Digital-twin-based automatic programming method for double-welding robot

Abstract

The application provides a kind of automatic programming method of double welding robot based on digital twinning. One of the automatic programming methods of double welding robot based on digital twinning specifically includes twinning of robot and work scene, weld seam trajectory optimization, automatic programming of robot motion code; obtain the three-dimensional model of robot and workpiece, establish twin space; add the parent-child relationship between each joint in the robot model, establish the kinematic model of robot; shoot the workpiece through binocular camera, collect weld seam position information and process; plan the multi-weld seam task of double robot, extract task parameters, establish algorithm operator, iterate to genetic algorithm convergence; generate the motion code of robot, and debug in the twin space, if the debugging is passed, send to the entity robot; the application is convenient and fast, good flexibility, lower skilled requirement for operator, easy to use, can improve the efficiency and precision of welding operation.

Description

Technical Field

[0001] This invention proposes an automatic programming method for welding robots based on digital twins, specifically relating to the field of welding robot teaching technology. Background Technology

[0002] Welding robots, with their advantages of high repeatability, high efficiency, stability, reliability, and ability to perform welding work even in extreme environments, have become an important means of reducing production costs and improving product competitiveness. Welding robots require methods to clearly define the relative position and orientation of the weld seam in space. Commonly used methods include drag-and-drop teaching, teach pendant teaching, and offline programming teaching.

[0003] Dragging teaching involves the worker dragging the robot's end effector along a welding trajectory to teach it to weld. While this method allows the robot to weld along a fixed path, it cannot achieve high precision due to operator observation errors and inherent joint damping in the robot. Furthermore, this technique has poor adaptability to external conditions; in complex environments or when the workpiece clamping error is large, it will produce even greater deviations.

[0004] Teach pendant teaching refers to controlling a robot via a teach pendant to walk along the required work trajectory and recording the trajectory, thereby completing the teaching process. This method of teaching and programming industrial robots places high demands on the programming skills of the operators, and it suffers from problems such as too many teach points and long programming cycles when programming complex trajectories.

[0005] Offline programming teaching is a computational graphics-based robot offline programming teaching technology. It involves reconstructing the robot model and the working environment model, calculating and planning the robot's motion trajectory offline for programming. Conventional offline programming is fast, but its adaptability is poor. Adjustments are cumbersome when the actual welding scenario changes, and it is easy to become out of touch with reality. Summary of the Invention

[0006] To address the shortcomings of existing welding robot programming technologies, this invention provides an automatic programming method for welding robots based on digital twin technology, which achieves high-precision acquisition of the spatial position of the weld seam while providing a more flexible and faster programming method.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an automatic programming method for a dual welding robot based on digital twins. The specific steps are as follows:

[0008] Step S1: Install a binocular point cloud camera at the end of robot R1, calibrate the positional relationship between robot R1 and robot R2 and the binocular point cloud camera, and obtain the homogeneous transformation matrix M1 from robot R2 to the base coordinates of robot R1 and the homogeneous transformation matrix M2 from the binocular camera to the base coordinates of robot R1.

[0009] Step S2: Use the DH robot kinematics modeling method to establish the forward and inverse kinematics models of robot R1 and robot R2.

[0010] Step S3: Draw 3D models of robots R1 and R2, the workpiece to be welded, and related equipment; establish a digital twin programming system for the welding robots; import the drawn 3D models; and recreate the digital twin space of the welding robots.

[0011] Step S4: Acquire workpiece point cloud data through a binocular point cloud camera, perform filtering, resampling, segmentation and other processing on the data, extract weld point cloud information, and apply homogeneous transformation matrix M2 to transform the point cloud coordinates into coordinates in the robot R1 base coordinate system.

[0012] Step S5: Perform twinning on the weld point cloud extracted in step S4, and perform B-spline curve fitting and equal-interval rediscretion on the weld to obtain discrete welding trajectory points.

[0013] Step S6: Perform robot welding task planning, assigning welding tasks to robots R1 and R2. Based on the robot motion code format used and the discrete welding trajectory points obtained in step S5, pre-determine the welding process parameters required for the robot motion code, and automatically generate the robot motion code.

[0014] Step S7: Based on the robot kinematic model established in step S2, calculate the joint axis rotation angle corresponding to the welding trajectory point, send it to the twin space for motion simulation, and observe whether the welding task requirements can be met. If not, repeat steps S4-S7.

[0015] Step S8: If the requirements are met, the motion code generated in step S6 is sent to the physical robot to complete the automatic programming of the welding robot's digital twin.

[0016] In step S3, the initial position of the coordinate system fixed at the rotation axis of robot R1 is set according to the external dimensions of robot R1, and the initial position of the coordinate system fixed at the rotation axis of robot R2 is set according to the external dimensions of robot R2. Further, the joint models of robot R1 and robot R2 are imported sequentially according to the rotation order. Parent-child relationships of robot joint axes are added and angle variables are bound so that motion simulation can be achieved through data in subsequent twinning.

[0017] In step S6, the step of optimizing the sorting of multiple welding tasks is as follows:

[0018] S6-1. For a single weld seam welding task, if the weld seam is only within the workspace of robot R1 or robot R2, then the weld seam is assigned to robot R1 or robot R2; if the weld seam is at the intersection of the workspaces of robot R1 and robot R2, then the intersection of the workspaces of robot R1 and robot R2 is used as the dividing point to assign their respective weld seams.

[0019] S6-2. For multi-weld welding tasks, perform optimal sorting of the two robots. The specific steps are as follows:

[0020] Extract the weld seam task assignment parameters. Constants: weld seam number sequence S, weld seam length L, robot end effector initial point P, robot running speed V, robot R... i Occupied workspace O i The number of algorithm iterations, z. Variable: The weld sequence U assigned by the robot. i :

[0021] U i ={u i1 ,u i2 ,u i3 ,u i4 ,u i5 ,......,u ij}

[0022] In the formula, U i The weld sequence assigned to the i-th robot, u ij The weld with sequence number j is assigned to the i-th robot.

[0023] The welding direction sequence D assigned by the robot to the weld seam i :

[0024] D i ={d i1 ,d i2 ,d i3 ,d i4 ,d i5 ,......,d ij}

[0025] In the formula, D i The welding direction sequence of the weld seam assigned to the i-th robot, d ij The welding direction of the weld seam with sequence number j assigned to the i-th robot. d ij ∈{0,1}, where 0 represents the welding robot welding from the start point to the end point of the weld, and 1 represents the welding robot welding from the end point to the start point of the weld.

[0026] S6-3. Sequentially determine the relationship between the weld and the workspace of each robot. If it belongs exclusively to robot R... iThe workspace is then classified as part of the weld sequence U. i If the robots are in the same workspace, robots with odd-numbered IDs are assigned to robot R1, and robots with even-numbered IDs are assigned to robot R2. The initial robot assignment weld sequence U is generated according to this rule. If the weld position exceeds the workspace of both robots simultaneously, an error is reported.

[0027] S6-4. Establish a dual-robot genetic algorithm operator, including one genetic algorithm chromosome representing weld seam assignment and sorting, and one genetic algorithm chromosome representing weld seam welding direction; its structure is as follows: Figure 7 As shown;

[0028] S6-5. Randomly generate 8 sets of operators as the initial genetic population.

[0029] S6-6. The shortest total distance traveled by both robots is used as the fitness function of the genetic algorithm. The total distance W can be expressed by the formula:

[0030]

[0031] In the formula, l i This represents the length of the i-th weld seam to be welded. This represents the travel distance between the i-th weld and the next weld to be welded. This represents the distance traveled without load from the initial position of the robot's end effector to the starting point of the first weld seam. This represents the distance traveled without moving from the end of the last weld seam back to the starting point of the robot's end effector.

[0032] S6-7 Calculate the genetic fitness function of the 8 sets of operators in the genetic population, and extract the optimal operator with the smallest fitness function value.

[0033] S6-8 Perform mutation, crossover, and other operations on the seven genetic operators other than the optimal operator.

[0034] Operator sequence mutation illustration as follows Figure 8 As shown;

[0035] Operator direction variation diagram as shown Figure 9 As shown;

[0036] S6-9. Calculate the fitness function W of the new operator and compare it with the fitness function of the optimal operator. If the fitness function of the new operator is smaller, replace it; otherwise, discard it.

[0037] S6-10: Determine the number of algorithm iterations. If the number of iterations has reached the set number z, exit the iteration and output the weld seam allocation scheme. Otherwise, return to S6-8 and perform mutation and crossover operations on the seven genetic operators other than the optimal operator.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] 1. The automatic programming method for dual welding robots based on digital twins described in this invention has good economic benefits. It acquires weld seam locations through a visual method, which is simple and easy to learn. Compared with traditional manual teaching, operators using this method can easily operate the welding robot to complete tasks such as welding workpieces without professional robot technology training, resulting in a cost advantage.

[0040] 2. The automatic programming method for dual welding robots based on digital twins described in this invention offers high overall safety. Compared to traditional manual drag-and-teach methods, this invention allows workers to program the robot remotely from the work environment, avoiding toxic and harmful substances such as arc light and fumes generated during welding.

[0041] 3. The automatic programming method for dual welding robots based on digital twins described in this invention is convenient, fast, and highly accurate. Compared with traditional manual drag-and-drop teaching, the method of this invention has a shorter overall time and avoids observation errors in manual teaching, resulting in higher overall accuracy. Attached Figure Description

[0042] Figure 1 This is a system flowchart of the present invention.

[0043] Figure 2 This is a schematic diagram of the hardware system device used in this invention.

[0044] Figure 3 This is a schematic diagram of the digital twin created by importing a robot model in this invention.

[0045] Figure 4 This is a schematic diagram of the twin space display interface and robot information and function operation interface established based on the concept of digital twins in this invention.

[0046] Figure 5 This is the automatic programming function interface of the present invention.

[0047] Figure 6 This is a schematic diagram of the weld seam extracted by the present invention.

[0048] Figure 7 This is a schematic diagram of chromosomes for the genetic algorithm that characterizes the welding direction of the weld seam according to the present invention.

[0049] Figure 8 This is a schematic diagram of operator sequence variation in this invention.

[0050] Figure 9 This is a schematic diagram of the operator direction variation of the present invention.

[0051] Among them, 3.1-Robot R1, 3.2-Robot R2, 2-Binocular point cloud camera, 1-Host computer. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] The embodiments of the present invention are implemented based on the technical solution of the present invention, and detailed implementation methods and specific operation processes are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0054] Reference Figure 2 This invention provides a rapid programming method based on digital twins. The hardware device used is shown in the figure, mainly including two programmable robots (3.1 - robot R1 and 3.2 - robot R2), one binocular point cloud camera, and one host computer. Figure 3 , 4 As described in points 5 and 6, the binocular vision module first captures images of the workpiece after calibration, then converts them into point cloud files. After filtering and sampling, these files are transmitted to the host computer. Finally, the position of the weld is identified using an algorithm and converted into code readable by the robot for use in welding operations. A flowchart illustrating the specific implementation principle is shown below. Figure 1 As shown.

[0055] S1. Install a binocular point cloud camera at the end of robot R1, calibrate the positional relationship between robot R1, robot R2, and the binocular point cloud camera, and obtain the homogeneous transformation matrix M1 from robot R2 to the base coordinates of robot R1, and the homogeneous transformation matrix M2 from the binocular camera to the base coordinates of robot R1.

[0056] S2. Using the DH robot kinematics modeling method, establish the forward and inverse kinematics models of robot R1 and robot R2.

[0057] S3. Draw 3D models of robots R1 and R2, the workpiece to be welded, and related equipment; establish a digital twin programming system for the welding robots; import the drawn 3D models; and recreate the digital twin space of the welding robots.

[0058] S4. Acquire workpiece point cloud data through a binocular point cloud camera, perform filtering, resampling, segmentation and other processing on the data, extract weld point cloud information, and apply homogeneous transformation matrix M2 to transform the point cloud coordinates into coordinates in the robot's R1 base coordinate system.

[0059] S5. Perform twinning on the weld point cloud extracted in step S4, and perform B-spline curve fitting and equal-interval rediscretion on the weld to obtain discrete welding trajectory points.

[0060] S6. Perform robot welding task planning, assigning welding tasks to robots R1 and R2. Based on the robot motion code format used and the discrete welding trajectory points obtained in step S5, pre-determine the welding process parameters required for the robot motion code, and automatically generate the robot motion code.

[0061] S7. Based on the robot kinematic model established in step S2, calculate the joint axis rotation angle corresponding to the welding trajectory point, send it to the twin space for motion simulation, and observe whether the welding task requirements can be met. If not, repeat steps S4-S7.

[0062] S8. If the requirements are met, the motion code generated in step S6 is sent to the physical robot to complete the automatic programming of the welding robot's digital twin.

[0063] Step 2, importing the 3D model, requires disassembling the overall robot model into individual joint models and fixing a coordinate system at the rotation axis of each joint. The initial position of the coordinate system is set according to its external dimensions. The joint models are then imported sequentially according to the rotation order, parent-child relationships are added, and variables are bound so that the twin robot's movement can be controlled via data within the twin space.

[0064] Step S6 involves planning the dual-robot multi-welding task. First, the weld seam assignment task parameters are extracted, the total welding trajectory distance is set as the genetic fitness function, an algorithm operator is established, and then the process is iterated until the genetic algorithm converges.

[0065] The following is a detailed implementation process of the present invention:

[0066] The present invention provides an automatic programming method for robots based on digital twins, the workflow of which is as follows: Figure 1 As shown, it mainly includes twin space establishment, welding trajectory optimization and task sequencing, and automatic generation of robot motion code.

[0067] The twin space is created. Using Visual Studio software, a twin space is built on the WPF (Windows Presentation Foundation) platform based on the .NET framework. The specific steps are as follows: In the XAML code of the WPF designer, import the Presentation3D namespace and the 3D design functions and methods within it for easy use later. Use the MeshGeometry3D method to define 3D objects, the Material and Light methods to define the materials and lighting effects of the 3D model respectively, and the PerspectiveCamera method to define the overall viewing angle of the interface. Add a Viewport3D control and load the above elements into the control to display the 3D model.

[0068] Obtain 3D models of robots, workpieces, etc. If using a mainstream brand robot, these can be obtained from the brand's official website. If an accurate 3D model is unavailable, it can be created using Solidworks software based on its external dimensions. The resulting 3D model must be in .stl format. If the model format is incompatible, it can be converted using Solidworks software. Import the 3D model into twin space, adding attributes such as rotation axes and parent-child relationships, and binding the data for subsequent motion simulation.

[0069] Welding task sequencing includes the following steps:

[0070] For a single weld seam welding task, if the weld seam is only within the workspace of robot R1 or robot R2, then the weld seam is assigned to robot R1 or robot R2; if the weld seam is at the intersection of the workspaces of robot R1 and robot R2, then the intersection of the workspaces of robot R1 and robot R2 is used as the dividing point to assign the weld seams to each robot.

[0071] For multi-weld welding tasks, S2 performs optimal sorting of the two robots. The specific steps are as follows:

[0072] Extract the weld seam task assignment parameters. Constants: weld seam number sequence S, weld seam length L, robot end effector initial point P, robot running speed V, robot R... i Occupied workspace O i The number of algorithm iterations, z. Variable: The weld sequence U assigned by the robot. i :

[0073] U i ={u i1 ,u i2 ,u i3 ,u i4 ,u i5 ,......,u ij}

[0074] In the formula, U i The weld sequence assigned to the i-th robot, u ij The weld with sequence number j is assigned to the i-th robot.

[0075] The welding direction sequence D assigned by the robot to the weld seam i :

[0076] D i ={d i1 ,d i2 ,d i3 ,d i4 ,d i5 ,......,d ij}

[0077] In the formula, D i The welding direction sequence of the weld seam assigned to the i-th robot, d ij The welding direction of the weld seam with sequence number j assigned to the i-th robot. d ij ∈{0,1}, where 0 represents the welding robot welding from the start point to the end point of the weld, and 1 represents the opposite.

[0078] S3 sequentially determines the relationship between the weld and the workspace of each robot. If it belongs exclusively to robot R... i The workspace is then classified as part of the weld sequence U. i If the robots are in the same workspace, robots with odd-numbered IDs are assigned to robot R1, and robots with even-numbered IDs are assigned to robot R2. The initial robot assignment weld sequence U is generated according to this rule. If the weld position exceeds the workspace of both robots simultaneously, an error is reported.

[0079] S4 establishes a dual-robot genetic algorithm operator, the structure of which is as follows: Figure 7 As shown;

[0080] S5 randomly generates 8 sets of operators as the initial genetic population.

[0081] S6 uses the shortest total distance traveled by the two robots as the fitness function for the genetic algorithm. The total distance W can be expressed by the formula:

[0082]

[0083] In the formula, l i This represents the length of the i-th weld seam to be welded. This represents the travel distance between the i-th weld and the next weld to be welded. This represents the distance traveled without load from the initial position of the robot's end effector to the starting point of the first weld seam. This represents the distance traveled without moving from the end of the last weld seam back to the starting point of the robot's end effector.

[0084] S7 calculates the genetic fitness functions of 8 sets of operators in the genetic population and extracts the optimal operator with the smallest fitness function value.

[0085] S8 performs mutation, crossover, and other operations on the seven genetic operators other than the optimal operator.

[0086] A schematic diagram of operator sequence variation is shown below. Figure 8 As shown;

[0087] Schematic diagram of operator direction variation is shown below Figure 9 As shown;

[0088] S9 calculates the fitness function W of the new operator and compares it with the fitness function of the optimal operator. If the fitness function of the new operator is smaller, it is replaced; otherwise, it is discarded.

[0089] S10 determines the number of algorithm iterations. If the set number z has been reached, the iteration exits and the weld seam allocation scheme is output. Otherwise, it returns to S8.

[0090] The above are preferred embodiments of the present invention. Any changes made to the technical solution of the present invention that do not exceed the scope of the technical solution of the present invention shall fall within the protection scope of the present invention.

Claims

1. An automatic programming method for a dual welding robot based on digital twins, characterized in that, Includes the following steps: S1, in the robot A binocular point cloud camera is installed at the end of the robot to monitor its movement. With robots The positional relationship between the binocular point cloud cameras is calibrated to obtain the robot's positional data. To Robot Homogeneous transformation matrix of base coordinates From binocular cameras to robots Homogeneous transformation matrix of base coordinates ; S2. Using the DH robot kinematics modeling method, establish the robot... With robots Forward and inverse kinematic models; S3, Drawing Robot ,robot A digital twin programming system for welding robots is established using a 3D model of the workpiece to be welded, and the drawn 3D model is imported to recreate the digital twin space of the welding robot. S4. Acquire workpiece point cloud data using a binocular point cloud camera. Perform filtering, resampling, and segmentation on the data to extract weld point cloud information. Apply a homogeneous transformation matrix. Transforming point cloud coordinates into robot Coordinates in a base coordinate system; S5. Perform twinning on the weld point cloud extracted in step S4. The operation of twinning the weld point cloud includes fitting the weld with a B-spline curve and re-discretizing at equal intervals to obtain discrete welding trajectory points. S6. Perform robot welding task planning and assign welding tasks to robots R1 and R2. This includes automatically generating robot motion code based on the robot motion code format used and the discrete welding trajectory points obtained in step S5, pre-given welding process parameters required for the robot motion code; S7. Based on the robot forward and inverse kinematics model established in step S2, calculate the joint axis rotation angle corresponding to the welding trajectory point, send it to the digital twin space for motion simulation, and observe whether the simulation result of the motion code can meet the welding task requirements. If it cannot meet the welding task requirements, repeat steps S4-S7. S8. When the simulation results of the motion code meet the requirements of the welding task, the motion code generated in step S6 is sent to the physical robot to complete the automatic programming of the welding robot's digital twin. The task of assigning weld seams in step S6 includes assigning weld seam number sequences, establishing a dual-robot genetic algorithm operator, establishing a dual-robot genetic algorithm fitness function, and performing algorithm iteration. The assigned weld number sequence includes the following: Step S6-1: For a single weld seam welding task, if the weld seam is only within the workspace of robot R1 or robot R2, then the weld seam is assigned to robot R1 or robot R2; if the weld seam is at the intersection of the workspaces of robot R1 and robot R2, then the intersection of the workspaces of robot R1 and robot R2 is used as the dividing point to assign the respective weld seams. Step S6-2: For multi-weld welding tasks, perform optimal sorting of the two robots. Establish the weld sequence assigned by the robot : ； In the formula, For the first The welding sequence assigned by the robot For the first The serial number assigned to each robot is The weld seam; Establish the welding direction sequence of the weld seams assigned by the robot. : ； In the formula, For the first The welding direction sequence assigned to the weld seams by the robot. For the first The serial number assigned to each robot is The welding direction of the weld seam; In this context, 0 represents the welding robot welding from the start point to the end point of the weld, while 1 represents the welding robot welding from the end point to the start point of the weld. Step S6-3: Sequentially determine the relationship between the weld and the workspace of each robot. When the weld belongs exclusively to a robot... The workspace then classifies it as part of the weld sequence. When two robots are in the same workspace, robots with odd-numbered IDs are assigned to different robots. Those with even numbers are assigned to robots. The initial robot-assigned weld sequence is generated according to this rule. If the weld seam is outside the working space of both robots, an error will be reported. The establishment of the dual-robot genetic algorithm operator and the establishment of the dual-robot genetic algorithm fitness function include: Step S6-4: Establish a dual-robot genetic algorithm operator, including one chromosome representing weld seam affiliation and sorting, and one chromosome representing weld seam welding direction; Step S6-5: Randomly generate 8 sets of operators as the initial genetic population; Step S6-6: Use the shortest total distance traveled by both robots as the fitness function for the genetic algorithm; the total distance W can be expressed by the following formula: ； in, Indicates the first The length of the weld to be welded Indicates the first The distance of travel between one weld and the next weld to be welded. This represents the distance traveled by the robot from its initial position at the end effector to the starting point of the first weld seam. This represents the distance traveled without moving from the end of the last weld seam back to the starting point of the robot's end effector.

2. The automatic programming method for a dual welding robot based on digital twins according to claim 1, characterized in that, In step S3, the robot is imported. ,robot Before importing the 3D model of the robot, it is necessary to import the robot... 3D models and robots The 3D model is broken down into individual joint models and in the robot A coordinate system is fixed at the rotation axis of each joint of the robot. A coordinate system is fixed at the rotation axis of each joint.

3. The automatic programming method for a dual welding robot based on digital twins according to claim 2, characterized in that, In step S3, according to the robot The initial position of the coordinate system fixed at its rotation axis is set according to the robot's external dimensions. The initial position of the coordinate system fixed at its rotation axis is set according to the external dimensions; further, the robot is sequentially imported according to the rotation sequence. Joint models and robots The joint model; further, adding the robot Establish parent-child relationships for joint axes and bind angle variables; add the robot. Establish the parent-child relationship of the joint axis and bind the angle variable.

4. The automatic programming method for a dual welding robot based on digital twins according to claim 1, characterized in that, The establishment of the dual-robot genetic algorithm operator and the dual-robot genetic algorithm fitness function also includes the following: Step S6-7: Calculate the fitness functions of the 8 sets of operators in the genetic population, and extract the optimal operator with the smallest fitness function value; Step S6-8: Perform mutation and crossover operations on the 7 groups of operators other than the optimal operator.

5. The automatic programming method for a dual welding robot based on digital twins according to claim 4, characterized in that, The iterative algorithm for the fitness function of the dual-robot genetic algorithm includes the following: Step S6-9: Calculate the fitness function of the new operator and determine the size of the fitness function of the new operator compared with that of the optimal operator; if the fitness function of the new operator is smaller than that of the optimal operator, then perform the replacement; if the fitness function of the new operator is greater than that of the optimal operator, then abandon the replacement.

6. The automatic programming method for a dual welding robot based on digital twins according to claim 5, characterized in that, The iterative algorithm for the fitness function of the dual-robot genetic algorithm includes the following: Step S6-10: Determine the number of algorithm iterations; When the algorithm iteration count has reached the set number If the iteration fails, the iteration ends and the weld seam task allocation scheme is output. Conversely, the seven sets of operators other than the optimal operator are subjected to mutation and crossover operations again.

Images