custom encoding area; 2-Binocular vision sensor; 3-Robot. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0041] like Figure 1 and Figure 2 As shown, the magnetic conformal positioning calibration block for welding path planning includes a first base 11 and a second base 12.

[0042] In this embodiment, both the first substrate 11 and the second substrate 12 are made of aluminum alloy profiles. The lightweight and high hardness of aluminum alloy profiles ensure structural rigidity (able to withstand magnetic attraction without deformation) while being lightweight and easy for operators to hold.

[0043] Both the first substrate 11 and the second substrate 12 have a checkerboard calibration surface 14 with diffuse reflection characteristics on their front surfaces. This checkerboard calibration surface 14 is made of aluminum oxide and forms a high-precision black and white checkerboard pattern through precision printing or laser engraving. It has diffuse reflection characteristics, significantly improving the accuracy and robustness of the binocular vision sensor in extracting feature points. In this embodiment, the checkerboard pattern of both the first substrate 11 and the second substrate 12 is a 2×2 grid array (with a total of 9 corner points, such as...). Figure 6 (As shown).

[0044] Both the first substrate 11 and the second substrate 12 have magnetic attraction surfaces 15 on their back sides for quick adsorption and adhesion to the surface of the metal workpiece. The magnetic attraction surfaces 15 may be composed of an array of neodymium magnets embedded in a groove on the back side of the substrate.

[0045] The magnetic conformal positioning and calibration block adopts a movable hinge structure in the middle, that is, one end of the first base 11 and the second base 12 are hinged by a movable hinge 13, so that the magnetic conformal positioning and calibration block can bend adaptively according to the bevel angle of the workpiece, and realize conformal fitting of the weld surface at any angle.

[0046] An encoding area 16 is provided on the front side of either the first substrate 11 or the second substrate 12. In this embodiment, the encoding area 16 is located on the first substrate 11, and the following description will assume that the encoding area 16 is located on the first substrate 11. The encoding area 16 consists of at least two color blocks with high contrast colors, used to encode the calibration type attribute. One color represents binary 0, and the other color represents binary 1. The high contrast colors can be black and white, red and green, blue and yellow, etc. This embodiment uses black and white to adapt to complex lighting environments. Black is decoded as 0, and white is decoded as 1. Different combinations represent different calibration types of the magnetic conformal positioning calibration block. Figure 3 As shown, the coding area 16 specifically includes:

[0047] Encoding start end 161: Fixed as a white block, decoded as 0, used to indicate the starting direction of decoding;

[0048] Encoding termination terminal 162: Fixed as a black block, decoded as 1, used to indicate the termination direction of decoding;

[0049] Custom Encoding Area 163: Consists of three configurable color blocks, each of which can be set to black or white. When set to black, it is decoded as 1; when set to white, it is decoded as 0. The three color blocks form a 3-bit binary number, resulting in 8 possible encoding combinations.

[0050] The eight encoding combinations and their corresponding calibration block type attributes are shown in Table 1 below:

[0051] Table 1: Correspondence between Encoding Combinations and Calibration Block Type Attributes

[0052]

[0053] Before attaching the calibration blocks to the workpiece, the operator configures the three color blocks of the coded custom area 163 to the corresponding colors according to the welding process plan by changing the pre-made color blocks, turning the mechanical switch or pasting colored stickers, thereby assigning semantic labels to each calibration block.

[0054] Teaching points of magnetic conformal positioning calibration blocks (e.g.) Figure 4 Point P in the diagram is the endpoint on the line of intersection between plane P1' on the back of the first substrate 11 and plane P2' on the back of the second substrate 12, near the encoding start end 161.

[0055] After the magnetic surface 15 is attached to the surface of the metal workpiece, P1' and P2' correspond to the surface planes on both sides of the workpiece bevel, and their intersection line corresponds to the edge line of the workpiece surface (i.e., the root line of the weld). The endpoint near the encoding start end 161 corresponds precisely to the key feature points in the weld trajectory (such as the weld start point, inflection point, and end point). In this way, a precise geometric correspondence is established between the physical placement of the calibration block and the actual key points of the weld. By embedding the calculation rules of the teaching points into the attributes of the calibration block, the vision system can directly and automatically calculate the three-dimensional coordinates of the key points of the weld according to these rules, without manual calculation, achieving the effect of "instant knowledge upon attachment".

[0056] The following section describes in detail the welding trajectory teaching method using the aforementioned magnetic conformal positioning marker block, with specific welding examples as examples.

[0057] Example 1:

[0058] This embodiment uses Figure 7 The example shown is a multi-segment straight fillet weld. Figure 7As shown, two steel plates are welded vertically to form a fillet weld. The weld trajectory consists of two straight lines. The key points of the weld trajectory include the weld start point A, the inflection point B, and the weld end point C. The inflection point B is the intersection of the two straight lines.

[0059] Step S1: Arrange magnetic conformal positioning calibration blocks

[0060] According to the welding process plan, the operator attaches the corresponding type of calibration block to the corresponding position on the workpiece bevel. In this embodiment, the calibration block coded 000 is attached to the weld start point A. Then, the calibration block coded 001 is attached to the inflection point B. The calibration block coded 110 is attached to the weld end point C. When attaching the calibration blocks, the movable hinge 13 of the calibration block is adaptively bent to 90° to adapt to the bevel angle, and the two magnetic suction surfaces 15 are tightly attached to the two side planes of the bevel.

[0061] Step S2: Calculate the coordinates of the teaching point

[0062] like Figure 5 As shown, the binocular vision sensor 2 is fixed to the end of the robot 3, controlling the robot 3 to move so that the binocular vision sensor 2 can sequentially and clearly capture the magnetically attached conformal positioning calibration blocks 1 at three corresponding positions: the weld start point A, the inflection point B, and the weld end point C. The sensor also calculates the pose of the teaching point for each magnetically attached conformal positioning calibration block 1. Taking the capture of the magnetically attached conformal positioning calibration block 1 (coded 001) at the inflection point B as an example, the specific acquisition and calculation process is as follows:

[0063] S21: Target Recognition

[0064] Using pre-trained deep learning models such as YOLO, the two checkerboard calibration surfaces 14 and the coding area 16 on the magnetic conformal positioning calibration block 1 are identified in the binocular images.

[0065] S22: Decoding

[0066] First, locate the white marker block at the start of encoding 161 and the black marker block at the end of encoding 162 to determine the range of encoding area 16 and the decoding order.

[0067] Then, the color information of the calibration block in the custom coding area 163 is extracted sequentially from the encoding start end 161 to the encoding end 162, and converted into corresponding binary numbers to obtain a 3-bit binary code, thereby identifying the type of the positioning block. In this embodiment, the color information is white, white, and black, and the decoding yields the binary sequence 001. Looking up Table 1, we find that the type of the calibration block is a welding trajectory point (linear interpolation).

[0068] S23: Chessboard Corner Point Recognition

[0069] The pixel coordinates of nine corner points (LP1, LP2, …, LP9) of the checkerboard pattern on the first substrate 11 and nine corner points (RP1, RP2, …, RP9) of the checkerboard pattern on the second substrate 12 are extracted. LP1, LP2, and LP3 are points closer to the encoding start point 161. Since the checkerboard pattern is a high-contrast black and white pattern, a threshold-based segmentation and edge extraction method can efficiently extract the pixel coordinates of the corner points. Specifically: Checkerboard images of the first substrate 11 and the second substrate 12 are acquired; adaptive threshold segmentation is performed on the acquired images to binarize the black and white regions; then, edge extraction algorithms (such as Canny edge detection) are used to obtain the boundary lines between the black and white squares; finally, the intersection points of each boundary line are calculated, which are the pixel coordinates of the corner points within the checkerboard pattern. After obtaining the pixel coordinates of each corner point, the checkerboard corner points in the binocular image are matched using the principle of binocular vision to obtain the three-dimensional coordinates of each checkerboard corner point in the sensor coordinate system.

[0070] S24: Calculation of teaching point coordinates

[0071] First, using the three-dimensional coordinates of the nine corner points on the first substrate 11 in the sensor coordinate system, the checkerboard plane P1 and normal vector of the first substrate 11 are fitted; using the three-dimensional coordinates of the nine corner points on the second substrate 12 in the sensor coordinate system, the checkerboard plane P2 and normal vector of the second substrate 12 are fitted.

[0072] Next, P1 is translated along the opposite direction of the normal vector by the thickness d of the calibration block to obtain the plane P1' containing the magnetic attraction surface 15; P2 is translated along the opposite direction of the normal vector by the thickness d of the calibration block to obtain the plane P2' containing the magnetic attraction surface 15.

[0073] Then, calculate the intersection of the two planes P1' and P2';

[0074] Finally, the corner point (such as LP1) of the chessboard grid of the first base 11 that is close to the encoding start end 161 is projected to P1' to obtain the projection point LP1'. The foot of the perpendicular from LP1' to the intersection line l is found, and the foot of the perpendicular is the teaching point P.

[0075] Repeat the above process to obtain the teaching points of each calibration block at positions A and C, and obtain the three-dimensional coordinates of each teaching point in the sensor coordinate system.

[0076] Step S3: Generate the robot executable program

[0077] First, based on the robot's end-effector pose matrix and hand-eye relationship transformation matrix at each shooting moment, the three-dimensional coordinates of all teaching points in the sensor coordinate system are transformed from the sensor coordinate system to the world coordinate system. This process is a mature existing technology and will not be elaborated here.

[0078] Then, select one welding start point (000) and one or more trajectory points (001, 010, 011) in sequence. In this embodiment, one trajectory point (001) and one welding end point (110) are used. Set the corresponding interpolation instructions and execute the welding arc start and end commands to automatically generate the robot program framework. The operator then sets specific welding parameters such as welding speed, welding current, voltage, and oscillation parameters to generate the final robot welding program.

[0079] For the fillet weld with two straight segments and one corner in this embodiment, traditional manual teaching requires an operator to teach point by point for about 2-4 minutes (about 1 minute per point). Using the method of this invention, the operator only needs to align three calibration blocks (taking about 4 seconds), while the robot automatically takes pictures and calculates (taking about 5 seconds per point, for a total of 35 seconds), for a total time of about 39 seconds, significantly improving efficiency. Simultaneously, the teaching point at corner B precisely corresponds to the actual intersection of the fillet weld root line, avoiding trajectory deviations caused by visual errors during manual teaching.

[0080] Example 2: Teaching the welding trajectory in a straight V-groove weld scenario

[0081] This embodiment uses Figure 8 Taking the straight V-groove weld as an example, this paper details the application of the present invention in the scenario of grooved welds (multi-layer, multi-pass welds). Figure 8 As shown, the workpiece consists of two 20mm thick steel plates joined together, with a V-groove (groove angle 45°), and the weld trajectory is a straight line (500mm in length). For multi-layer, multi-pass welding, the key points of the weld trajectory include: welding start point A, left edge of the groove L, right edge of the groove R, and welding end point D. The teaching process is similar to that of Example 1, except that the position selection of the magnetic conformal positioning calibration block is different.

[0082] By setting edge calibration blocks coded 100 and 101 on the bevel surface, the coordinates of key bevel points are obtained, and the three-dimensional geometry of the bevel is automatically calculated. Combined with a multi-layer, multi-pass path planning algorithm, a multi-layer, multi-pass welding program can be generated at once, further improving teaching efficiency. Meanwhile, due to the adaptive bending of the movable hinge, the calibration blocks can closely fit bevel surfaces at different angles, eliminating the need to prepare multiple calibration blocks with different angles.

[0083] Example 3: Teaching the welding trajectory in a curved V-groove weld scenario

[0084] This embodiment uses Figure 9 Taking the curved V-groove weld as an example, this paper details the application of the present invention in the scenario of grooved welds (multi-layer, multi-pass welds). Figure 9As shown, the workpiece is a cylinder and end cap with a circular arc weld, a V-shaped bevel, and the weld trajectory is a circular arc. The key points of the weld trajectory include: welding start point A, the midpoint of the circular arc segment B, the left edge point L of the bevel, the right edge point R of the bevel, and the welding end point D.

[0085] Step S1: Arrange magnetic conformal positioning calibration blocks

[0086] Place the calibration block coded 000 at the weld start point A. Then place the calibration block coded 010 at the midpoint B of the arc segment. Place the calibration block coded 100 at the left edge point L of the bevel. Place the calibration block coded 101 at the left edge point R of the bevel. Finally, place the calibration block coded 110 at the weld end point C.

[0087] When the calibration block is attached, the movable hinge 13 of the calibration block is adaptively bent to adapt to the bevel angle, and the two magnetic surfaces 15 are tightly attached to the two sides of the bevel plane.

[0088] It should be noted that for curved welds, the calibration block itself cannot be bent (the movable hinge only provides one degree of bending freedom and cannot bend along the curve). Therefore, when the calibration block is attached to the curved weld, only the centerline of the calibration block (along the welding direction) is tangent to the weld. However, since the size of the calibration block is much smaller than the radius of curvature of the curved weld, it is locally approximately a straight line, and the teaching point after attachment can still accurately characterize the weld root point at that location. For curves with a small radius of curvature (e.g., R < 100 mm), the calibration blocks can be arranged more densely along the curve trajectory.

[0089] Step S2: Calculate the coordinates of the teaching point

[0090] The robot 3 is controlled to move so that the binocular vision sensor 2 can sequentially and clearly capture the magnetically attached conformal positioning calibration blocks 1 at the corresponding positions of the welding start point A, the midpoint of the arc segment B, the left edge point L of the bevel, the right edge point R of the bevel, and the welding end point D. The pose of the teaching points of each magnetically attached conformal positioning calibration block 1 is calculated according to steps S21-S24 above, obtaining the three-dimensional coordinates of each viewpoint in the sensor coordinate system.

[0091] Step S3: Generate the robot executable program

[0092] First, based on the robot's end-effector pose matrix and hand-eye relationship transformation matrix at each shooting moment, the three-dimensional coordinates of all teaching points in the sensor coordinate system are transformed from the sensor coordinate system to the world coordinate system. This process is a mature existing technology and will not be elaborated here.

[0093] Then, by sequentially selecting one welding start point (000), one circular interpolation trajectory point (010), and one welding end point (110), and coordinating with two bevel edge feature points (100, 101), and setting the corresponding interpolation commands and welding arc initiation and extinguishing actions, a multi-layer, multi-pass robot program framework can be automatically generated. The operator can then set specific welding parameters such as welding speed, welding current, voltage, and oscillation parameters to generate the final multi-layer, multi-pass robot welding program.

Claims

1. A magnetic conformal positioning calibration block for welding path planning, characterized in that, It includes a first substrate (11) and a second substrate (12). The front side of the first substrate (11) and the second substrate (12) are provided with a checkerboard calibration surface (14) with diffuse reflection characteristics, and the back side is provided with a magnetic suction surface (15) for adsorbing metal workpieces. One end of the first substrate (11) and the second substrate (12) is hinged by a movable hinge (13); an encoding area (16) is provided on the front of the first substrate (11) or the second substrate (12), the encoding area (16) is composed of at least two color blocks with high contrast colors, and is used to encode the type attribute.

2. The magnetic conformal positioning calibration block for welding path planning according to claim 1, characterized in that, The encoding area (16) includes an encoding start end (161), an encoding end end (162), and an encoding custom area (163). The encoding start end (161) is fixedly set to one color block, and the encoding end end (162) is fixedly set to another color block; the encoding custom area (163) is located between the encoding start end (161) and the encoding end end (162), and is composed of at least one color block with configurable color.

3. The magnetic conformal positioning calibration block for welding path planning according to claim 2, characterized in that, The coded custom area (163) consists of three color blocks, which form eight codes through binary combination of two colors. Each code corresponds to the following meanings: welding start point, linear interpolation trajectory point, circular interpolation trajectory point, spline interpolation trajectory point, left edge point of bevel, right edge point of bevel, welding end point, and welding reserved position.

4. The magnetic conformal positioning calibration block for welding path planning according to claim 1, characterized in that, The checkerboard marking surface (14) is made of aluminum oxide and has a black and white checkerboard pattern; the first substrate (11) and the second substrate (12) are made of aluminum alloy profiles.

5. A welding path planning system, characterized in that, The system includes a magnetically attached conformal positioning calibration block (1) for welding path planning as described in any one of claims 1-4, a binocular vision sensor (2), a robot (3), and a controller; the magnetically attached conformal positioning calibration block (1) is attached to the key weld points of the workpiece to be taught; the binocular vision sensor (2) is installed at the end of the robot (3) to collect images of the magnetically attached conformal positioning calibration block (1); the controller calculates the three-dimensional coordinates of the teaching point corresponding to each magnetically attached conformal positioning calibration block (1) in the sensor coordinate system based on the collected images of the magnetically attached conformal positioning calibration block (1); and converts the coordinates of each teaching point to the world coordinate system according to the encoding attributes of each magnetically attached conformal positioning calibration block (1), and generates a robot welding program containing welding trajectory and process instructions.

6. A welding trajectory teaching method based on the welding path planning system of claim 5, characterized in that, Includes the following steps: S1. Arrange calibration blocks: attach the corresponding type of magnetic suction conformal positioning calibration block (1) to the corresponding position of the workpiece weld, and make the movable hinge (13) of the magnetic suction conformal positioning calibration block (1) bend adaptively according to the weld bevel angle, so that the two bases are respectively attached to the two sides of the bevel plane. S2. Calculate the coordinates of the teaching point: Control the robot (3) to move so that the binocular vision sensor (2) fixed to the robot (3) can collect the image of the magnetic conformal positioning calibration block (1) and calculate the three-dimensional coordinates of the teaching point of each magnetic conformal positioning calibration block (1) in the sensor coordinate system based on the image. S3. Generate robot executable program: Transform the three-dimensional coordinates of the teaching points collected at different locations in the sensor coordinate system to the world coordinate system, and generate a robot executable welding program containing welding trajectory and process instructions based on the encoding attributes of each magnetic conformal positioning calibration block (1).

7. The welding trajectory teaching method according to claim 6, characterized in that, Step S2 specifically includes: S21. Target recognition: Using a deep learning model, the two checkerboard calibration surfaces (14) and the coding area (16) of the magnetic conformal positioning calibration block (1) are identified in the binocular image. S22, Decoding: Perform color segmentation on the encoding area (16), identify the start and end of the encoding, extract the binary code of the encoding custom area, and determine the type attribute of the current magnetic suction conformal positioning calibration block (1); S23, checkerboard corner point recognition: extract the corner pixel coordinates of the checkerboard calibration surface (14) of the first substrate (11) and the second substrate (12), and calculate the three-dimensional coordinates of each corner point in the sensor coordinate system through the principle of binocular vision; S24. Calculation of teaching point coordinates: Using the three-dimensional coordinates of the corner points of the checkerboard calibration surface (14) of the first base (11) and the second base (12), fit the planes P1 and P2 of the two checkerboard calibration surfaces (14) and their normal vectors respectively; translate P1 and P2 along the opposite direction of their normal vectors by the thickness d of the magnetic suction conformal positioning calibration block (1) to obtain the planes P1' and P2' where the magnetic suction surface (15) is located; calculate the intersection line of P1' and P2'; project the checkerboard corner point near the starting end of the coding area (16) onto P1' to obtain the projection point, and then find the foot of the perpendicular from the projection point to the intersection line. The foot of the perpendicular is the teaching point P.

8. The welding trajectory teaching method according to claim 7, characterized in that, In S24, the corner point of the chessboard near the starting end of the coding area (16) is one of the three corner points in the first base (11) or the second base (12) that are closest to the end where the coding area (16) is located.

9. The welding trajectory teaching method according to claim 6, characterized in that, In step S3, the encoded attributes include the welding start point, the welding end point, one or more trajectory interpolation points, and N bevel feature points. The trajectory interpolation points are linear interpolation trajectory points, circular interpolation trajectory points, or spline interpolation trajectory points, and N is an even number equal to 0 or greater than 0. Arc ignition and arc extinguishing commands are automatically generated based on the welding start point and the welding end point, and linear interpolation, circular interpolation, or spline interpolation commands are automatically generated according to the trajectory point type.

10. The welding trajectory teaching method according to claim 6, characterized in that, In step S3, when transforming the teaching point coordinates from the sensor coordinate system to the world coordinate system, the robot end-effector pose matrix T at the time of shooting is used for the transformation.

Images