Intelligent welding multi-robot workstation for C-shaped frame type bridge rib plate and working method

Abstract

The invention relates to a C-shaped frame type bridge rib plate intelligent welding multi-robot workstation and a working method, a plurality of inverted C-shaped beams walking on a workstation ground rail are respectively provided with a workstation control cabinet, a welding power supply and a gun cleaning station in a matched mode, a 3D large scanning device is installed on the front side of an extending arm at the upper end of each C-shaped beam, and a 3D small camera is installed on the lower portion of the front portion of the extending arm at the upper end of each C-shaped beam. A welding robot body is installed on the lower portion of the C-shaped beam. The device has the advantages that manual teaching, a tool clamp and drawing introduction can be avoided, flexible welding is achieved, the size of a covered workpiece is large, the site is saved through robot collaborative welding, and the requirements for small batches and multiple types of bridge rib plates are effectively met.

Description

Technical Field

[0001] The present invention relates to a C-type frame bridge rib intelligent welding multi-robot workstation and working method which can avoid manual teaching, fixtures and drawing import, realize flexible welding and cover large workpiece sizes, save space through robot collaborative welding, and effectively solve the requirements of small batches and multiple varieties of bridge ribs. Background Art

[0002] With the development and progress of the world, infrastructure such as building construction, bridge construction, shipbuilding, and railway tracks are becoming more and more important to today's society. These infrastructures have led to a growing demand for steel structures, and the processing of steel structures has increased day by day. However, due to the harsh working environment, heavy workload, and high risk factor of discrete steel structure processing and welding, it has become an inevitable trend to introduce robot technology into discrete steel structure processing and welding to replace manual operations. However, the characteristics of discrete steel structure components are: low repetition rate, many non-standard parts, complex node forms, and large plate thickness, which not only increase the difficulty and complexity of processing operations, but also make the quality of processing operations using traditional robots unguaranteed. Summary of the invention

[0003] Design purpose: To avoid the shortcomings of the background technology, design a C-type frame bridge rib intelligent welding multi-robot workstation and working method that can avoid manual teaching, fixtures, and drawing import, and can achieve flexible welding and cover large workpiece sizes. At the same time, it can save space through robot collaborative welding, and effectively solve the small batch and multi-variety requirements of bridge ribs.

[0004] Design plan: Intelligent manufacturing has become an emerging field that countries around the world are vigorously developing. Applying intelligent robot technology to discrete steel structure processing has become a feasible solution to the problem of difficult quality assurance in traditional discrete steel structure processing operations. However, the current discrete workpiece processing and welding in the pan-steel structure industry faces the following three challenges. (1) Intelligent robots have difficulty in perceiving and sensing complex environments. The processing environment of discrete workpieces in the pan-steel structure industry is highly unstructured and complex and changeable. Traditional deep learning models are often trained on specific data sets for specific perception and understanding tasks, and have poor generalization, making it difficult for robots to accurately perceive and model complex and changeable processing environments. (2) The current RGB cameras and depth cameras are inefficient in collecting data about the surrounding environment, and efficient data collection equipment and sensor equipment are urgently needed. (3) Workpieces in the bridge industry generally present "small batches and multiple types". Many batches of workpieces do not have model drawings, or the model drawings are not supported or universal drawings on the market. When applying similar offline programming, it is inconvenient to import the system, resulting in low efficiency.

[0005] In order to solve the above problems, the patent of the present invention provides the following method of intelligent welding multi-robot workstation for bridge ribs based on multimodal data model-free, which effectively solves the above difficulties and the required technical improvement requirements. In terms of technical solution design, this workstation includes a visual system, a C-frame welding robot system and a control system, wherein the C-frame is 5 meters high and 4 meters wide, each with a six-axis industrial robot arm, and the ground rail is designed to be 20 meters long. The visual system includes a large-range line laser sensor, and the structured light sensor is installed on the six-axis robot arm. The workstation control system is used to control the line laser sensor and the structured light sensor to take pictures to obtain the bridge rib workpiece information and its weld position and posture information, and to control the movement of the C-frame and the welding robot, the distribution of the welding path, and the collaborative obstacle avoidance of multiple robots through algorithm settlement.

[0006] Compared with the background technology, the present invention can eliminate the need for manual teaching, fixtures, and drawing import, and can flexibly weld and cover large workpiece sizes. Robot-assisted welding saves space and effectively solves the requirements of small batches and multiple varieties of bridge ribs. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of a multi-robot workstation for intelligent welding of C-type bridge ribs. DETAILED DESCRIPTION

[0008] Embodiment 1: refer to the attached Figure 1 A C-frame bridge rib intelligent welding multi-robot workstation, multiple inverted C-beams 1 walking on the workstation ground rail 7 are equipped with a workstation control cabinet 5, a welding power supply 6 and a gun cleaning station 2, a 3D large scanning device 3 is installed on the front side of the upper end of the extended arm of the C-beam 1, a 3D small camera 8 is installed on the lower front of the upper end of the extended arm of the C-beam 1, and a welding robot body 4 is installed on the lower part of the C-beam 1. The height of the C-frame is 4.5 to 5.5 meters, preferably 5 meters; the width is 3.6 to 4.5 meters, preferably 4 meters; the design length of the ground rail is 15 to 25 meters, preferably 20 meters. Each inverted six-axis industrial robot arm.

[0009] The visual system (3, 8) includes a large-range line laser sensor and a structured light sensor installed on a six-axis robotic arm. The workstation control cabinet uses a Windows 10 i7 high-performance PC, which can realize the functions of the host computer and process the point cloud algorithm of the line laser and structured light sensors. It is used to control the line laser sensor and structured light sensor to take pictures to obtain the bridge rib workpiece information and its weld position and posture information, and control the movement of the C-frame and the welding robot body 4 and the distribution of the welding path through algorithm settlement and the coordinated obstacle avoidance of multiple welding robot bodies.

[0010] The welding power source 6, the gun cleaning station 2, the welding robot body 4 and the six-axis industrial robot arm are existing technologies and can be purchased directly.

[0011] The workstation control cabinet is replaced by a Windows 10 i7 high-performance PC, which can realize the functions of the host computer and process the point cloud algorithms of the line laser and structured light sensors.

[0012] Embodiment 2: Based on Embodiment 1, a working method of a C-frame bridge rib intelligent welding multi-robot workstation is provided. Step 1: First, the robot and structured light need to be calibrated, and then the robot hand-eye calibration is performed. Since the structured light camera carried by the robot is installed in front of the robot welding gun, it belongs to the eye-in-hand solution. The robot is driven to shoot a fixed chessboard at 5 positions to obtain 5 sets of robot posture matrices, and then the corresponding 5 sets of chessboard feature points are extracted by Zhang's method. The hand-eye matrix is ​​calculated by AX=XB. Through the hand-eye matrix and the corresponding robot posture matrix, the structured light point cloud camera coordinate system can be converted to the robot base coordinate system. Then the line laser and robot are calibrated. Place the large standard fixture under the line laser sensor, shoot the point cloud, and select 3 point cloud coordinates on the fixture point cloud. This is the coordinate in the camera coordinate system. Then drive the end of the robot to touch the actual position of the fixture. This is the position corresponding to the above three points in the robot coordinate system. Then, the two groups of three points are cross-multiplied to obtain the conversion relationship of the coordinate system. The line laser and robot can complete the calibration.

[0013] Step 2: The bridge rib workpiece to be welded is hoisted to the C-frame welding station by a gantry crane and randomly placed. The ground rail is controlled by the workstation control unit to move from the zero point to the end point. Since the line laser sensor is installed above the C-frame, the line laser sensor can obtain the overall scene point cloud below the C-frame through this movement. The control unit transmits the single-frame line laser data to the point cloud processing unit. The single-frame point cloud extracts the information of the line laser height and brightness change through the grayscale centroid method to obtain the scene point cloud.

[0014] Step 3: Since line laser calibration + robot hand-eye calibration have been done in step 1, the scene points captured at this time are already point clouds in the world coordinate system. First, perform straight-through filtering on the scene point cloud to filter out the point cloud on the ground, and then perform the Euclidean clustering segmentation algorithm for the point cloud. Then perform plane fitting on the point cloud using the RANSAC algorithm. The principle of the RANSAC algorithm is to randomly extract three points from the overall point cloud each time, fit a plane with the three points according to formula 1, and then calculate the distance from all the remaining point clouds to this plane. According to the set threshold, the point clouds that meet the requirements are considered to be point clouds that satisfy the plane equation. Repeatedly select randomly extracted points. When the number of points that satisfy a certain plane equation is the largest, it can be used as the final plane equation: Principle of the Ransac algorithm: randomly select 3 points, calculate the sampling plane formed by the 3 points, calculate the distance from all points to the plane based on the sampling plane, and divide the points into inner points and outer points according to the parameters (threshold of the distance). Count the number of inner points and outer points, and update the maximum number of iterations. K=(log*(1-p)) / log(1-w^n) (1) Repeat the above process until the stopping condition (maximum number of iterations, internal point ratio, etc.) is reached. After plane equation extraction, the workpiece can be simply divided into the base plate and other features.

[0015] (2) Step 4: Filter out the bottom plate point cloud, which is the interference point cloud of the welding platform or the frame, for the above segmented point cloud. The remaining point cloud is the workpiece point cloud. The structured light 3D vision sensor installed at the end of each six-axis robot arm takes the local point cloud of the weld to complete the line laser point cloud: Since the line laser point cloud obtained in step 2 is a point cloud in the world coordinate system, extract the bounding box of the workpiece point cloud: Minimum bounding box calculation process: 1. Use PCA principal component analysis to obtain the three main directions of the point cloud, obtain the centroid, calculate the covariance, obtain the covariance matrix, and find the eigenvalue and eigenvector of the covariance matrix. The eigenvector is the main direction; 2. According to the main direction and center of mass obtained in 1, the input point cloud is transformed to the origin, and the main direction coincides with the direction of the coordinate system, and the point cloud bounding box transformed to the origin is established.

[0016] 3. Set the main direction and bounding box for the input point cloud, which is achieved by the inverse transformation of the input point cloud to the origin point cloud.

[0017] According to the size of the bounding box, the shooting points of the structured light camera carried by the robotic arm are automatically planned, and the structured light camera is driven to capture point clouds at different angles around the workpiece multiple times. Since the hand-eye calibration has been done in step 1, the complete workpiece point cloud can be obtained by transferring the captured point cloud to the world coordinate system.

[0018] Step 5: Combine the complete point cloud information of the workpiece with the RANSAC algorithm of the point cloud to perform plane fitting and extract the intersection points of the three planes. Solve the linear equations: Assume that the normal vectors of the three planes are (a1, b1, c1) ( a 1, b 1, c 1), (a2,b2,c2)( a 2, b 2, c 2), (a3,b3,c3)( a 3, b 3, c 3), then each plane can be expressed as ax+by+cz=d ax + by + cz = d These equations are combined to form a linear system: a 1 +b 1 y+c 1 z=d 1 a 2 +b 2 y+c 2 z=d 2 a 3 +b 3 y+c 3 z=d 3 By solving this system of linear equations, we can get the coordinates (x, y, z) of the intersection of the three planes. Plane 1 a*x + b*y + c*z + d == 0; Plane2 e*x + f*y + g*z + h == 0; Plane equation: ax+by+cz+d=0 Two plane parameters: p1=(a1,b1,c1,d1), p2=(a2,b2,c2,d2) Direction vector of the intersection line: n=n1×n2=(b1c2-c1b2,c1a2-a1c2,a1b2-b1a2) Let any point on the line be (x,y,z) Let x = 0, we get: b1y+c1z+d1=0, b2y+c2z+d2=0, Right now y=-(c1z+d1) / b1=-(c2z+d2) / b2 The solution is: z=-(d1 / b1-d2 / b2) / (c1 / b1-c2 / b2) y=-c1 / b1z-d1 / b1 =c1*(d1 / b1-d2 / b2) / b1 / (c1 / b1-c2 / b2)-d1 / b1 The straight line can be determined by the straight line direction vector and the coordinates of a point Point-form equation of a line: A(x−x0)+B(y−y0)+C(z−z0)=0 Among them, (x0, y0, z0) = (0, x, y), (A, B, C) = n = (b1c2-c1b2, c1a2-a1c2, a1b2-b1a2) The obtained intersection line is the weld position of the bridge rib workpiece point cloud according to the arc instructions and straight line instructions that the robot can execute, and the robot can be driven to weld.

[0019] The above method uses point method to obtain plane equations and intersection lines between faces.

[0020] It should be understood that although the above embodiments provide a relatively detailed textual description of the design ideas of the present invention, these textual descriptions are only simple textual descriptions of the design ideas of the present invention, rather than limitations on the design ideas of the present invention. Any combination, addition or modification that does not exceed the design ideas of the present invention shall fall within the protection scope of the present invention.

Claims

1. A C-type bridge rib intelligent welding multi-robot workstation, characterized by: walking A plurality of inverted C-shaped beams (1) are mounted on a workstation floor rail (7) and are respectively equipped with a workstation control cabinet (5), a welding power source (6) and a gun cleaning station (2); a 3D large scanning device (3) is mounted on the front side of an extending arm at the upper end of the C-shaped beam 1; a 3D small camera (8) is mounted on the lower front side of the extending arm at the upper end of the C-shaped beam 1; and a welding robot body (4) is mounted on the lower part of the C-shaped beam (1).

2. The C-frame bridge rib intelligent welding multi-robot workstation according to claim 1 is characterized by: The C-frame is 4.5 to 5.5 meters high and 3.6 to 4.5 meters wide, and each frame is inverted with a six-axis industrial robot arm.

3. The C-frame bridge rib intelligent welding multi-robot workstation according to claim 1 is characterized by: The designed length of the ground rail is 15 to 25 meters.

4. The C-frame bridge rib intelligent welding multi-robot workstation according to claim 1 is characterized by: The visual system (3, 8) includes a large-range line laser sensor and a structured light sensor installed on a six-axis robotic arm. The workstation control cabinet (5) is used to control the line laser sensor and the structured light sensor to take pictures to obtain the bridge rib workpiece information and its weld position and posture information, and to control the movement of the C-frame and the welding robot body (4) and the distribution of the welding path through algorithm settlement and the coordinated obstacle avoidance of multiple welding robot bodies.

5. A working method of a C-frame bridge rib intelligent welding multi-robot workstation, characterized by: Step 1: 1) Calibrate the robot and structured light, and then perform robot hand-eye calibration. The camera is installed in front of the robot welding gun. Drive the robot to shoot a fixed chessboard at 5 positions to obtain 5 sets of robot posture matrices. Then use Zhang's method to extract the corresponding 5 sets of chessboard feature points. Calculate the hand-eye matrix through AX=XB. Through the hand-eye matrix and the corresponding robot posture matrix, the structured light point cloud camera coordinate system can be converted to the robot base coordinate system; 2) Perform line laser and robot calibration. Place a large standard fixture under the line laser sensor, shoot the point cloud, and select 3 point cloud coordinates on the fixture point cloud as the coordinates in the camera coordinate system. Then drive the end of the robot to touch the actual position of the fixture. This is the position corresponding to the above three points in the robot coordinate system. Then, the two groups of three points are cross-multiplied to obtain the conversion relationship of the coordinate system. The line laser and robot can complete the calibration; Step 2: The bridge rib workpiece to be welded is hoisted to the C-frame welding station by a gantry car and randomly placed. The ground rail is controlled by the workstation control unit to move from the zero point to the end point. Since the line laser sensor is installed above the C-frame, the line laser sensor can obtain the overall scene point cloud below the C-frame through this movement. The control unit transmits the single-frame line laser data to the point cloud processing unit. The single-frame point cloud extracts the information of the line laser height and brightness change through the grayscale centroid method to obtain the scene point cloud; Step 3: Since the line laser calibration + robot hand-eye calibration has been done in step 1, the scene points captured at this time are already point clouds in the world coordinate system. First, the scene point cloud is filtered directly to filter out the point cloud on the ground. Then the Euclidean clustering segmentation algorithm of the point cloud is performed, and then the RANSAC algorithm of the point cloud is used for plane fitting. The principle of the RANSAC algorithm is roughly to randomly extract three points from the overall point cloud each time. According to formula 1, the three points fit a plane, and then all the remaining point clouds are calculated to the distance of this plane. According to the set threshold, the point clouds that meet the requirements are screened and considered to be the point clouds that satisfy the plane equation. The randomly extracted points are repeatedly selected. When the number of points that satisfy a certain plane equation is the largest, it can be used as the final plane equation: Ransac algorithm principle: randomly select 3 points, and calculate the sampling plane formed by the 3 points, calculate the distance from all points to the plane according to the sampling plane, and divide the points into inner points and outer points according to the parameter (distance threshold). Count the number of inner points and outer points, and update the maximum number of iterations K=(log*(1-p)) / log(1-w^n) (1) Repeat the above process until the stop condition (maximum number of iterations, internal point ratio, etc.) is reached. After plane equation extraction, the workpiece can be simply divided into the base plate and other features. （2） Step 4: For the above segmented point cloud, filter out the base plate point cloud, which is the interference point cloud of the welding platform or the frame, and the remaining point cloud is the workpiece point cloud. The structured light 3D vision sensor installed at the end of each six-axis robot arm takes the local point cloud of the weld to complete the line laser point cloud: Since the line laser point cloud obtained in step 2 is a point cloud in the world coordinate system, extract the bounding box of the workpiece point cloud: Minimum bounding box calculation process: ①Use PCA principal component analysis to obtain the three main directions of the point cloud, obtain the centroid, calculate the covariance, obtain the covariance matrix, and find the eigenvalue and eigenvector of the covariance matrix. The eigenvector is the main direction; ② According to the main direction and center of mass obtained in ① above, the input point cloud is transformed to the origin, and the main direction coincides with the direction of the coordinate system, and a bounding box of the point cloud transformed to the origin is established; ③ Set the main direction and bounding box for the input point cloud, which is achieved by the inverse transformation of the input point cloud to the origin point cloud; ④ According to the size of the bounding box, the shooting points of the structured light camera carried by the robot arm are automatically planned, and the structured light camera is driven to take point clouds at different angles around the workpiece multiple times. Since the hand-eye calibration has been done in step 1, the point cloud taken can be converted to the world coordinate system to obtain the complete workpiece point cloud; Step 5: Combine the complete point cloud information of the workpiece with the RANSAC algorithm of the point cloud to perform plane fitting and extract the intersection points of the three planes. Solve the linear equations: Assume that the normal vectors of the three planes are (a1, b1, c1) ( a 1, b 1, c 1), (a2,b2,c2)( a 2, b 2, c 2), (a3,b3,c3)( a 3, b 3, c 3), then each plane can be expressed as ax+by+cz=d ax + by + cz = d Combining these equations into a linear system of equations: a1+b1y+c1z=d1 a2+b2y+c2z=d2 a3+b3y+c3z=d3 By solving this set of linear equations, we can obtain the coordinates of the intersection of the three planes (x, y, z), the intersection of the two planes, Plane 1 a*x + b*y + c*z + d == 0; Plane2 e*x + f*y + g*z + h == 0; Plane equation: ax+by+cz+d=0 Two plane parameters: p1=(a1,b1,c1,d1), p2=(a2,b2,c2,d2) Direction vector of the intersection line: n=n1×n2=(b1c2-c1b2,c1a2-a1c2,a1b2-b1a2) Let any point on the line be (x,y,z) Let x = 0, we get: b1y+c1z+d1=0, b2y+c2z+d2=0, Right now y=-(c1z+d1) / b1=-(c2z+d2) / b2 The solution is: z=-(d1 / b1-d2 / b2) / (c1 / b1-c2 / b2) y=-c1 / b1z-d1 / b1 =c1*(d1 / b1-d2 / b2) / b1 / (c1 / b1-c2 / b2)-d1 / b1 The straight line can be determined by the straight line direction vector and the coordinates of a point Point-form equation of a line: A(x−x0)+B(y−y0)+C(z−z0)=0 Among them, (x0, y0, z0) = (0, x, y), (A, B, C) = n = (b1c2-c1b2, c1a2-a1c2, a1b2-b1a2) The obtained intersection line is the weld position of the bridge rib workpiece point cloud according to the arc instructions and straight line instructions that the robot can execute, and the robot can be driven to weld.

Images