A robotically intelligent gouging welding repair die station and method of controlling the same

Abstract

The present application relates to the technical fields of intelligent gas gouging and welding repair, and discloses a robot intelligent gas gouging welding repair mold workstation and a control method thereof, which comprises a robot, a welding machine, a 3D vision camera, a welding gun, a gas gouging power supply, a gas gouging gun and a gun changing disc; the gun changing disc is composed of one male module and two female modules, and the two female modules are female module A and female module B respectively. Through the six-axis movement of the robot, the 3D vision camera and the gas gouging gun are driven to move, and the workpiece is coarsely positioned by combining the autonomous photographing and point cloud processing of the 3D vision camera on the mold contour. Then, the three-dimensional coordinates of the mold cavity and the adaptive space posture of the gas gouging gun are automatically identified by the secondary photographing and point cloud analysis of the 3D vision camera on the mold cavity, so as to realize the accurate positioning of the gas gouging path and the intelligent matching of the gas gouging gun posture, shorten the gas gouging cycle of single workpiece, and solve the problem of low gas gouging efficiency of the mold.

Description

Technical Field

[0001] This invention relates to the field of intelligent air gouging and welding repair technology, specifically to a robotic intelligent air gouging welding repair mold workstation and its control method. Background Technology

[0002] Intelligent air gouging and welding repair technology, as one of the core technologies in the machinery manufacturing industry, is widely used in critical and vulnerable fields such as molds, mine car buckets, coal processing, and heavy containers, where there is direct contact with blanks and raw materials with relative movement. As the core load-bearing component of automatic stamping, the mold is made of high-strength mold steel, and its contour shape has the complexity of a three-dimensional curved surface. The welding quality directly determines the mechanical properties of the entire mold cavity and the quality stability of the product.

[0003] Currently, the air gouging and welding production of molds is mainly achieved through a fixed workbench and special tooling fixtures. The workpiece needs to be placed precisely in the tooling fixture according to the preset position. The operator clamps the workpiece manually, first uses a carbon gouging tool to cut and remove material from the cavity of the mold at the damaged area, then manually grinds the carbon gouging surface, then manually preheats it with a flame, and then manually welds the additive material to fill the predetermined size range. Finally, the machine tool processes and removes the surface of the welding additive material to the size requirements marked on the drawing, thus completing the mold repair work for the damaged part. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a robotic intelligent air gouging and welding repair mold workstation and its control method, which solves the problems of poor working environment and low efficiency in current air gouging and welding of molds.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a robotic intelligent air gouging welding repair mold workstation, comprising: a robot, a welding machine, a 3D vision camera, a welding torch, an air gouging power supply, an air gouging torch, and a torch changing disc;

[0006] The gun-changing disc consists of one male module and two female modules, namely female module A and female module B.

[0007] The public module is installed at the end of the sixth axis of the robot;

[0008] The welding torch is fixed to the female module A of the torch changing plate, and can be detached and installed on one side of the robot through the connection and cooperation of the male module and the female module A.

[0009] The air planer gun is fixed to the female module B of the tool changing disc, and can be detached and installed on one side of the robot through the connection and cooperation of the male module and the female module B.

[0010] The 3D vision camera is set on the other side of the robot to take the first picture of the mold outline, obtain the spatial point cloud of the mold, process it, obtain the relative position value of the mold's outer reference distance from the robot coordinates, and obtain the spatial position data of the mold reference point in the X, Y, and Z directions as the reference coordinate values.

[0011] The 3D vision camera is also used to take a second picture of the mold outline, obtain the spatial point cloud of the mold cavity, process it, obtain the spatial position data of the mold cavity in the X, Y and Z directions, and calculate the spatial posture of the corresponding air gouging gun and welding gun.

[0012] By adopting the above technical solutions, a robot drives a 3D vision camera, a gouging gun, and a welding gun to move. Combined with the autonomous photography and point cloud processing of the mold contour by the 3D vision camera, the three-dimensional coordinate values ​​of the workpiece reference points can be quickly obtained without manual positioning or special tooling, completing the rough positioning of the workpiece. Then, the 3D vision camera autonomously photographs and processes the mold cavity a second time. By analyzing the point cloud within the effective area of ​​the photograph, the three-dimensional coordinates of the mold cavity and the spatial posture of the matching gouging gun and welding gun are automatically identified. This achieves precise positioning of the mold cavity and intelligent matching of the posture of the gouging gun and welding gun, shortening the welding cycle of a single workpiece and reducing the ineffective time consumption caused by manual intervention. It realizes the automated and intelligent continuous production of gouging and welding of mold cavities, solving the problem of low efficiency in gouging and welding of molds.

[0013] Preferably, the spatial orientation of the air gouging gun and welding gun includes coordinate values ​​of six degrees of freedom: X, Y, Z, Ra, Ry, and Rz.

[0014] The robot is a six-axis robot.

[0015] Preferably, obtaining the relative position values ​​of the mold's outer shape reference distance to the robot coordinates, and obtaining the reference coordinate values ​​of the mold's reference point in the X, Y, and Z directions, specifically includes the following steps:

[0016] A 3D vision camera emits a surface-vibrating laser beam to irradiate the surface of the mold. The robot moves in six axes and takes continuous pictures from a preset initial point to collect spatial point cloud data of the target area of ​​the mold's outline.

[0017] The spatial point cloud data is processed by outlier removal based on the improved RANSAC algorithm. Outliers are removed through least squares iterative optimization. Then, Euclidean distance clustering analysis is used to divide the data into blocks, resulting in several continuous local point cloud blocks.

[0018] Based on the Bézier curve fitting theory in differential geometry, contour features are extracted for each local point cloud block, and the parametric equation of the contour curve is constructed by solving the coordinates of the control points of the Bézier curve.

[0019] By determining the continuity of curve segments, the parametric equations of each local point cloud block are spliced ​​into a continuous contour curve, and the effective right-angled sides of the contour that meet the preset benchmark distance threshold are locked.

[0020] Based on the parametric equation of the right-angled side of the effective contour, the linear equation of the reference side is derived. Combined with the origin coordinate of the robot coordinate system, the relative position value of the mold's outer reference distance from the robot coordinate system is calculated, and the spatial position data of the mold's reference point in the X, Y, and Z directions is output.

[0021] Preferably, the improved RANSAC algorithm is an algorithm that introduces an adaptive inlier threshold adjustment mechanism, which dynamically corrects the inlier threshold according to the point cloud density;

[0022] The block processing involves setting a clustering radius based on the size characteristics of the mold cavity, and dividing the collected point cloud of the large workpiece into several continuous local point cloud blocks.

[0023] The coordinates of the control points of the Bézier curve were obtained by the least squares iterative method.

[0024] Preferably, obtaining the spatial position data of the mold cavity in the X, Y, and Z directions, and calculating the spatial orientation of the corresponding air gouging gun and welding gun, specifically includes the following steps:

[0025] Based on the light scattering compensation model, three-dimensional spatial point cloud data of the air-gouging area is collected by taking pictures with a 3D vision camera.

[0026] Wavelet transform is used to denoise the three-dimensional spatial point cloud data. Through decomposition, thresholding, and reconstruction steps, high-frequency noise signals are removed to obtain calibration data.

[0027] Based on the geodesic theory in differential geometry, the calibration data is fitted with an air-gouging trajectory to obtain the trajectory curve;

[0028] By analyzing the curvature of the trajectory curve, the straight line segment and the arc segment of the air gouging are divided to generate a complete air gouging spatial trajectory. The coordinates of each feature point on the air gouging spatial trajectory are extracted simultaneously to obtain the three-dimensional coordinate values ​​of the X, Y, and Z of the air gouging trajectory points.

[0029] Based on the spatial angle form of the air planer, an attitude optimization model for the air planer gun is established. The calculated solution of the air planer gun attitude is obtained based on the Jacobian matrix of robot kinematics. The attitude parameters of the air planer gun are optimized by pseudo-inverse iteration of the Jacobian matrix to generate spatial attitude parameters of the air planer gun that are adapted to the requirements of the air planer process.

[0030] Preferably, the light scattering compensation model is used to correct the reflection coefficient of the laser on the workpiece surface in order to correct the point cloud coordinate offset, and the reflection coefficient is calibrated in real time by laser reflection intensity detection;

[0031] In the air gouging trajectory fitting, the trajectory of the straight line segment is solved by two geodesics, and the trajectory of the circular arc segment is constructed by three geodesics to form the circular arc equation.

[0032] The Jacobian matrix maps the relationship between the linear velocity, angular velocity, and joint angular velocity at the tip of the welding torch.

[0033] Preferably, after the robot has finished planing the mold cavity with the air planer gun, the robot end automatically switches the air planer gun to a welding gun via a gun-changing disc;

[0034] The 3D vision camera takes pictures of the mold cavity after air gouging, obtains the relative position value of the mold's outer shape reference distance from the robot coordinates, and obtains the spatial position data of the mold cavity reference point in the X, Y, and Z directions as reference coordinate values. The robot is then controlled to carry the welding gun to weld and fill the mold cavity.

[0035] Preferably, a control method for a robotic intelligent air gouging welding repair mold workstation includes the following steps:

[0036] S1. Set the mold on the ground or platform within the air gouging and welding area of ​​the workstation;

[0037] S2. Use a 3D vision camera to take pictures of the mold outline, obtain the spatial point cloud of the mold, process it, obtain the relative position value of the mold's outer reference distance from the robot coordinates, and obtain the spatial position data of the mold reference point in the X, Y, and Z directions as the reference coordinate values.

[0038] S3. Take pictures of the mold cavity with a 3D vision camera to obtain the spatial point cloud of the air planer path, and process it to obtain the spatial position data of the air planer path in the X, Y and Z directions, as well as the spatial attitude of the corresponding air planer gun.

[0039] S4. Based on the reference coordinate values, air planer path coordinate values, and the corresponding air planer gun's spatial attitude, adjust the air planer path parameters and the air planer gun's attitude.

[0040] S5. The robot moves and air-gougs the surface of the mold cavity.

[0041] S6. After air gouging is completed, the surface of the mold cavity after air gouging is photographed by a 3D vision camera to obtain the spatial point cloud of the welding path. The data is then processed to obtain the spatial position data of the welding path, the coordinate values ​​of the welding path in the X, Y, and Z directions, and the spatial posture of the corresponding welding torch.

[0042] S7. Based on the reference coordinate values, welding path coordinate values, and the corresponding spatial posture of the welding torch, adjust the parameters of the welding path and the posture of the welding torch.

[0043] S8. The robot moves and uses a welding torch to weld and fill the surface of the mold cavity after air gouging.

[0044] This invention provides a robotic intelligent air gouging welding repair mold workstation and its control method. It has the following beneficial effects:

[0045] 1. This invention utilizes the six-axis motion of a robot to drive the movement of a 3D vision camera and an air planer gun. Combined with the autonomous photography and point cloud processing of the mold contour by the 3D vision camera, the coarse positioning of the workpiece is completed. Then, through secondary photography and point cloud analysis of the mold cavity by the 3D vision camera, the three-dimensional coordinates of the mold cavity and the spatial posture of the air planer gun are automatically identified, achieving precise positioning of the air planing path and intelligent matching of the air planer gun posture. This shortens the air planing cycle of a single workpiece and solves the problem of low air planing efficiency for current molds.

[0046] 2. This invention uses a robot's six-axis motion to move a 3D vision camera and a welding torch. By combining the 3D vision camera's imagery and point cloud analysis of the gouging surface of the mold cavity, the invention automatically identifies the three-dimensional coordinates of the gouging surface of the mold cavity and the corresponding spatial posture of the welding torch. This achieves precise positioning of the welding path and intelligent matching of the welding torch posture, while also solving the problem of low welding efficiency in current molds. Attached Figure Description

[0047] Figure 1 This is a three-dimensional structural diagram of a robotic intelligent air gouging welding and repair mold workstation proposed in this invention;

[0048] Figure 2 This is a schematic diagram of a robotic intelligent air gouging welding and repair mold workstation proposed in this invention after the fence has been removed.

[0049] Figure 3 This is a schematic diagram of the main structure of a robotic intelligent air gouging welding repair mold workstation proposed in this invention;

[0050] Figure 4 This is a top view of the robotic intelligent air gouging welding and repair mold workstation proposed in this invention.

[0051] Figure 5 This is a side view of a robotic intelligent air gouging welding and repair mold workstation proposed in this invention.

[0052] The components are as follows: 1. Fence; 2. Robot; 3. 3D vision camera; 4. Welding torch; 5. Mold; 6. Platform; 7. Control cabinet; 8. Air gouging torch; 9. Female module B; 10. Fixed bracket; 11. Barreled welding wire drum; 12. Air gouging power supply; 13. Welding machine; 14. Robot control cabinet; 15. Male module; 16. Female module A; 17. Torch cleaning station. Detailed Implementation

[0053] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Please see the appendix Figure 1 -Appendix Figure 5 This invention provides a robotic intelligent air gouging welding repair mold workstation, including: a robot 2, a welding machine 13, a 3D vision camera 3, a welding torch 4, an air gouging power supply 12, an air gouging torch 8, and a torch changing disc;

[0055] The gun changing plate consists of one male module 15 and two female modules, namely female module A16 and female module B9.

[0056] The male module 15 is installed at the end of the sixth axis of robot 2;

[0057] The welding torch 4 is fixed to the female module A16 of the torch changing plate. It can be detached and installed on one side of the robot 2 through the connection and cooperation of the male module 15 and the female module A16.

[0058] The air planer gun 8 is fixed to the female module B9 of the gun changing disc. It can be detached and installed on one side of the robot 2 through the connection and cooperation of the male module 15 and the female module B9.

[0059] The 3D vision camera 3 is set on the other side of the robot 2 to take the first picture of the outline of the mold 5, obtain the spatial point cloud of the mold 5, process it, obtain the relative position value of the mold 5's outer reference distance from the robot 2 coordinates, and obtain the spatial position data of the mold 5's reference point in the X, Y, and Z directions as the reference coordinate values.

[0060] The 3D vision camera 3 is also used to take a second picture of the outline of the mold 5, obtain the spatial point cloud of the cavity of the mold 5, and process it to obtain the spatial position data of the cavity of the mold 5 in the X, Y and Z directions, and calculate the spatial attitude of the corresponding air gouging gun 8 and welding gun 4; the spatial attitude of the air gouging gun 8 and welding gun 4 includes the coordinate values ​​of six degrees of freedom: X, Y, Z, Ra, Ry and Rz.

[0061] Robot 2 is a six-axis robot.

[0062] Specifically, the robot 2, along with its 3D vision camera 3 and air planer 8, moves in multiple directions to adapt to the curved surface structure of the mold 5's cavity, facilitating air planing operations on the mold 5's cavity. The 3D vision camera 3 takes pictures of the mold 5's outline to obtain its spatial point cloud, which is then processed to obtain the relative position values ​​of the mold 5's outer reference distance from the robot 2's coordinates. This yields the spatial position data of the mold 5's reference point in the X, Y, and Z directions, enabling autonomous coarse positioning of the mold 5 without manual intervention. This achieves rapid reference calibration after any workpiece placement, avoiding the tedious operations and error accumulation of traditional manual positioning.

[0063] The 3D vision camera 3 takes a second picture of the cavity surface of the mold 5 to obtain the spatial point cloud of the cavity surface. After processing, the spatial position data of the cavity surface, the coordinate values ​​of the cavity surface in the X, Y and Z directions, and the spatial posture of the corresponding air planer gun 8 are obtained. This allows the actual position and shape of the air planer trajectory to be identified, and the optimal working posture of the air planer gun 8 to be automatically adapted. This achieves the autonomous and precise positioning of the air planer and the intelligent matching of the posture of the air planer gun 8, ensuring the accuracy and consistency of the air planer operation.

[0064] Furthermore, the two fixed supports 10 form a dual workstation, allowing the robot 2 to perform air planing on the mold 5 on one fixed support 10, while the operator on the other fixed support 10 can load and unload the mold 5 or perform manual processing, reducing the waiting time between the operator and the machine and improving work efficiency.

[0065] Through the organic collaboration of various components, a complete operation process is constructed, from rough positioning of the workpiece, fine positioning of the cavity surface of mold 5 to intelligent air planing. No manual teaching programming and cumbersome tooling adjustments are required, realizing the intelligent and automated operation of mold 5 and solving the problem of low air planing efficiency of mold 5.

[0066] Furthermore, the relative position values ​​of the outer datum of mold 5 and robot 2 are obtained, and the spatial position data of the datum point of mold 5 in the X, Y, and Z directions are obtained. This specifically includes the following steps:

[0067] The surface of the mold 5 is irradiated by a surface vibrating laser beam emitted by the 3D vision camera 3. The robot 2 takes continuous pictures from a preset initial point along six axes, collecting spatial point cloud data of the target area of ​​the mold 5's outline.

[0068] Outlier removal is performed on spatial point cloud data based on the improved RANSAC algorithm. Outliers are removed through least squares iterative optimization, and then Euclidean distance clustering analysis is used for block processing to obtain several continuous local point cloud blocks.

[0069] Based on the Bézier curve fitting theory in differential geometry, contour features are extracted for each local point cloud block, and the parametric equation of the contour curve is constructed by solving the coordinates of the control points of the Bézier curve.

[0070] By determining the continuity of curve segments, the parametric equations of each local point cloud block are spliced ​​into a continuous contour curve, and the effective right-angled sides of the contour that meet the preset benchmark distance threshold are locked.

[0071] Based on the parametric equation of the right-angled side of the effective contour, the linear equation of the reference side is derived. Combined with the origin coordinate of the robot 2 coordinate system, the relative position value of the outer reference distance of the mold 5 from the robot 2 coordinate system is calculated, and the spatial position data of the reference point of the mold 5 in the X, Y, and Z directions is output.

[0072] The improved RANSAC algorithm is an algorithm that introduces an adaptive inlier threshold adjustment mechanism, which dynamically corrects the inlier threshold according to the point cloud density.

[0073] The block processing involves setting the cluster radius based on the dimensional characteristics of the mold cavity 5, dividing the collected point cloud of the large workpiece into several continuous local point cloud blocks.

[0074] The coordinates of the control points of the Bézier curve are obtained by the least squares iterative method.

[0075] Specifically, the system controls robot 2 to move 3D vision camera 3 to a preset observation height of approximately 300mm to 500mm above mold 5. 3D vision camera 3 activates its surface-vibrating laser generator, projecting high-brightness laser stripes onto the surface of mold 5. Simultaneously, robot 2 moves at a constant speed along a pre-planned "bow" or "swivel" shaped trajectory. The system triggers the camera shutter at fixed time intervals, continuously acquiring and stitching data to form an original 3D point cloud dataset covering the outline of mold 5. , where each point Include Coordinate information.

[0076] For raw point cloud data Outlier noise caused by welding sparks in the workshop or reflections from oil stains on the metal surface of molds is addressed by the system executing an improved RANSAC algorithm. First, a spatial index is established for any point in the dataset. Search its radius The nearest neighbor set within the neighborhood, counting the number of nearest neighbors. As local density The system sets a basic distance threshold. And calculate the dynamic threshold based on density. The calculation formula is as follows:

[0077] ;

[0078] In the formula: Indicates the first The threshold for determining each data point;

[0079] This indicates the preset basic error tolerance (e.g., 0.5mm).

[0080] Indicates the density weighting coefficient;

[0081] Indicates the density attenuation coefficient;

[0082] Indicates the first Local point cloud density of a point;

[0083] Given the base of the natural logarithm, the algorithm enters an iterative loop: randomly select 3 points to construct a local planar model, calculate the distance from the remaining points to the plane, and if the distance is less than the dynamic threshold of the corresponding point... If the point is an interior point, then mark it as an interior point after the maximum number of iterations. Then, the point set corresponding to the model with the most interior points is retained as the valid surface data. Remove the remaining noise points.

[0084] Subsequently, the system processed the cleaned point cloud. Perform Euclidean distance clustering and set clustering tolerance. and minimum number of cluster points The system selects an unprocessed point from the point cloud as a seed and uses breadth-first search to find all points with a Euclidean distance less than 1. The neighboring points are added to the current cluster until no further expansion is possible, thus segmenting the contour point cloud of the large mold 5 into... A separate local point cloud patch For each point cloud block The system uses the least squares method to fit the Bézier curve, parameterizing the points in the point cloud block according to spatial order. Construct the observation matrix and basis function matrix The control point matrix is ​​determined by solving a system of linear equations. The solution formula is:

[0085] ;

[0086] In the formula: for The control point coordinate matrix; for Bernstein basis function matrix, The number of points in the point cloud block. The order of the curve; for The transpose of the matrix; for The inverse matrix; for The coordinate matrix of the observation point cloud.

[0087] After obtaining the parametric equations of each curve segment, the system performs continuity and orthogonality checks and calculates the Euclidean distance between the endpoints of adjacent curve segments. Angle with tangent vector ,like Less than the distance threshold and If the angle is less than the threshold, splicing is performed. The system identifies straight line feature segments on the spliced ​​complete contour and extracts two straight line segment vectors with an angle of approximately 90 degrees between them. The corner coordinates of mold 5 in the camera coordinate system are obtained by calculating their intersection points. Using the hand-eye calibration matrix of robot 2 Transform the coordinates to the robot 2 base coordinate system using the following calculations:

[0088] ;

[0089] In the formula: This represents the homogeneous coordinates of the mold 5 reference point in the robot base coordinate system; This represents the coordinate system from the camera coordinate system to the robot base coordinate system. A homogeneous transformation matrix, which includes rotation matrices and translation vectors; This represents the homogeneous coordinates of the mold 5 reference point in the camera coordinate system, calculated as follows. The first three components are the relative position values ​​of the outer datum of mold 5 to the coordinates of robot 2. .

[0090] This effectively overcomes the interference of complex lighting conditions in industrial settings on visual measurement, and enables millimeter-level automated calibration of the five reference coordinates of the mold.

[0091] Furthermore, the spatial position data of the cavity of mold 5, including the coordinate values ​​of the mold cavity in the X, Y, and Z directions, are obtained, and the spatial orientation of the corresponding air gouging gun 8 and welding gun 4 is calculated. This specifically includes the following steps:

[0092] Based on the light scattering compensation model, three-dimensional spatial point cloud data of the air-gouging area is collected by taking pictures with a 3D vision camera.

[0093] Wavelet transform is used to denoise the three-dimensional spatial point cloud data. Through decomposition, thresholding, and reconstruction steps, high-frequency noise signals are removed to obtain calibration data.

[0094] Based on the geodesic theory in differential geometry, the calibration data is fitted with an air-gouging trajectory to obtain the trajectory curve;

[0095] By analyzing the curvature of the trajectory curve, the straight line segment and the arc segment of the air gouging are divided to generate a complete air gouging spatial trajectory. The coordinates of each feature point on the air gouging spatial trajectory are extracted simultaneously to obtain the three-dimensional coordinate values ​​of the X, Y, and Z of the air gouging trajectory points.

[0096] Based on the spatial angle form of the air planer, an attitude optimization model for the air planer gun 8 is established. The calculated solution of the attitude of the air planer gun 8 is obtained based on the Jacobian matrix of the robot 2 kinematics. The attitude parameters of the air planer gun 8 are optimized by pseudo-inverse iteration of the Jacobian matrix to generate spatial attitude parameters of the air planer gun 8 that are adapted to the air planing process requirements.

[0097] The light scattering compensation model is used to correct the reflection coefficient of the laser on the workpiece surface in order to correct the point cloud coordinate offset. The reflection coefficient is calibrated in real time by detecting the laser reflection intensity.

[0098] In air gouging trajectory fitting, the trajectory of a straight line segment is solved by two geodesics, and the trajectory of a circular arc segment is constructed by three geodesics to form the circular arc equation.

[0099] The Jacobian matrix maps the relationship between the linear velocity, angular velocity, and joint angular velocity at the end of the welding torch 4.

[0100] Specifically, the system controls robot 2 to move above the cavity of mold 5, which has been roughly positioned. The 3D vision camera 3 then activates a high-resolution scanning mode to take detailed pictures of the air-gouged area. Considering the reflection and scattering phenomena on the metal cavity surface, the system applies a light scattering compensation model to preprocess the acquired raw 3D point cloud data. The system also reads the laser reflection intensity value fed back by the camera sensor in real time. Using a pre-calibrated light scattering offset coefficient For the depth coordinates of each sampling point The correction is made, and the calculation formula is as follows:

[0101] ;

[0102] In the formula: This indicates the corrected depth coordinates; Represents the original depth coordinates; This represents the scattering coefficient related to the mold material; This represents the normalized value of the laser reflection intensity; This represents the laser incident angle. This model is used to correct the Gaussian distribution center shift caused by multiple scattering of photons inside the material, thus obtaining more accurate calibration point cloud data.

[0103] To address the slight high-frequency noise still present in the corrected point cloud data, the system employs wavelet transform for denoising, selecting the Daubechies wavelet basis to perform denoising on the three-dimensional coordinate signal. Layered discrete wavelet decomposition yields approximation coefficients and detail coefficients; for high-frequency detail coefficients... A soft thresholding function is used for processing, and the threshold formula is:

[0104] ;

[0105] In the formula: Represents the processed wavelet coefficients; Represents the original detail coefficients; Represents a symbolic function; The general threshold is represented by the formula: ,in Indicates the standard deviation of noise. The signal length is represented by a reconstruction algorithm, which is then used to restore the processed coefficients, resulting in smooth calibration data that retains geometric features.

[0106] Based on this, the system applies differential geometry geodesic theory to fit the smoothed point cloud data. The system first calculates the discrete curvature of each discrete point on the path. Set curvature threshold When the curvature of continuous points Less than When it is determined to be a straight line segment, a two-point geodesic fitting is used; when Greater than or equal to When a segment is identified as an arc, three adjacent points are selected to construct a spatial arc equation. The system connects the fitted curve segments to generate a complete air-scraping spatial trajectory, which is then discretized with a fixed step size, outputting the three-dimensional coordinates of each feature point on the trajectory. .

[0107] To generate the spatial attitude of the air gouging gun 8 that is adapted to the air gouging process, the system establishes an attitude optimization model and defines the target end pose vector according to the air gouging process requirements. Using the kinematic Jacobian matrix of robot 2 Establish joint angular velocity With end Cartesian velocity Mapping relationship between In order to solve the optimal joint angle Iterative optimization is performed using the pseudo-inverse of the Jacobian matrix, with the iterative formula as follows:

[0108] ;

[0109] In the formula: Indicates the amount of joint angle adjustment; Representing the Jacobian matrix The pseudo-inverse matrix is ​​defined as ; Indicates the desired end-effector pose; This represents the currently calculated end-effector pose; Represents the identity matrix; This represents the null space projection matrix, used to handle redundant degrees of freedom; The gradient vector represents the objective function, used for obstacle avoidance or moving away from joint limits. The system iterates multiple times until the position error is reached. Less than the preset precision, the final output contains The six-degree-of-freedom attitude parameter commands are given to the robot controller.

[0110] This not only eliminates the impact of metal reflection on visual accuracy, but also achieves automatic generation of air gouging paths and smooth transition of robot posture on complex curved surfaces through mathematical optimization algorithms, ensuring the consistency of air gouging depth and high quality of welding repair.

[0111] After the robot 2, carrying the air planer gun 8, finishes planing the cavity of the mold 5, the end of the robot 2 automatically switches the air planer gun 8 to the welding gun 4 through the gun changing disc;

[0112] The 3D vision camera 3 takes a picture of the cavity of the gouged mold 5 to obtain the relative position value of the outer reference distance of the mold 5 to the coordinates of the robot 2. The spatial position data of the reference point of the cavity of the mold 5 in the X, Y and Z directions are obtained. The robot 2 is controlled to carry the welding gun 4 to weld and fill the cavity of the mold 5.

[0113] Specifically, after the robot 2 controls the air planer gun 8 to complete the material removal operation at the cavity defect of the mold 5 along the planned trajectory, the control system issues a gun change command, controlling the robot 2 to move the air planer gun 8 to the storage position of the gun changer system.

[0114] The male module 15 at the end of robot 2 performs a pneumatic or mechanical unlocking action to release the female module B9 connected to the air gouging gun 8 and place it on the support. Then, robot 2 moves to the station where the welding gun 4 is stored, aligns and presses the female module A16 connected to the welding gun 4 through the male module 15, performs a locking action and connects the air and electrical signals, completing the automatic physical switch from air gouging tool to welding tool.

[0115] Subsequently, to ensure the accuracy of the welding fill, robot 2 drives 3D vision camera 3 to perform a third image scan of the cavity area of ​​mold 5, which has just completed the air gouging operation. The camera collects the three-dimensional topographic point cloud data of the U-shaped or V-shaped grooves formed after air gouging. The system uses the aforementioned improved RANSAC algorithm and light scattering compensation model to process the data and eliminate the interference of metallic luster caused by cutting.

[0116] The system performs cross-sectional analysis on the processed slot point cloud, identifies the bottom center line and the edges of the two side walls of the air-gouged slot, and calculates the depth and position of the new surface generated after the air-gouging removes the material. This updates the spatial position data of the reference point of the mold cavity 5 in the X, Y, and Z directions, which are the precise zero points for welding arc initiation.

[0117] Based on the width and depth data of the gouging grooves, the system automatically plans a multi-layer, multi-pass welding filling path and oscillation parameters, and calculates the posture of the welding torch 4 at each point along the path. Finally, based on the updated reference coordinate values ​​and path parameters, the control system controls the robot 2 to move the welding torch 4 to the arc ignition point, ignite the arc, and precisely weld and fill the gouging grooves of the mold 5 cavity layer by layer along the planned trajectory until the mold surface contour is restored.

[0118] This embodiment also provides a control method for a robotic intelligent air gouging welding repair mold workstation, including the following steps:

[0119] S1. Place the mold 5 on the ground or platform 6 within the air gouging and welding area of ​​the workstation;

[0120] S2. Use 3D vision camera 3 to take pictures of the outline of mold 5, obtain the spatial point cloud of mold 5, process it, obtain the relative position value of the outer reference distance of mold 5 from robot 2 coordinates, and obtain the spatial position data of mold 5 reference point in the X, Y and Z directions reference coordinate values.

[0121] S3. Take pictures of the cavity of mold 5 using 3D vision camera 3 to obtain the spatial point cloud of the air planer path, and process it to obtain the spatial position data of the air planer path in the X, Y and Z directions, the coordinate values ​​of the air planer path, and the corresponding spatial attitude of the air planer gun 8.

[0122] S4. Based on the reference coordinate values, the air planer path coordinate values, and the spatial attitude of the corresponding air planer gun 8, adjust the parameters of the air planer path and the attitude of the air planer gun 8.

[0123] S5. The robot 2 moves and air-gougs the surface of the cavity of mold 5 using air-gouging gun 8.

[0124] S6. After the air gouging is completed, the surface of the cavity of mold 5 after air gouging is photographed by 3D vision camera 3 to obtain the spatial point cloud of the welding path, and process it to obtain the spatial position data of the welding path in the X, Y and Z directions, the welding path coordinate values, and the corresponding spatial posture of the welding gun 4.

[0125] S7. Based on the reference coordinate values, welding path coordinate values, and the spatial posture of the corresponding welding torch 4, adjust the parameters of the welding path and the posture of the welding torch 4.

[0126] S8. The robot 2 moves and the welding gun 4 welds and fills the surface of the mold cavity after air gouging.

[0127] Specifically, firstly, the operator uses an overhead crane or forklift to lift the mold 5 to be repaired onto the platform 6 or fixed bracket 10 in the workstation. Since this system has a visual calibration function, there is no need to perform precise mechanical positioning of the mold 5 or use special tooling fixtures. The mold 5 can be placed at any angle, as long as it is within the reachable working range of the robot 2.

[0128] Subsequently, the control system instructs robot 2 to drive 3D vision camera 3 to perform the first large-scale image scan of the outer contour of mold 5. The system executes the aforementioned reference coarse positioning algorithm steps, collects the original point cloud including the edge of the mold, uses the improved RANSAC algorithm to remove environmental background noise, extracts the edge features of the mold through the Bezier curve fitting algorithm and identifies the mutually perpendicular right-angled sides, calculates the displacement vector of the right-angled vertex relative to the base coordinate system of robot 2, thereby obtaining the spatial position data of the reference point of mold 5 in the X, Y, and Z directions, and thus establishing the transformation relationship between the workpiece coordinate system and the robot coordinate system.

[0129] Based on the coarse positioning data, robot 2 moves the 3D vision camera 3 above the defect area of ​​the mold cavity 5 for a second, finer photograph. The system executes the aforementioned air-gouging path fine positioning algorithm steps, applies a light scattering compensation model to correct the depth data of the metal surface, and obtains a high-fidelity point cloud through wavelet transform denoising. Using geodesic theory, it fits the air-gouging trajectory curve on the defect surface, extracts the X, Y, and Z coordinates of each discrete point on the trajectory, and simultaneously calculates the optimal spatial posture of the air-gouging gun 8 at each point on the trajectory to meet the process back tilt angle requirements through a Jacobian matrix pseudo-inverse iterative optimization algorithm. The control system fuses the obtained global reference coordinate values ​​with the local air-gouging path coordinate values ​​to generate air-gouging motion command codes that robot 2 can execute, and compensates and corrects the air-gouging path according to the TCP parameters of the air-gouging gun 8 to ensure that the end of the air-gouging carbon rod is accurately aligned with the defect position.

[0130] Subsequently, the end effector of robot 2 locks the female module B9 of the air gouging gun 8 via male module 15, and the control system starts the air gouging power supply 12. Robot 2 guides the air gouging gun 8 along the surface of the mold cavity 5 according to the planned path and posture, igniting the electric arc and using compressed air to blow away the molten metal, precisely removing defects and fatigue layers in the cavity. After air gouging is completed, robot 2 automatically performs a gun-changing action, switching the air gouging gun 8 to the welding gun 4, and driving the 3D vision camera 3 to take a third picture of the U-shaped or V-shaped groove formed after air gouging.

[0131] The system reprocesses the collected slot point cloud, analyzes the slot width, depth, and cross-sectional shape, recalculates the path coordinates required for welding filling, and uses the Jacobian matrix algorithm to calculate the optimal spatial posture of the welding torch 4 within the slot to avoid collisions and facilitate molten pool formation. Based on the baseline coordinate values ​​and the latest welding path coordinate values, combined with the slot volume data, the system automatically plans a multi-layer, multi-pass welding strategy. Simultaneously, it corrects the welding trajectory according to the TCP parameters of the welding torch 4, starts the welding machine 13 and wire feeding mechanism, and controls the robot 2 to move the welding torch 4 to the arc initiation point. According to the adjusted path and posture parameters, it performs layer-by-layer welding filling on the slot after gouging the cavity of the mold 5 until the repair area is welded to the predetermined height, completing the repair operation.

[0132] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robotic intelligent hot wire welding repair mold station, characterized by, The application relates to a welding robot system, which comprises a robot (2), a welding machine (13), a 3D vision camera (3), a welding torch (4), a gouging power supply (12), a gouging torch (8) and a gun changing disc. The gun changing disc is composed of a male module (15) and two female modules, namely a female module A (16) and a female module B (9); The male module (15) is installed at the end of the sixth shaft of the robot (2); The welding torch (4) is fixed to the female module A (16) of the gun changing disc and can be detachably arranged on one side of the robot (2) through the connection and cooperation of the male module (15) and the female module A (16); The gouging torch (8) is fixed to the female module B (9) of the gun changing disc and can be detachably arranged on one side of the robot (2) through the connection and cooperation of the male module (15) and the female module B (9); The 3D vision camera (3) is arranged on the other side of the robot (2) and is used for taking a first photo of the profile of a mold (5), acquiring the spatial point cloud of the mold (5) and processing the spatial point cloud to obtain the relative position value of the shape reference distance of the mold (5) from the coordinates of the robot (2) and the spatial position data X, Y and Z of the reference point of the mold (5); The 3D vision camera (3) is also used for taking a second photo of the profile of the mold (5), acquiring the spatial point cloud of the mold cavity of the mold (5) and processing the spatial point cloud to obtain the mold cavity coordinate value of the spatial position data X, Y and Z of the mold cavity of the mold (5) and the spatial attitude corresponding to the gouging torch (8) and the welding torch (4); The mold cavity coordinate value of the spatial position data X, Y and Z of the mold cavity of the mold (5) and the spatial attitude corresponding to the gouging torch (8) and the welding torch (4) are obtained through the following steps: Based on a light scattering compensation model, the 3D vision camera (3) is used for taking a photo and collecting three-dimensional spatial point cloud data of a gouging area; Wavelet transform is used for denoising the three-dimensional spatial point cloud data, high-frequency noise signals are removed through decomposition, threshold processing and reconstruction steps, and calibration data are obtained; Based on the geodesic line theory in differential geometry, the calibration data are subjected to gouging trajectory fitting to obtain a trajectory curve; Through curvature analysis of the trajectory curve, a complete gouging space trajectory is generated by dividing the straight line segment and the circular arc segment of the gouging, and the coordinates of each feature point on the gouging space trajectory are synchronously extracted to obtain the three-dimensional coordinate values of the gouging trajectory points X, Y and Z; Combined with the spatial angle form of the gouging, an attitude optimization model of the gouging torch (8) is established, the calculation solution of the attitude of the gouging torch (8) is solved based on the Jacobian matrix of the kinematics of the robot (2), the attitude parameters of the gouging torch (8) are optimized through the pseudo-inverse iteration of the Jacobian matrix, and the spatial attitude parameters of the gouging torch (8) that adapt to the requirements of the gouging process are generated; The light scattering compensation model is used for correcting the reflection coefficient of laser on the workpiece surface to correct the point cloud coordinate offset, and the reflection coefficient is calibrated in real time through laser reflection intensity detection; In the gouging trajectory fitting, the straight line segment trajectory is solved through two-point geodesic lines, and the circular arc segment trajectory is constructed through three-point geodesic lines to form a circular arc equation. ​ The Jacobian matrix maps the linear velocity and angular velocity of the end of the welding gun (4) to the joint angular velocity.

2. A robotic intelligent gas tung arc welding repair fixture work station according to claim 1, wherein, The spatial pose of the air gouging gun (8) and the welding gun (4) includes coordinate values of six degrees of freedom of X, Y, Z, Ra, Ry, and Rz. The robot (2) is a six-axis robot.

3. A robotic intelligent gas tung arc welding repair fixture work station according to claim 1 wherein, The relative position value of the mold (5) from the robot (2) coordinate is obtained, and the spatial position data X, Y, and Z of the reference point of the mold (5) are obtained. The surface of the mold (5) is irradiated with a face vibration laser emitted by the 3D vision camera (3), and the six-axis robot (2) is continuously photographed from the preset initial point to collect the spatial point cloud data of the target region of the mold (5) profile contour. Based on the improved RANSAC algorithm, outliers are removed from the spatial point cloud data, and abnormal points are removed through least square iteration optimization, and then Euclidean distance clustering analysis is performed for block processing to obtain a plurality of continuous local point cloud blocks. Based on the Bezier curve fitting theory in differential geometry, the profile feature of each local point cloud block is extracted, and the parameter equation of the profile curve is constructed by solving the control point coordinates of the Bezier curve. Through the continuity judgment of the curve segment, the parameter equations of the local point cloud blocks are spliced into a continuous profile curve, and the effective profile right angle edge satisfying the preset reference distance threshold is locked. Based on the parameter equation of the effective profile right angle edge, the straight line equation of the reference edge is derived, the origin coordinates of the robot (2) coordinate are combined, the relative position value of the mold (5) from the robot (2) coordinate is calculated, and the reference coordinate values of the spatial position data X, Y, and Z of the mold (5) reference point are output.

4. A robotic intelligent gas tung arc welding repair fixture work station according to claim 3, wherein, The improved RANSAC algorithm is an algorithm that introduces an adaptive inlier threshold adjustment mechanism, which dynamically corrects the inlier threshold according to the point cloud density. The block processing is to set the clustering radius according to the size characteristics of the mold (5) cavity, and the collected point cloud of the large-size workpiece is divided into a plurality of continuous local point cloud blocks. The control point coordinates of the Bezier curve are solved by the least square iteration method.

5. A robotic intelligent gas tung arc welding repair fixture work station according to claim 1 wherein, After the robot (2) with the air gouging gun (8) finishes gouging the mold (5) cavity, the robot (2) end automatically switches the air gouging gun (8) to the welding gun (4) through the gun changing disc. The 3D vision camera (3) photographs the gouged mold (5) cavity to obtain the relative position value of the mold (5) from the robot (2) coordinate, and obtains the reference coordinate values of the spatial position data X, Y, and Z of the mold (5) cavity reference point, and controls the robot (2) with the welding gun (4) to weld the mold (5) cavity.

6. A control method for a robotic intelligent gas gouging weld repair mold workstation, the method comprising: A robot intelligent air gouging and welding repair mold workstation according to any one of claims 1-5, comprising the following steps: S1, place the mold (5) on the ground or platform (6) in the air gouging and welding area of the workstation; S2, the profile of the mold (5) is photographed by using the 3D vision camera (3), the space point cloud of the mold (5) is obtained, and the relative position value of the shape reference distance robot (2) coordinate is obtained by processing, the space position data X, Y, Z three direction reference coordinate value of the mold (5) reference point is obtained; S3, the mold cavity of the mold (5) is photographed by the 3D vision camera (3), the space point cloud of the gas blowing path is obtained, and the space position data X, Y, Z three direction gas blowing path coordinate value of the gas blowing path is obtained by processing, and the space attitude of the corresponding gas blowing gun (8) is obtained; S4, according to the reference coordinate value, the gas blowing path coordinate value, and the space attitude of the corresponding gas blowing gun (8), the parameters of the gas blowing path and the attitude of the gas blowing gun (8) are adjusted; S5, through the movement of the robot (2), and through the gas blowing gun (8), the surface of the mold cavity of the mold (5) is gas blown; S6, after the gas blowing is completed, the surface of the mold cavity of the mold (5) after the gas blowing is photographed by the 3D vision camera (3), the space point cloud of the welding path is obtained, and the space position data X, Y, Z three direction welding path coordinate value of the welding path is obtained by processing, and the space attitude of the corresponding welding gun (4) is obtained; S7, according to the reference coordinate value, the welding path coordinate value, and the space attitude of the corresponding welding gun (4), the parameters of the welding path and the attitude of the welding gun (4) are adjusted; S8, through the movement of the robot (2), and through the welding gun (4), the surface of the mold cavity of the mold (5) after the gas blowing is welded and filled.

Images