A welding robot welding control method and system for use on construction sites

Abstract

This invention discloses a welding control method and system for welding robots applied in construction sites. The invention maps the robot's link pose information and component surface point cloud data to a base coordinate system, constructs a geometric topological feature model, and generates a dynamic envelope model covering the robotic arm links and joints to achieve real-time interference detection. When potential interference is detected, flexible obstacle avoidance is achieved during continuous welding. Simultaneously, the posture deviation is calculated by combining the welding torch posture and the groove normal relationship, and the arc parameters and wire feed speed are adjusted accordingly to achieve adaptive compensation control of the welding process. This invention achieves high-precision spatial interference detection in complex environments, improving the robot's perception ability in unstructured work spaces. By linking posture deviation with welding process parameters, dynamic compensation of arc characteristics and wire feed process is achieved, thereby ensuring the consistency and stability of weld formation quality under complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of welding robot technology, and specifically to a welding robot welding control method and system applied to construction sites. Background Technology

[0002] Welding robot technology is an automated operation field that integrates robotics, welding technology, automatic control theory, and sensing technology. Its core logic lies in using a multi-axis robotic arm to simulate manual welding actions, precisely scheduling servo drive mechanisms through a control system to achieve coordinated movement of each joint, and cooperating with the welding power source and end effector to complete the metal joining task. This technical field encompasses structural design, kinematic modeling, spatial path planning, and real-time control of welding process parameters, typically utilizing position sensors and arc status monitoring to achieve programmed execution of the welding process.

[0003] For steel structure welding in complex environments such as outdoor construction sites, existing automated control solutions primarily focus on adaptive management of on-site conditions. In implementing these solutions, existing control systems typically involve key aspects such as bevel recognition, trajectory correction, and multi-layer, multi-path planning. Specifically, current solutions often employ laser structured light scanning technology to acquire the three-dimensional geometric coordinates of the weld seam, and combine this with proportional-integral-derivative (PID) control strategies to dynamically adjust the end effector's pose. Regarding path determination, existing technologies mainly rely on offline programming systems or teach-and-replay methods, using preset or guided recording of welding paths to drive the actuator to complete the predetermined welding operation process on the construction site.

[0004] Existing welding robot control technologies for complex construction sites typically separate kinematic models from environmental perception, making it difficult to deeply integrate point clouds of unstructured components with robot link poses. This results in blind spots in interference detection and insufficient early warning accuracy. Furthermore, when attitude adjustments are required to trigger interference, existing methods struggle to achieve flexible obstacle avoidance while maintaining a constant welding torch end pose and welding continuity. More importantly, existing control strategies generally lack a coupling mechanism between the robotic arm's spatial reconstruction and underlying welding process parameters. They ignore microscopic normal deviations such as push-pull angles and working angles caused by obstacle avoidance adjustments, and cannot proactively compensate for spatial attitude disadvantages through adaptive and collaborative adjustment of arc stiffness and wire feed speed. This easily leads to uneven stress on the molten pool under complex working conditions, making it difficult to ensure the stability and consistency of weld formation throughout the entire process. Summary of the Invention

[0005] This invention provides a welding control method and system for welding robots applied in construction sites. It achieves precise interference early warning by constructing a dynamic envelope model through the fusion of multi-source geometric topology. Under the premise of maintaining a constant welding torch end pose, it uses the curvature gradient of the component surface to guide redundant joints to flexibly avoid obstacles. At the same time, it quantifies the attitude normal deviation to coordinately control the arc stiffness and wire feed speed. This solves the technical problems of easy blind spots in interference detection in unstructured environments in complex construction sites, easy interruption of welding continuity in obstacle avoidance, and uneven stress on the molten pool and unstable weld formation quality caused by process deviations due to spatial pose adjustment.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows: A welding control method for welding robots applied on construction sites includes: S1: Obtain the robot joint rotation angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topological feature matrix; S2: Construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing; S3: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, calculate the surface curvature gradient of the component based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing surface curvature gradient of the component, and generate a joint space compensation vector. S4: Based on the joint space compensation vector, extract the pointing vector of the welding gun end and calculate the angle between it and the normal vector of the groove center to obtain the welding gun attitude normal deviation value; S5: Substitute the welding torch posture normal deviation value and cosine value into the preset pulse frequency operator to output the control coefficient for adjusting the arc stiffness, so as to shorten the peak current duration, increase the pulse frequency and correct the wire feeding speed.

[0007] Further, step S1 includes: S11: Obtain robot joint angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors. Based on the joint angle variables and link length constant, construct a homogeneous transformation matrix. Perform homogeneous transformation operation to obtain the three-dimensional coordinates of the endpoints of the link center axis. Extract the boundary extrema of the three axes of the endpoint three-dimensional coordinates. Arrange the boundary extrema and the endpoint three-dimensional coordinates as column vectors to establish the link spatial pose matrix. S12: Call the point cloud coordinates and sampling point normal vectors of the component surface, perform spatial translation and rotation transformation using the preset basic coordinate system transformation matrix, obtain the projection coordinates and projection normal vectors in the basic coordinate system, perform subtraction operation between the projection coordinates and the dimension coordinates corresponding to the link spatial pose matrix, extract the absolute value of the coordinate difference and perform array arrangement to obtain the surface projection difference set. S13: For the set of surface projection differences, construct an initial feature matrix, extract the absolute values ​​within the set and place them on the main diagonal of the initial feature matrix, fill the projection coordinates and projection normal vectors into the off-diagonal positions of the initial feature matrix according to the associated dimensions, perform normalization operation on the initial feature matrix, scale the matrix elements to a preset reference range, and generate a geometric topological feature matrix for the work space.

[0008] Further, step S2 includes: S21: Based on the geometric topology feature matrix, extract the three-dimensional coordinates of the endpoints of the central axis of the link spatial pose matrix, call the link radius constant and perform radial expansion vector superposition on the central axis to construct a set of closed cylindrical surface points around the central axis of each link segment, merge and arrange the closed cylindrical surface point set with the link spatial pose matrix and perform coordinate mapping to generate a set of link envelope surface features; S22: Based on the feature set of the connecting rod envelope surface, obtain its projected coordinates in the geometric topological feature matrix, calculate the Euclidean distance between the coordinates of each point in the feature set of the connecting rod envelope surface and its projected coordinates, and perform vectorization and arrangement of the Euclidean distance values ​​according to the sampling time order to obtain the spatial coordinate point-to-point distance sequence. S23: Call the spatial coordinate point pair distance sequence, sort the values ​​in the spatial coordinate point pair distance sequence in ascending order, and extract the first minimum value as the minimum interference distance.

[0009] Furthermore, the specific operations for constructing the feature set of the link envelope surface in step S21 are as follows: At the endpoints of the central axis of two adjacent connecting rods, a hemispherical envelope surface point set is constructed with the connecting rod radius constant as the radius. Perform a Boolean union operation on the hemispherical envelope surface point set and the closed cylindrical surface point set to form a capsule-shaped envelope model covering the robot joint area; The spatial coordinates of all vertices in the capsule-shaped envelope model are mapped to the geometric topological feature matrix, thereby expanding the feature set of the link envelope surface.

[0010] Further, step S3 includes: S31: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, call the sampling point normal vector and the corresponding component surface point cloud coordinates, calculate the absolute value of the three-axis component difference of the normal vector of adjacent sampling points, obtain the spatial Euclidean distance between adjacent sampling points, perform a division operation between the three-axis component difference and the spatial Euclidean distance, and generate the component surface curvature gradient value. S32: For the surface curvature gradient value of the component, obtain the sampling angle sequence in the redundant joint angle search space, calculate the gradient change rate of the surface curvature gradient value at the corresponding position of the sampling angle, extract the extreme value terms with negative values ​​in the gradient change rate, arrange the deflection orientation corresponding to the extreme value terms, and obtain the joint deflection guide vector. S33: Based on the joint deflection guide vector, call the current redundant joint actual rotation angle and the preset deflection step coefficient, perform a multiplication operation on the joint deflection guide vector and the preset deflection step coefficient to obtain the rotation angle compensation offset, and perform a matrix addition operation on the rotation angle compensation offset and the actual rotation angle to generate the joint space compensation vector.

[0011] Furthermore, the redundant joint angle search space is defined by the limiting range from the second axis to the fourth axis of the robotic arm; The preset deflection step size coefficient is generated by performing a mapping operation between the minimum interference spacing and the surface curvature gradient value; The mapping operation includes: obtaining the deviation value of the minimum interference spacing relative to the preset warning boundary, performing a multiplication operation between the reciprocal term of the deviation value and the surface curvature gradient value, and generating a preset deflection step size coefficient that increases as the deviation value decreases.

[0012] Further, step S4 includes: S41: Based on the joint space compensation vector, call the preset forward kinematics transformation matrix of the robotic arm to perform a multiplication operation, extract the direction vector elements of the posture rotation matrix in the matrix operation result, perform normalization processing on the direction vector elements, and obtain the three-dimensional component of the welding gun pointing. S42: Call the welding torch pointing three-dimensional component and the bevel center normal vector, perform dot product operation on the corresponding coordinate axis components of the two, and extract the absolute value of the dot product operation result to obtain the normal dot product scalar; S43: For the normal dot product scalar, call the inverse cosine calculation formula to obtain the spatial angle value. After converting the spatial angle value into standard angle units, perform a subtraction operation with the preset reference vertical angle constant to generate the welding gun posture normal deviation value.

[0013] Furthermore, the preset reference vertical angle constant is composed of an ideal normal reference and a preset welding process deflection compensation angle. The welding process deflection compensation angle includes the welding torch working angle parameter set based on a specific bevel type, and the push-pull angle parameter set along the welding path travel direction.

[0014] Further, step S5 includes: S51: Call the welding torch attitude normal deviation value and substitute it into the preset pulse frequency operator to obtain the preset reference period parameter and bias constant. Multiply the welding torch attitude normal deviation value and the reference period parameter to obtain the period deviation term. Add the period deviation term to the bias constant to generate the current period control value. Shorten the peak current duration and increase the pulse frequency through the current period control value. S52: Perform cosine operation on the normal deviation value of the welding torch attitude to extract the cosine value, call the preset reference wire feeding speed, multiply the cosine value by the reference wire feeding speed to obtain the basic variable of wire feeding, and perform division operation on the current cycle control value and the basic variable of wire feeding to obtain the wire feeding speed compensation amount. S53: Call the preset proportional gain parameter, multiply the wire feeding speed compensation amount by the proportional gain parameter to obtain the gain compensation term, add the gain compensation term to the current cycle control value to obtain the comprehensive control value, compare the comprehensive control value with the preset extreme value constant for amplitude limiting, and output the control coefficient used to adjust the arc stiffness.

[0015] A welding robot control system for construction sites, used to execute welding control methods, includes: Data acquisition and topology mapping module: used to acquire robot joint rotation angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topology feature matrix; Envelope construction and interference calculation module: used to construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing; Curvature-guided posture reconstruction module: used to keep the welding torch end pose unchanged when the minimum interference spacing reaches the preset warning boundary, calculate the curvature gradient of the component surface based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing curvature gradient of the component surface, and generate joint space compensation vector. Normal deviation quantification module: Based on the joint space compensation vector, extract the pointing vector of the welding torch end and calculate the angle between it and the normal vector of the groove center to obtain the normal deviation value of the welding torch attitude; Arc and wire feeding coordinated control module: used to substitute the welding torch posture normal deviation value into a preset pulse frequency operator, shorten the peak current duration and increase the pulse frequency, use the cosine value of the welding torch posture normal deviation value to correct the wire feeding speed, and output a control coefficient for adjusting the arc stiffness.

[0016] The beneficial effects of this invention are as follows: This invention constructs a geometric topological feature matrix by uniformly projecting robot kinematic parameters and component surface point cloud features onto a basic coordinate system, and then generates a dynamic capsule envelope model covering links and joints. This effectively eliminates collision detection blind spots in complex construction site environments and significantly improves the comprehensiveness of equipment's perception of unstructured work spaces and the accuracy of interference warning calculations. This invention proposes a redundant joint adaptive obstacle avoidance strategy that keeps the welding torch end pose constant. When an interference warning is triggered, the redundant joint is automatically guided to flexibly deflect in the direction of gradient decrease by utilizing the curvature gradient of the component surface. This allows the robot to achieve smooth obstacle avoidance without interrupting the current welding trajectory, greatly enhancing the system's continuous operation capability in narrow or complex working conditions. This invention deeply integrates robot kinematics reconstruction calculation with actual welding process. When quantifying the spatial deviation between the actual pointing of the welding torch and the ideal normal of the bevel, it introduces process reference parameters such as push-pull angle and working angle, accurately extracts the micro normal error of the welding torch caused by the adjustment of the robot arm posture, and realizes the effective conversion of spatial geometric displacement into welding process error. This invention establishes a dynamic compensation mechanism for arc and wire feeding based on electromechanical coordination. The quantified attitude normal deviation is introduced into the pulse frequency operator, which shortens the peak current duration, increases the pulse frequency, and corrects the wire feeding speed. By dynamically adjusting the arc stiffness, it actively compensates for the welding torch posture disadvantage caused by obstacle avoidance, effectively overcoming the problem of uneven force on the molten pool under complex spatial postures, and ensuring the consistency of weld formation throughout the entire process. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram illustrating the process of constructing the geometric topological feature matrix according to the present invention; Figure 3 This is a schematic diagram of the process for obtaining the minimum interference distance according to the present invention; Figure 4 This is a schematic diagram illustrating the process of calculating the joint space compensation vector according to the present invention; Figure 5 This is a schematic diagram illustrating the process of obtaining the normal deviation value of the welding torch attitude according to the present invention; Figure 6 This is a schematic diagram illustrating the process of adjusting the output arc stiffness coefficient according to the present invention. Detailed Implementation

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below 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.

[0019] In the description of this invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0020] Figure 1 This is an exemplary flowchart illustrating a welding robot control method for construction sites, according to some embodiments of this specification. In some embodiments, the welding robot control method for construction sites can be executed by processing logic, which may include hardware (e.g., circuits, dedicated logic, programmable logic, microcode, etc.), software (instructions running on a processing device to execute hardware simulations), and any combination thereof. In some embodiments, Figure 1 One or more operations in the flowchart of the construction site-oriented welding robot control method shown can be implemented by processing equipment and / or terminal equipment. For example, the construction site-oriented welding robot control method can be stored in a storage device in the form of computer programs and / or instructions, and invoked and / or executed by processing equipment and / or terminal equipment.

[0021] like Figure 1 As shown, the welding robot control method for construction sites disclosed in this invention specifically includes the following steps: S1: Obtain the robot joint rotation angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topological feature matrix; In one specific embodiment of the present invention, the acquisition of the underlying raw data mainly relies on a high-precision industrial sensing hardware system. Specifically, the "component surface point cloud coordinates" and related topographic features are acquired in real time during operation by a laser line scanning sensor rigidly integrated into the end of the welding torch, which scans the component surface to reproduce the complex weld seam and surrounding geometry at the construction site with high fidelity. The "robot joint rotation angle variables" are read in real time by absolute encoders of servo motors installed on each axis of the robotic arm. The above hardware configuration ensures the synchronization and accuracy of the robot's body state and environmental perception data, providing reliable physical data support for subsequent anti-interference calculations and attitude reconstruction.

[0022] S2: Construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing; S3: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, calculate the surface curvature gradient of the component based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing surface curvature gradient of the component, and generate a joint space compensation vector. S4: Based on the joint space compensation vector, extract the pointing vector of the welding gun end and calculate the angle between it and the normal vector of the groove center to obtain the welding gun attitude normal deviation value; S5: Substitute the welding torch posture normal deviation value and cosine value into the preset pulse frequency operator to output the control coefficient for adjusting the arc stiffness, so as to shorten the peak current duration, increase the pulse frequency and correct the wire feeding speed.

[0023] Based on the above method, the present invention also discloses a system for performing the method, specifically: The data acquisition and topology mapping module is used to acquire robot joint rotation angle variables, link length constants, link radius constants, bevel center normal vectors, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topology feature matrix; The envelope construction and interference calculation module is used to construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing. The curvature-guided attitude reconstruction module is used to keep the welding torch end pose unchanged when the minimum interference spacing reaches the preset warning boundary, calculate the curvature gradient of the component surface based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing curvature gradient of the component surface, and generate a joint space compensation vector. The normal deviation quantization module is used to extract the pointing vector of the welding torch end based on the joint space compensation vector and calculate the angle between it and the normal vector of the groove center to obtain the normal deviation value of the welding torch attitude. The arc and wire feeding coordinated control module is used to substitute the welding torch posture normal deviation value into a preset pulse frequency operator, shorten the peak current duration and increase the pulse frequency, use the cosine value of the welding torch posture normal deviation value to correct the wire feeding speed, and output a control coefficient for adjusting the arc stiffness.

[0024] It should be noted that, in order to ensure the effective implementation of the control algorithm described in this invention on the physical device side, after completing the calculation of the control coefficient and various compensation quantities, the main control system will send the underlying control commands, including the "comprehensive control value", to the underlying all-digital welding machine communication board and wire feeder servo driver through an industrial-grade real-time communication bus (such as EtherCAT or DeviceNet protocol); for some traditional interface types of welding power supplies, the main control system can also convert the control parameters into standard analog voltage signals through the corresponding digital-to-analog conversion module for control.

[0025] Through the aforementioned communication link, the system can intervene in the physical motor and welding power source with a microsecond delay, thereby achieving physical closed-loop control of arc stiffness adjustment and wire feed speed correction.

[0026] like Figure 2 As shown, in some embodiments of the present invention, step S1 specifically includes S11-S13, which are specifically executed by the data acquisition and topology mapping module: S11: Obtain robot joint angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors. Based on the joint angle variables and link length constant, construct a homogeneous transformation matrix. Perform homogeneous transformation operation to obtain the three-dimensional coordinates of the endpoint of the link's central axis. Extract the three-axis boundary extreme values ​​of the endpoint's three-dimensional coordinates. Arrange the boundary extreme values ​​and the endpoint's three-dimensional coordinates as column vectors to establish the link's spatial pose matrix. In one specific embodiment of the present invention, the "link radius constant" not only represents the actual physical outer diameter of the robotic arm link, but also includes a built-in safety buffer margin. For example, if the maximum physical radius of a certain link of the robotic arm is 80mm, to prevent sudden dynamic interference at the construction site and to consider the following error when the robotic arm moves at high speed, the system reserves a safety buffer margin of 20mm, and the link radius constant for that segment is set to 100mm. Through this setting, a virtual protective shield can be formed in advance when constructing the dynamic cylindrical envelope model of the link, effectively improving the fault tolerance rate of obstacle avoidance planning and construction safety.

[0027] S12: Call the point cloud coordinates and sampling point normal vectors of the component surface, perform spatial translation and rotation transformation using the preset basic coordinate system transformation matrix, obtain the projection coordinates and projection normal vectors in the basic coordinate system, perform subtraction operation between the projection coordinates and the corresponding dimension coordinates of the link spatial pose matrix, extract the absolute value of the coordinate difference and perform array arrangement to obtain the surface projection difference set. S13: For the set of surface projection differences, construct an initial feature matrix, extract the absolute values ​​within the set and place them on the main diagonal of the initial feature matrix, fill the projection coordinates and projection normal vectors into the off-diagonal positions of the initial feature matrix according to the associated dimensions, perform normalization operation on the initial feature matrix, scale the matrix elements to a preset reference range, and generate a geometric topological feature matrix for the work space.

[0028] In the process of generating the geometric topological feature matrix, the "preset reference interval" for performing normalization operations on the initial feature matrix is ​​preferably [0, 1] or [-1, 1]. Scaling the matrix elements to this interval primarily aims to eliminate the dimensional differences and numerical magnitude effects between physical quantities of different dimensions (e.g., millimeter-level spatial coordinates and dimensionless normal vectors). This not only prevents large numerical features from abnormally dominating subsequent calculations but also significantly improves the convergence speed and numerical stability of spatial point-pair distance calculation and interferometric detection algorithms.

[0029] In a specific implementation scenario of this invention, the extracted six-axis rotation angle variable sequence of a construction site robot is assigned values ​​of 0.52, 0.78, -1.04, 0, 1.57, and 0 radians, and the corresponding six-segment link length constant sequence is assigned values ​​of 100, 400, 400, 0, 0, and 100 millimeters. The physical maximum radius of the third link of the robotic arm (80 millimeters) is added to a preset safety buffer margin of 20 millimeters to obtain a link radius constant set to 100 millimeters. The bevel normal vector is assigned values ​​of 0, 0, and 1, the component surface coordinates are assigned values ​​of 500, 200, and 100, and the sampled normal vector is assigned values ​​of 0, 0.707, and 0.707. The components corresponding to the rotation angle variable sequence and the link length constant sequence are substituted into the corresponding trigonometric function terms of a four-by-four standard transformation matrix to perform multiplication and addition operations to obtain the rotation and translation transformation matrix of each joint. The transformation matrices of adjacent joints are then subjected to moment... The matrix multiplication operation yields a cumulative matrix from the base to the ends of each link. The first three rows of the fourth column of the cumulative matrix are extracted as the three-dimensional coordinates of the corresponding link axis endpoints. Taking the third link as an example, let its starting point three-dimensional coordinates be (100, 100, 200) and its endpoint three-dimensional coordinates be (300, 200, 400). The X-axis, Y-axis and Z-axis coordinate values ​​of the starting point and endpoint are compared respectively, and the boundary maxima and minima of the three axes are extracted, namely the X-axis maxima 300 and minima 100, the Y-axis maxima 200 and minima 100, and the Z-axis maxima 400 and minima 200. The obtained minima sequence 100, 100, 200 and maxima sequence 300, 200, 400, and the endpoint three-dimensional coordinates (300, 200, 400) are arranged in column order to form a 3×3 two-dimensional array, thus establishing the spatial pose numerical array corresponding to the third link. The point cloud coordinate parameters with values ​​of 500, 200, and 100, and the normal vector parameters with values ​​of 0, 0.707, and 0.707 are retrieved. A pre-defined basic translation transformation term consisting of a 10 mm X-axis translation, a 10 mm Y-axis translation, and a 0 mm Z-axis translation is introduced. The coordinate parameters and the translation transformation term are then added sequentially along the corresponding X, Y, and Z axes to obtain the projected coordinates in the basic coordinate system as (510, 210, 100). Simultaneously, the normal vector parameters are multiplied by the third-order unit rotation matrix of the basic coordinate system to obtain the basic coordinate system... The projection normal vectors are kept constant at 0, 0.707, and 0.707. The projection coordinates (510, 210, 100) are subtracted from the corresponding values ​​of the endpoint three-dimensional coordinate column vectors 300, 200, and 400 in the pose array constructed in the previous step in three axial dimensions to obtain coordinate difference vectors of 210, 10, and -300. The absolute values ​​of each value in the coordinate difference vectors are extracted to obtain absolute value vectors of 210, 10, and 300. The absolute value vectors are then directly arranged vertically in the column direction to generate a single-column surface projection difference dataset composed of the absolute value vectors. Based on the single-column difference dataset of element values ​​210, 10, and 300 obtained in the previous steps, a 3x3 all-zero initial matrix is ​​constructed. These three absolute values ​​are extracted from the dataset and used to replace the zero elements in the first row, first column, second row, second column, and third row, third column of the all-zero initial matrix. The previously obtained projection coordinate components 510, 210, and 100 are then sequentially filled into the positions of the first row, second column, second row, third column, and third row, first column of the all-zero initial matrix. Similarly, the projection normal vector components 0, 0.707, and 0.707 are sequentially filled into the positions of the first row, third column, second row, first column, and third row, second column. Finally, the system-defined linear mapping scaling formula is retrieved. , Where parameter x norm The value represents the result after mapping, and x represents the value in the original square matrix that has been manipulated. min With x max These represent the minimum and maximum values ​​within the original matrix, respectively; both are numerical and dimensionless. The minimum element 0 and the maximum element 510 in the matrix are retrieved. Each element value in the matrix, along with the minimum element 0 and the maximum element 510, is substituted into the linear mapping scaling formula, and subtraction and division operations are performed sequentially. This maps all values ​​in the original matrix proportionally to a preset reference interval consisting of the minimum value 0 and the maximum value 1, outputting a geometrically topologically characteristic numerical matrix where all element values ​​are strictly defined between zero and one and possess a unified dimension.

[0030] like Figure 3As shown, in some embodiments of the present invention, step S2 specifically includes S21-S23, which are specifically executed by the envelope construction and interference calculation module: S21: Based on the geometric topology feature matrix, extract the three-dimensional coordinates of the endpoints of the central axis in the link spatial pose matrix, call the link radius constant and perform radial expansion vector superposition on the central axis, construct a set of closed cylindrical surface points around the central axis of each link segment, merge and arrange the closed cylindrical surface point set with the link spatial pose matrix and perform coordinate mapping to generate a set of link envelope surface features; Then, at the endpoints of the central axis of two adjacent connecting rods, a hemispherical envelope surface point set is constructed with the connecting rod radius constant as the radius; the hemispherical envelope surface point set and the closed cylindrical surface point set are subjected to a Boolean union operation to form a capsule-shaped envelope model covering the robot joint; the spatial coordinates of all vertices in the capsule-shaped envelope model are mapped to the geometric topological feature matrix to complete the expansion of the connecting rod envelope surface feature set.

[0031] S22: Based on the feature set of the connecting rod envelope surface, obtain the projected coordinates in the geometric topology feature matrix, calculate the Euclidean distance between the coordinates of each point in the feature set of the connecting rod envelope surface and the projected coordinates, and perform vectorization and arrangement of the Euclidean distance values ​​according to the sampling time order to obtain the spatial coordinate point-to-point distance sequence; S23: Call the spatial coordinate point pair distance sequence, sort the values ​​in the spatial coordinate point pair distance sequence in ascending order, and extract the first minimum value as the minimum interference distance.

[0032] Furthermore, following the aforementioned example, based on the geometric topological feature matrix, the three-dimensional coordinates of the starting point (100, 100, 200) and the three-dimensional coordinates of the endpoints (300, 200, 400) of the central axis of the third link segment are extracted from the link spatial pose matrix. The link radius constant of 100 mm, obtained from the previous steps, is used. A radial expansion vector with a polar radius of 100 is added to the axis coordinates in a circumferential direction perpendicular to the link axis with a rotation step of 15 degrees. This generates a discrete cylindrical surface coordinate point matrix covering the entire length of the link. The coordinates of the starting point are then calculated. A three-dimensional spherical coordinate system transformation with a fixed length of 100 mm is performed at the 3D coordinates (100, 100, 200) and the endpoint 3D coordinates (300, 200, 400). The coordinates of the sampling points with azimuth angles in the range of 0 to 2π and elevation angles in the range of 0 to π are extracted to form a hemispherical shell point set. Coordinate deduplication and Boolean union operations are performed on the cylindrical surface coordinate point matrix and the hemispherical shell point set to form a capsule body envelope model covering the robot joint. The obtained discrete coordinate points are mapped to the geometric description dimension of the feature matrix in column vector format to complete the expansion of the link envelope surface feature set.

[0033] Based on the feature set of the link envelope surface, the projected coordinates (510, 210, 100) obtained from the pre-transformation in the geometric topology feature matrix are obtained. The coordinates (300, 300, 400) of the i-th vertex in the expanded envelope surface point set are extracted. The values ​​of the X-axis, Y-axis, and Z-axis corresponding to the two sets of coordinates are subtracted sequentially to obtain component difference terms 210, -90, and -300. The squares of each component difference term are accumulated and the square root is taken. The calculation is performed using the formula: , Where parameter d represents the Euclidean distance between the two points, x env With y env and z env These represent the three-axis components of the vertex coordinates of the envelope plane, x and y. sur With y sur and z sur These represent the three-axis components of the projected coordinates of the component surface, all of which are numerical and dimensionless. Substituting the aforementioned data, the distance scalar is approximately 377.09 mm. By traversing all vertices within the envelope surface point set and performing the above operation, the scalar values ​​are filled into a one-dimensional vector according to the sampling time order, resulting in a spatial coordinate point-to-distance sequence. The spatial coordinate point distance sequence is invoked to obtain the set of distance scalar values ​​in the sequence, including 377.09 and 280.50, and 415.20 and 300.00. All values ​​in the set are compared and arranged in ascending order. The value 280.50 is determined to be the smallest value in the sequence. The safety buffer limit set by the system based on the maximum braking deceleration parameter of the robotic arm is retrieved and a warning critical interval of 0 to 300 mm is defined. The smallest value is substituted into the warning critical interval for numerical comparison. The first element after sorting is extracted, and the first minimum value is extracted as the minimum interference distance.

[0034] like Figure 4 As shown, in another embodiment of the present invention, step S3 specifically includes S31-S33, which are specifically executed by the curvature-guided attitude reconstruction module: S31: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, call the sampling point normal vector and the corresponding component surface point cloud coordinates, calculate the absolute value of the difference between the three-axis components of the normal vectors of adjacent sampling points, obtain the spatial Euclidean distance between adjacent sampling points, perform a division operation between the three-axis component difference and the spatial Euclidean distance, and generate the component surface curvature gradient value. Maintaining the welding torch end position unchanged means that obstacle avoidance is achieved by adjusting the posture of the intermediate link of the robotic arm. When making the adjustment, it is necessary to keep the target position and target posture of the welding torch tool center point (TCP) relative to the bevel center line or welding trajectory without deviating from the target position and target posture that would affect the welding quality. At this time, the welding torch end still operates continuously according to the original welding trajectory. The relative position between the welding wire end and the bevel, the welding torch pointing, the working angle and the push-pull angle are kept within the preset welding process tolerance range. At the same time, the intermediate link of the robotic arm changes its own spatial occupancy through joint posture adjustment, thereby achieving the avoidance of obstacles on the component surface or on site.

[0035] S32: For the surface curvature gradient value of the component, obtain the sampling angle sequence in the redundant joint angle search space, calculate the gradient change rate of the surface curvature gradient value at the corresponding position of the sampling angle, extract the extreme value terms with negative values ​​in the gradient change rate, arrange the deflection orientation corresponding to the extreme value terms, and obtain the joint deflection guide vector. When performing obstacle avoidance, the posture of the intermediate link of the robotic arm is adjusted in three-dimensional space. It is necessary to prioritize the robotic arm links that can participate in posture reconstruction and meet the requirements of three-dimensional adjustment, and then select the links that have a significant impact on the spatial posture but will not significantly affect the process posture of the welding torch end.

[0036] The redundant joint angle search space is defined by the limiting range of the second to fourth axes of the robotic arm. The movement directions of the second to fourth axes are X, Y, and Z, respectively. Adjusting the second to fourth axes allows for movement adjustment in these three directions within three-dimensional space. Since the first axis is the overall rotation axis of the base, its large adjustment range can easily cause overall deviation of the welding trajectory, making it unsuitable as a priority adjustment axis for localized flexible obstacle avoidance. The fifth and sixth axes directly relate to the welding torch working angle, push-pull angle, and welding wire orientation; frequent use of the fifth and sixth axes for obstacle avoidance can easily disrupt the welding process posture. The second to fourth axes, located in the middle of the robotic arm, have the greatest impact on the enveloping space of the links in the upper arm, forearm, and forearm, effectively changing the relative distance between the robot's intermediate links and the component surface. Therefore, they are limited to the second to fourth axes. Furthermore, the second to fourth axes are adjusted first; only when adjusting the second to fourth axes alone cannot meet the adjustment requirements are the first, fifth, or sixth axes of the robotic arm adjusted.

[0037] S33: Based on the joint deflection guide vector, call the current redundant joint actual rotation angle and the preset deflection step coefficient, perform a multiplication operation on the joint deflection guide vector and the preset deflection step coefficient to obtain the rotation angle compensation offset, and perform a matrix addition operation on the rotation angle compensation offset and the actual rotation angle to generate the joint space compensation vector.

[0038] It should be noted that the preset deflection step size coefficient is generated by performing a mapping operation between the minimum interference spacing and the surface curvature gradient value; the mapping operation includes: obtaining the deviation value of the minimum interference spacing relative to the preset warning boundary, performing a multiplication operation between the reciprocal term of the deviation value and the surface curvature gradient value, and generating a preset deflection step size coefficient that increases as the deviation value decreases.

[0039] Furthermore, the "preset warning boundary" is the trigger threshold for the minimum interference distance, which is the critical condition for initiating redundant joint obstacle avoidance deflection. The preset warning boundary includes an inner forced obstacle avoidance boundary and an outer warning boundary. The setting of the preset warning boundary value comprehensively considers the communication delay of on-site sensors, the operation cycle of the control system, and the physical braking distance of the robotic arm movement. When the outer warning boundary is reached, early flexible obstacle avoidance is triggered, which is set to 300mm. When the inner forced obstacle avoidance boundary is reached, extreme forced obstacle avoidance is triggered. In a typical on-site construction scenario, the inner forced obstacle avoidance boundary is set to 50mm. When the minimum interference distance between the envelope model surface and the component surface point cloud reaches or is less than 300mm, the system determines that there is a potential collision risk and immediately initiates curvature-guided attitude reconstruction to ensure that the obstacle avoidance action is completed within a very short response time.

[0040] For example, in a specific implementation scenario, the minimum interference distance value of 280.50 mm output from the previous step is retrieved and compared with the set warning threshold parameter of 300 mm. When 280.50 mm is within the trigger range of less than 300 mm, the current pulse duty cycle of each drive motor of the control robot arm is locked and no new position update commands are sent. The first projected position coordinates (510, 210, 100) of the component surface in the basic coordinate system stored in the control system and the adjacent second position coordinates (513, 214, 100) are extracted. At the same time, the normal vector component 0 corresponding to the first position obtained in the previous step is retrieved. The components of the normal vector at the second position are 0, 0.707, 0.707, and 0, 0.697, 0.717. The corresponding axis coordinates of the two sets of position coordinates are subtracted, the sum of squares is calculated, and the square root is taken to obtain the spatial straight-line distance between the two positions as 5 mm. The three dimensional components of the two sets of normal vectors are subtracted and the absolute values ​​are extracted to obtain the dimensional differences as 0, 0.010, 0.010, respectively. The three values ​​of 0, 0.010, 0.010 are used as dividends and divided by the spatial straight-line distance of 5 mm, which is used as the divisor, to generate the surface curvature gradient values ​​of the component with the specific values ​​of 0, 0.002, 0.002. The surface curvature gradient values ​​of the three components (0, 0.002, and 0.002) are obtained. The lower limit of the second axis operating boundary (-1.57 radians and upper limit of +1.57 radians), the third axis boundary (-3.14 radians and +1.57 radians), and the fourth axis boundary (-3.14 radians and +3.14 radians) stored in the robotic arm control dictionary are retrieved. Numerical extraction is performed within these three limit intervals at fixed intervals of 0.05 radians, and multiple sets of rotation test sequences are arranged and combined in a matrix row. These multiple sets of rotation test sequences are then successively fused into the system's partial differential equations. The derivative operation of the surface curvature gradient variable is performed, and the gradient change rate data set corresponding to each test sequence is output. The magnitude judgment operation of each value in the data set and the baseline parameter 0 is performed, and the minimum value item with the value in the range of less than 0 is filtered out. The positive deflection identifier +1 of the second axis, the negative deflection identifier -1 of the third axis, and the stationary identifier 0 of the fourth axis associated with the minimum value item are extracted. The above three orientation identifiers +1, -1, and 0 are spliced ​​and arranged in the order of axis number to obtain the joint deflection guide vector with each dimension component being +1, -1, and 0 respectively. Read the joint deflection guide vectors generated above with dimension components of +1, -1, and 0. Obtain the encoder feedback data of each motor at the current moment and retrieve the actual angles of the second axis (0.78 radians), the third axis (-1.04 radians), and the fourth axis (0 radians) from the previous steps. Subtract the warning critical limit parameter of 300 mm from the minimum interference spacing value of 280.50 mm to obtain the difference of 19.50 mm. Substitute this difference into the formula to perform the calculation: , In the formula, parameter K represents the step size adjustment coefficient, which is dimensionless, and parameter d... warn The trigger limit parameter value represents 300 mm, dimensionless, parameter d. min Represents the minimum pitch value of 280.50 mm, dimensionless, parameter G. max The largest numerical term in the previously acquired surface curvature gradient, 0.002, is the curvature driving quantity after normalization and angle mapping gain processing. Its value ranges from 0 to 1 and is dimensionless. Redundant joint obstacle avoidance deflection is only activated when a warning is triggered; therefore, d... warn Greater than d min Parameter G maxThe value is selected from the surface curvature gradient of the component, and can be zero. The reciprocal of the phase difference of 19.50 mm is calculated to be approximately 0.051, and multiplied by 0.002 to obtain a coefficient value of 0.0001 radians. The above three guiding components of +1, -1, and 0 are multiplied by 0.0001 radians respectively to output deflection corrections of 0.0001 radians, -0.0001 radians, and 0 radians. The 0.0001 radians, -0.0001 radians, and 0 radians are arranged into column vectors and then added to the corresponding terms of the three current state vectors of 0.78 radians, -1.04 radians, and 0 radians to generate joint space compensation vectors with each row element of 0.7801 radians, -1.0401 radians, and 0 radians respectively.

[0041] like Figure 5 As shown, in some embodiments of the present invention, step S4 specifically includes S41-S43, which are specifically executed by the normal deviation quantization module: S41: Based on the joint space compensation vector, call the preset forward kinematics transformation matrix of the robotic arm to perform a multiplication operation, extract the direction vector elements of the posture rotation matrix in the matrix operation result, perform normalization processing on the direction vector elements, and obtain the three-dimensional component of the welding gun pointing. The preset forward kinematics transformation matrix of the robotic arm is obtained by multiplying the pose transformation matrix of the robot end flange and the calibration matrix of the welding gun tool center point; the direction vector element of the posture rotation matrix in the extracted matrix operation result is specifically extracted as the Z-axis approach vector along the welding wire extension direction in the calibration matrix of the welding gun tool center point, so as to characterize the actual pointing of the welding gun in the actual physical space.

[0042] S42: Call the welding torch pointing three-dimensional component and the bevel center normal vector, perform dot product operation on the corresponding coordinate axis components of the two, and extract the absolute value of the dot product operation result to obtain the normal dot product scalar; S43: For the normal dot product scalar, call the inverse cosine calculation formula to obtain the spatial angle value. After converting the spatial angle value into standard angle units, perform a subtraction operation with the preset reference vertical angle constant to generate the welding gun posture normal deviation value.

[0043] It should be noted that the preset reference vertical angle constant is composed of an ideal normal reference and a preset welding process deflection compensation angle. The welding process deflection compensation angle includes a welding torch working angle parameter set based on a specific groove type, and a push-pull angle parameter set along the welding path travel direction.

[0044] Furthermore, the "welding process deflection compensation angle" is a correction basis tailored to the specific bevel type and welding process requirements of the construction site. Taking common T-joints or V-groove multi-pass welding as an example, this compensation angle specifically includes the welding torch working angle and push-pull angle parameters. In the actual implementation of T-joint fillet welding, the working angle parameter is set to 45 degrees to ensure uniform fusion of the base materials on both sides; at the same time, the push-pull angle parameter along the welding path is set to a push angle of 10 degrees (i.e., the welding torch tilts 10 degrees in the welding direction) to obtain a smoother weld surface and appropriate penetration depth. These specific parameters, together with the ideal normal reference, constitute the preset reference vertical angle constant.

[0045] For example, based on the joint space compensation vectors generated in the previous steps, with each row containing elements of 0.7801 radians, -1.0401 radians, and 0 radians respectively, a preset flange pose transformation matrix for the robotic arm end effector is called. The three-dimensional translation coordinate constants of the flange pose transformation matrix are set to 100.0 mm, 200.0 mm, and 300.0 mm, respectively, and the rotation parameter is a third-order unit orthogonal matrix. The welding torch tool center point calibration matrix is ​​called, and the X-axis and Y-axis position offset constants of the calibration matrix are set to 0.0 mm, the Z-axis position offset constant to 150.0 mm, and the rotation angle constant around the Y-axis to 30 degrees. The flange pose transformation matrix is ​​then... The corresponding values ​​in each row and column of the calibration matrix of the welding gun tool center point are multiplied and added to obtain the forward kinematic transformation matrix of the robotic arm. The numerical coordinates of the first to third rows of the third column in the forward kinematic transformation matrix of the robotic arm are extracted to form the Z-axis approach vector along the welding wire extension direction. The initial three-axis component values ​​of the approach vector are obtained as 0.5, 0.0 and 0.866. The sum of squares of the initial three-axis component values ​​is calculated as 1.0 and its positive square root is extracted to obtain the modulus parameter 1.0. The three coordinate items of the initial three-axis component values ​​are divided by the modulus parameter 1.0 respectively to perform a proportional scaling operation to obtain the three-dimensional component of the welding gun pointing.

[0046] The welding torch pointing three-dimensional component is called, along with the bevel center normal vector assigned values ​​of 0, 0, and 1 in the previous steps. The X-axis component (0.5), Y-axis component (0.0), and Z-axis component (0.866) of the welding torch pointing three-dimensional component obtained in the previous operation are extracted. The X-axis component (0.5) of the welding torch pointing three-dimensional component is multiplied by the X-axis component (0) of the bevel center normal vector to obtain the first value (0.0). The Y-axis component (0.0) of the welding torch pointing three-dimensional component is multiplied by the Y-axis component (0) of the bevel center normal vector to obtain the second value (0.0). The first value 0.0 is multiplied by the Z-axis component 0.866 of the welding torch pointing three-dimensional component and the Z-axis component 1 of the bevel center normal vector to obtain a third value 0.866. The first value 0.0, the second value 0.0, and the third value 0.866 are then added together to obtain a scalar sum value 0.866. The scalar sum value 0.866 is determined to be in the positive number determination interval greater than or equal to zero. The scalar sum value 0.866 itself is extracted to replace the absolute value transformation operation to obtain the normal dot product scalar.

[0047] For the previously obtained normal dot product scalar value of 0.866, substitute it into the inverse trigonometric function operation term to obtain the radian spatial angle value of 0.5236. Multiply the radian spatial angle value of 0.5236 by the angle transformation constant 180 and divide by pi 3.14159 to obtain the standard angle unit value of 30.0 degrees. Set the ideal normal reference angle parameter for a specific bevel type of T-joint at the construction site to 0.0 degrees. Extract the preset welding torch working angle parameter of 45.0 degrees and the push angle parameter of 10.0 degrees set along the welding path travel direction. Add the ideal normal reference angle parameter of 0.0 degrees, the welding torch working angle parameter of 45.0 degrees, and the push angle parameter of 10.0 degrees to obtain the preset reference vertical angle constant of 55.0 degrees. Use the formula to integrate the aforementioned numerical relationships: , Wherein, parameter Dnorm represents the output normal deviation value in degrees; parameter Sdot represents the obtained normal dot product scalar, which is the absolute value of the dot product of the welding torch pointing unit vector and the bevel center unit normal vector. Before substituting parameter Sdot into the inverse cosine function, the welding torch pointing unit vector and the bevel center unit normal vector are normalized, and Sdot is limited to the range [0,1], dimensionless; parameter θideal represents the set ideal normal reference angle in degrees; parameter θwork represents the called welding torch working angle in degrees; parameter θ... push The reference angle, in degrees, is used to calculate the difference of -25.0 degrees by subtracting the standard angle unit value of 30.0 degrees from the reference vertical angle constant of 55.0 degrees, thus generating the welding torch attitude normal deviation value.

[0048] like Figure 6As shown, in some embodiments of the present invention, step S5 specifically includes S51-S53, which are specifically executed by the arc and wire feeding coordinated control module: S51: Call the welding torch attitude normal deviation value and substitute it into the preset pulse frequency operator to obtain the preset reference period parameter and bias constant. Multiply the welding torch attitude normal deviation value and the reference period parameter to obtain the period deviation term. Add the period deviation term to the bias constant to generate the current period control value. Shorten the peak current duration and increase the pulse frequency through the current period control value. S52: Perform cosine operation on the normal deviation value of the welding torch attitude to extract the cosine value, call the preset reference wire feeding speed, multiply the cosine value by the reference wire feeding speed to obtain the basic variable of wire feeding, and perform division operation on the current cycle control value and the basic variable of wire feeding to obtain the wire feeding speed compensation amount. S53: Call the preset proportional gain parameter, multiply the wire feeding speed compensation amount by the proportional gain parameter to obtain the gain compensation term, add the gain compensation term to the current cycle control value to obtain the comprehensive control value, compare the comprehensive control value with the preset extreme value constant for amplitude limiting, and output the control coefficient used to adjust the arc stiffness.

[0049] The "preset extreme value constant" used for comparison and limiting is a key safety constraint to ensure the stability of the welding process. In actual welding operations, if the control coefficients of wire feed speed and arc stiffness are not rigidly limited, it is easy for thin plates to burn through due to excessive instantaneous current compensation, or for the arc to break due to excessive reduction in wire feed speed. Therefore, in this embodiment, the preset extreme value constant is usually set as ±20% of the reference parameter as the upper and lower limit fluctuation range. For example, if the calculated comprehensive control value exceeds 120% of the reference value, the system will forcibly limit its output to within 120%, thereby achieving dynamic adjustment while firmly maintaining the bottom line of welding arc stability and forming quality.

[0050] Following the previous example, the welding torch attitude normal deviation value of -25.0 degrees generated in the previous steps is called and substituted into the operation storage node of the preset pulse frequency operator. The system preset reference period parameter of 0.004 and the bias constant of 1.0 are obtained. The welding torch attitude normal deviation value of -25.0 degrees and the reference period parameter of 0.004 are multiplied to obtain a period deviation term of -0.1. The period deviation term of -0.1 and the bias constant of 1.0 are added to obtain an addition result of 0.9, and the current period control value is generated. The welding torch attitude normal deviation value of -25.0 degrees obtained in the previous step is retrieved and converted to -0.436 radians. Then, a cosine trigonometric function extraction operation is performed to obtain a cosine value of 0.906. The reference wire feed speed parameter of 5.0 preset in the system processing database is called. The cosine value of 0.906 and the reference wire feed speed parameter of 5.0 are multiplied to obtain a wire feed basic variable of 4.530. The current cycle control value of 0.9 generated in the previous step is extracted as the dividend and the wire feed basic variable of 4.530 is used as the divisor to perform a division operation to obtain a quotient of 0.198, thus obtaining the wire feed speed compensation amount. The system calls the preset proportional gain parameter of 0.5 within the control unit. It then multiplies the wire feed speed compensation (value 0.198) obtained in the previous step with the proportional gain parameter of 0.5 to obtain a gain compensation term of 0.099. Next, it adds the gain compensation term of 0.099 with the current cycle control value of 0.9 extracted in the previous step to obtain a comprehensive control value of 0.999. Finally, it retrieves the system-set reference parameter of 1.0 and, combined with a preset floating boundary coefficient of 20%, performs subtraction and addition operations to obtain a preset extreme value constant range containing a lower limit of 0.8 and an upper limit of 1.2. Finally, it performs a numerical boundary comparison operation using the limiting calculation formula. , Where parameter C ctrl The output control value is dimensionless, parameter V. comp The value representing the wire feed speed compensation is dimensionless, and the parameter K is used. p The value of the proportional gain parameter, T, is dimensionless and represents the input value. adj The value representing the current cycle regulation value is dimensionless, and the parameter L is... min The lower limit value to be extracted is dimensionless, and the parameter L is... max The value represents the upper limit of the extracted value and is dimensionless. The comprehensive control value of 0.999 is sequentially compared with the lower limit value of 0.8 and the upper limit value of 1.2 to determine their magnitude. If the comprehensive control value of 0.999 is determined to be within the range of greater than 0.8 and less than 1.2, the comprehensive control value of 0.999 itself is directly extracted as the output value, and the control coefficient is output.

[0051] The basic concepts have been described above. Obviously, for those skilled in the art, the detailed disclosure above is merely illustrative and does not constitute a limitation of this specification. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this specification. Such modifications, improvements, and corrections are suggested in this specification and therefore remain within the spirit and scope of the exemplary embodiments described herein.

[0052] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0053] Furthermore, those skilled in the art will understand that various aspects of this specification can be described and illustrated in several patentable ways or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Accordingly, various aspects of this specification can be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. All of the above hardware or software may be referred to as a “data block,” “module,” “engine,” “unit,” “component,” or “system.” Furthermore, various aspects of this specification may be represented as a computer product located on one or more computer-readable media, including computer-readable program code.

[0054] Computer storage media may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and suitable combinations thereof. Computer storage media can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer storage medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0055] The computer program code required for the operation of each part of this manual can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc.; conventional procedural programming languages ​​such as C, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP; dynamic programming languages ​​such as Python, Ruby, and Groovy; or other programming languages. This program code can run entirely on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as Software as a Service (SaaS).

[0056] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A welding control method for a welding robot applied on a construction site, characterized in that, include: S1: Obtain the robot joint rotation angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topological feature matrix; S2: Construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing; S3: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, calculate the surface curvature gradient of the component based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing surface curvature gradient of the component, and generate a joint space compensation vector. Step S3 includes: S31: If the minimum interference spacing reaches the preset warning boundary, keep the welding torch end pose unchanged, call the sampling point normal vector and the corresponding component surface point cloud coordinates, calculate the absolute value of the three-axis component difference of the normal vector of adjacent sampling points, obtain the spatial Euclidean distance between adjacent sampling points, perform a division operation between the three-axis component difference and the spatial Euclidean distance, and generate the component surface curvature gradient value. S32: For the surface curvature gradient value of the component, obtain the sampling angle sequence in the redundant joint angle search space, calculate the gradient change rate of the surface curvature gradient value at the corresponding position of the sampling angle, extract the extreme value terms with negative values ​​in the gradient change rate, arrange the deflection orientation corresponding to the extreme value terms, and obtain the joint deflection guide vector. S33: Based on the joint deflection guide vector, call the current redundant joint actual rotation angle and preset deflection step coefficient, perform multiplication operation on the joint deflection guide vector and the preset deflection step coefficient to obtain the rotation angle compensation offset, perform matrix addition operation on the rotation angle compensation offset and the actual rotation angle to generate the joint space compensation vector; S4: Based on the joint space compensation vector, extract the pointing vector of the welding gun end and calculate the angle between it and the normal vector of the groove center to obtain the welding gun attitude normal deviation value; S5: Substitute the welding torch posture normal deviation value and cosine value into the preset pulse frequency operator to output the control coefficient for adjusting the arc stiffness, so as to shorten the peak current duration, increase the pulse frequency and correct the wire feeding speed. Step S5 includes: S51: Call the welding torch attitude normal deviation value and substitute it into the preset pulse frequency operator to obtain the preset reference period parameter and bias constant. Multiply the welding torch attitude normal deviation value and the reference period parameter to obtain the period deviation term. Add the period deviation term to the bias constant to generate the current period control value. Shorten the peak current duration and increase the pulse frequency through the current period control value. S52: Perform cosine operation on the normal deviation value of the welding torch attitude to extract the cosine value, call the preset reference wire feeding speed, multiply the cosine value by the reference wire feeding speed to obtain the basic variable of wire feeding, and perform division operation on the current cycle control value and the basic variable of wire feeding to obtain the wire feeding speed compensation amount. S53: Call the preset proportional gain parameter, multiply the wire feeding speed compensation amount by the proportional gain parameter to obtain the gain compensation term, add the gain compensation term to the current cycle control value to obtain the comprehensive control value, compare the comprehensive control value with the preset extreme value constant for amplitude limiting, and output the control coefficient used to adjust the arc stiffness.

2. The welding control method for a welding robot applied to a construction site according to claim 1, characterized in that the steps are as follows: S1 includes: S11: Obtain robot joint angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors. Based on the joint angle variables and link length constant, construct a homogeneous transformation matrix. Perform homogeneous transformation operation to obtain the three-dimensional coordinates of the endpoints of the link center axis. Extract the boundary extrema of the three axes of the endpoint three-dimensional coordinates. Arrange the boundary extrema and the endpoint three-dimensional coordinates as column vectors to establish the link spatial pose matrix. S12: Call the point cloud coordinates and sampling point normal vectors of the component surface, perform spatial translation and rotation transformation using the preset basic coordinate system transformation matrix, obtain the projection coordinates and projection normal vectors in the basic coordinate system, perform subtraction operation between the projection coordinates and the dimension coordinates corresponding to the link spatial pose matrix, extract the absolute value of the coordinate difference and perform array arrangement to obtain the surface projection difference set. S13: For the set of surface projection differences, construct an initial feature matrix, extract the absolute values ​​within the set and place them on the main diagonal of the initial feature matrix, fill the projection coordinates and projection normal vectors into the off-diagonal positions of the initial feature matrix according to the associated dimensions, perform normalization operation on the initial feature matrix, scale the matrix elements to a preset reference range, and generate a geometric topological feature matrix for the work space.

3. The welding control method for a welding robot applied on a construction site according to claim 2, characterized in that, Step S2 includes: S21: Based on the geometric topology feature matrix, extract the three-dimensional coordinates of the endpoints of the central axis of the link spatial pose matrix, call the link radius constant and perform radial expansion vector superposition on the central axis to construct a set of closed cylindrical surface points around the central axis of each link segment, merge and arrange the closed cylindrical surface point set with the link spatial pose matrix and perform coordinate mapping to generate a set of link envelope surface features; S22: Based on the feature set of the connecting rod envelope surface, obtain its projected coordinates in the geometric topological feature matrix, calculate the Euclidean distance between the coordinates of each point in the feature set of the connecting rod envelope surface and its projected coordinates, and perform vectorization and arrangement of the Euclidean distance values ​​according to the sampling time order to obtain the spatial coordinate point-to-point distance sequence. S23: Call the spatial coordinate point pair distance sequence, sort the values ​​in the spatial coordinate point pair distance sequence in ascending order, and extract the first minimum value as the minimum interference distance.

4. The welding control method for a welding robot applied on a construction site according to claim 3, characterized in that, The specific operations for constructing the link envelope surface feature set in step S21 are as follows: At the endpoints of the central axis of two adjacent link segments, construct a hemispherical envelope surface point set with the link radius constant as the radius; perform a Boolean union operation on the hemispherical envelope surface point set and the closed cylindrical surface point set to form a capsule-shaped envelope model covering the robot joint; map the spatial coordinates of all vertices in the capsule-shaped envelope model to the geometric topological feature matrix to complete the expansion of the link envelope surface feature set.

5. The welding control method for a welding robot applied on a construction site according to claim 1, characterized in that, The redundant joint rotation angle search space is defined by the limiting interval from the second axis to the fourth axis of the robotic arm; the preset deflection step size coefficient is generated by performing a mapping operation between the minimum interference spacing and the surface curvature gradient value; The mapping operation includes: obtaining the deviation value of the minimum interference spacing relative to the preset warning boundary, performing a multiplication operation between the reciprocal term of the deviation value and the surface curvature gradient value, and generating a preset deflection step size coefficient that increases as the deviation value decreases.

6. The welding control method for a welding robot applied on a construction site according to claim 1, characterized in that, Step S4 includes: S41: Based on the joint space compensation vector, call the preset forward kinematics transformation matrix of the robotic arm to perform a multiplication operation, extract the direction vector elements of the posture rotation matrix in the matrix operation result, perform normalization processing on the direction vector elements, and obtain the three-dimensional component of the welding gun pointing. S42: Call the welding torch pointing three-dimensional component and the bevel center normal vector, perform dot product operation on the corresponding coordinate axis components of the two, and extract the absolute value of the dot product operation result to obtain the normal dot product scalar; S43: For the normal dot product scalar, call the inverse cosine calculation formula to obtain the spatial angle value. After converting the spatial angle value into standard angle units, perform a subtraction operation with the preset reference vertical angle constant to generate the welding gun posture normal deviation value.

7. A welding control method for a welding robot applied at a construction site according to claim 6, characterized in that, The preset reference vertical angle constant is composed of an ideal normal reference and a preset welding process deflection compensation angle; the welding process deflection compensation angle includes a welding torch working angle parameter set based on a specific groove type, and a push-pull angle parameter set along the welding path travel direction.

8. A welding robot control system for construction sites, used to perform the method according to any one of claims 1-7, characterized in that, include: Data acquisition and topology mapping module: used to acquire robot joint rotation angle variables, link length constant, link radius constant, bevel center normal vector, component surface point cloud coordinates and sampling point normal vectors, and project them onto the robot's basic coordinate system to generate a geometric topology feature matrix; Envelope construction and interference calculation module: used to construct a dynamic cylindrical envelope model of the link based on the geometric topological feature matrix, calculate the minimum Euclidean distance between the point cloud coordinates of the envelope model surface and the component surface, and obtain the minimum interference spacing; Curvature-guided attitude reconstruction module: used to keep the welding torch end pose unchanged when the minimum interference spacing reaches the preset warning boundary, calculate the curvature gradient of the component surface based on the sampling point normal vector, drive the redundant joint to deflect in the direction of decreasing curvature gradient of the component surface, and generate a joint space compensation vector; Normal deviation quantization module: used to extract the welding torch end pointing vector based on the joint space compensation vector and calculate the angle between it and the bevel center normal vector to obtain the welding torch attitude normal deviation value; Arc and wire feeding coordinated control module: used to substitute the welding torch posture normal deviation value into a preset pulse frequency operator, shorten the peak current duration and increase the pulse frequency, use the cosine value of the welding torch posture normal deviation value to correct the wire feeding speed, and output a control coefficient for adjusting the arc stiffness.

Images