Visual guidance wind power blade cutting method
Through the vision-guided wind turbine blade cutting method, using 3D camera scanning and robot automatic cutting, the accuracy and efficiency problems of wind turbine blade cutting in the existing technology are solved, and high-precision, low-cost automated cutting is achieved.
Patent Information
- Application Number
- CN202510724997.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-03
- Publication Date
- 2025-10-10
AI Technical Summary
In the existing technology, wind turbine blade cutting has high risks and low precision in manual operation, and traditional mechanical equipment is inconvenient to position, resulting in inflexible cutting operations and difficulty in achieving efficient and accurate separation of composite materials.
A vision-guided wind turbine blade cutting method is adopted. Standard and actual point clouds are acquired through 3D camera scanning, pre-processed and registered, cutting points are marked, cutting trajectories are fitted, and robots are used for automatic cutting, which reduces manual intervention and improves positioning accuracy and cutting efficiency.
It achieves high-precision cutting without human intervention, reduces costs, is suitable for a variety of complex environments, improves cutting accuracy and efficiency, and is suitable for a variety of complex working environments. The process of obtaining the cutting trajectory effectively segments the blade point cloud and the ground point cloud, thereby improving the accuracy of the cutting trajectory.
Smart Images
Figure CN120765533A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the field of wind turbine blade cutting, and particularly relates to a visual guidance wind turbine blade cutting method. BACKGROUND
[0002] With the gradual depletion of traditional fossil energy such as coal, oil and natural gas, wind energy as a clean and renewable energy has been widely used. As the core energy conversion component of a wind turbine generator, the production, retirement and recycling of a wind turbine blade have become a key link in the whole life cycle management.
[0003] A wind turbine blade is mainly composed of glass fiber reinforced epoxy resin composite material and has a service life of about 20-25 years. According to the prediction of the European Wind Energy Association, more than 50,000 tons of composite material blades will face retirement by 2030. The traditional landfill or incineration treatment method has environmental risks such as soil pollution and carbon emissions, while emerging technologies such as mechanical recycling and pyrolysis recycling require fine cutting as a prerequisite. However, the blade has the characteristics of super large size, special curved surface structure and internal multi-cavity reinforcing rib, and the traditional cutting technology faces many difficulties.
[0004] At present, in the prior art, for wind turbine blade cutting, most of them are manually cut or use a linear track to convey the wind turbine blade and then cut it by a saw blade. Manual operation is risky and has low precision, while cutting the wind turbine blade on a linear track requires high positioning accuracy of the blade, and additional mechanical equipment is needed to convey the wind turbine blade, resulting in additional costs. The equipment is not convenient to control in the azimuth movement, making the cutting operation not flexible enough. Therefore, it is necessary to develop a mobile cutting equipment with environmental adaptability and precise separation of composite material structure, which has important strategic significance for the industrialization and popularization of blade recycling. SUMMARY
[0005] The purpose of the present application is to provide a visual guidance wind turbine blade cutting method, which can effectively improve the cutting precision and cutting efficiency.
[0006] The present application is realized by the following technical scheme: The visual guidance wind turbine blade cutting method comprises the following steps: Step S1, in a specific place, a 3D camera is used to perform multi-angle scanning on a wind turbine blade as a template to obtain a standard point cloud, and the standard point cloud is preprocessed to remove ground point clouds, thereby obtaining an initial template point cloud containing only blade point clouds; Step S2, uniformly down-sampling the initial template point cloud to obtain a template point cloud, and in the down-sampling process, different sampling intervals are determined according to the relationship between the average Gaussian curvature of the initial template point cloud and the principal curvatures of each point cloud; Step S3: driving a transport truck equipped with a cutting robot to the location of the wind turbine blade to be cut, moving the wind turbine blade to be segmented to a location where the robot can cut it, scanning the wind turbine blade to be segmented using a 3D camera to obtain an actual point cloud, and preprocessing the actual point cloud to remove the ground point cloud and downsample it to obtain a processed point cloud containing only the point cloud of the blade to be cut; Step S4: Mark three initial cutting points in the template point cloud as required, align the template point cloud with the point cloud to be processed to obtain a rotation and translation matrix, and obtain three actual cutting points in the point cloud to be processed corresponding to the three initial cutting points according to the rotation and translation matrix; Step S5: Fit a trajectory circle based on the three actual cutting points, and the points on the trajectory circle constitute the blade cutting trajectory. For any point on the blade cutting trajectory, an orthogonal coordinate system is established with this point to provide posture information for the robot. The robot cuts the wind turbine blade to be cut according to the cutting trajectory and posture information, and after cutting, it is transported to the designated location by a transport truck.
[0007] Furthermore, in step S1, the wind turbine blade is shifted multiple times at a set rotation angle, and a 3D camera is used to scan after each rotation, and the data obtained after each scan are spliced to obtain a standard point cloud.
[0008] Furthermore, in step S1, preprocessing the standard point cloud includes: Step S11: traverse all point clouds in the standard point cloud. i Point Cloud p i =( x i ,y i , z i )of z i If the value is greater than the set height threshold, the point cloud is attributed to the leaf point cloud set; otherwise, the point cloud is attributed to the ground set G; Step S12: Select the ground set G z The smallest coordinate k Point clouds form a seed point cloud set. For each seed point cloud s j , if the point cloud within the spherical neighborhood of the set radius p j The angle between the normal vector of the seed point cloud and the seed point cloud is not greater than the set angle threshold, then the seed point cloud is moved into the ground set G. When there is no point cloud in the seed point cloud set and it is moved into the ground set G, the final leaf point cloud set Bc is obtained, where point p j Belongs to standard point cloud; Step S13: Using a density-based noise spatial clustering algorithm to cluster the final leaf point cloud set Bc to obtain an initial template point cloud.
[0009] Furthermore, in step S2, the average Gaussian curvature of the initial template point cloud is With principal curvature The relationship between The point cloud is divided into the first type of point cloud, The point cloud is divided into the second type of point cloud, The point cloud is divided into the third category of point cloud, and the sampling interval ratio of the first category of point cloud, the second category of point cloud and the third category of point cloud is set to 3:2:1, where , n represents the number of point clouds in the initial template point cloud, Represents the point cloud in the initial template point cloud p i' The principal curvature of .
[0010] Furthermore, in step S4, the template point cloud and the point cloud to be processed are registered using the RANSCA+ICP algorithm to obtain the rotation and translation matrix T. m The actual cutting point is expressed as ,in, m =1,2,3.
[0011] Furthermore, the step S5 specifically includes the following steps: Step S51: Fit a trajectory circle based on the three actual cutting points. The points on the trajectory circle form the blade cutting trajectory. The coordinates of each point on the blade cutting trajectory are calculated as follows: the center, radius, and normal vector of the trajectory circle are calculated. The normal vector is normalized within the trajectory circle to calculate two unit vectors with the center as the origin and perpendicular to each other. The coordinates of any point on the trajectory circle can then be calculated. Step S52: For the points on the cutting trajectory q , with a point q Three orthogonal unit vectors are established for the origin. The rotation matrix of the robot posture is obtained from the three orthogonal unit vectors and the normal vector of the trajectory circle. The rotation matrix is converted into Euler angles to obtain the robot posture information.
[0012] Furthermore, the pre-processing of step S1 further includes filtering the standard point cloud using voxel filtering, and then proceeding to step S11 after the filtering.
[0013] Furthermore, in step S3, before scanning the wind turbine blade to be segmented, a correspondence between the camera coordinate system and the robot coordinate system is established through hand-eye calibration.
[0014] Furthermore, the transport truck is provided with a cutting system, which includes a truck crane, a robot, a high-pressure water jet head and a water receiving tank. The lower end of the truck crane is fixedly set on the transport truck, the robot is horizontally slidably set on the transport truck, and the high-pressure water jet head is set at the end of the robot wrist.
[0015] Furthermore, a control room is also provided on the transport truck, which includes a water tank, a water softener, a sand supply machine and a high-pressure water pump. The high-pressure water pump is connected to the water tank, the water softener is connected between the high-pressure water pump and the high-pressure water jet head to soften hard water, and the sand supply machine is connected to the sand storage container of the high-pressure water jet head.
[0016] The present invention has the following beneficial effects: 1. The present invention first uses a 3D camera to perform multi-angle scanning on the wind turbine blade as a template in a specific scene to obtain a standard point cloud, pre-processes the standard point cloud to remove the ground point cloud, and obtains an initial template point cloud containing only the blade point cloud. Then, the initial template point cloud is uniformly down-sampled to obtain a template point cloud, and then a transport truck equipped with a cutting system is driven to the position of the wind turbine blade to be cut, and the wind turbine blade to be segmented is transported to a position where the cutting system can cut. The wind turbine blade to be segmented is scanned by a 3D camera to obtain an actual point cloud, and the actual point cloud is pre-processed to remove the ground point cloud and down-sampled to obtain a processed point cloud containing only the blade point cloud to be cut. Three initial cutting points are marked in the template point cloud according to requirements, and the template point cloud is aligned with the point cloud to be processed to obtain a rotation and translation matrix. According to the rotation and translation matrix, the points in the point cloud to be processed corresponding to the three initial cutting points are obtained. The three actual cutting points of the cutting point are finally fitted with a trajectory circle based on the three actual cutting points. The blade cutting trajectory is composed of points on the trajectory circle. For any point on the blade cutting trajectory, an orthogonal coordinate system is established with the point to provide posture information for the robot. The robot cuts the wind turbine blade to be cut according to the cutting trajectory and posture information, and is transported to the designated location by a transport truck after cutting. No human participation is required in this process, which effectively reduces human risks. The transport truck drives to the position of the wind turbine blade to be cut for cutting, which reduces the positioning accuracy requirements of the wind turbine blade and does not require the use of additional mechanical transmission machines. The robot uses visual guidance to cut according to the cutting trajectory, which can effectively improve the cutting accuracy and efficiency, greatly save costs, and is suitable for a variety of complex working environments. The process of obtaining the cutting trajectory effectively segments the blade point cloud and the ground point cloud, thereby improving the accuracy of the cutting trajectory. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] The present invention will be further described in detail below with reference to the accompanying drawings.
[0018] Figure 1 It is a flow chart of the present invention.
[0019] Figure 2Schematic diagram of a transport truck of the present invention.
[0020] Figure 3 It is a schematic structural diagram of the cutting system and control room of the present invention.
[0021] Figure 4 This is a schematic diagram of cutting a wind turbine blade according to the present invention. DETAILED DESCRIPTION
[0022] like Figures 1 to 4 As shown, the vision-guided wind turbine blade cutting method includes the following steps: Step S1: In a specific location, a wind turbine blade serving as a template is scanned at multiple angles using a 3D camera to obtain a standard point cloud, and the standard point cloud is preprocessed to remove the ground point cloud to obtain an initial template point cloud containing only the blade point cloud; The specific location can be a laboratory or a standard production workshop for wind turbine blades. The specific process of obtaining the standard point cloud is as follows: the wind turbine blade is shifted multiple times at a set rotation angle (45°), scanned with a 3D camera after each rotation, and the data obtained from each scan is spliced to obtain a standard point cloud. The specific splicing process is based on existing technology. The collected standard point cloud includes blade point cloud and ground point cloud, so the standard point cloud needs to be preprocessed, including: Step S10: filtering the standard point cloud using voxel filtering; Step S11: Create a leaf point cloud set and a ground point cloud set G and initialize them, traverse all point clouds in the filtered standard point cloud, and if i Point Cloud p i =( x i ,y i , z i )of z i The value is greater than the set height threshold h th , then the point cloud is attributed to the leaf point cloud set, otherwise, the point cloud is attributed to the ground set G; Step S12: Select the ground set G z The smallest coordinate k point clouds form a seed point cloud set (in actual operation, all point clouds in the ground set G can be z Sort the coordinates in ascending order, select the first k point clouds form a seed point cloud set), for each seed point cloud s j , if the point cloud within the spherical neighborhood of the set radius pj the seed point cloud is moved into the ground set G, and the above process is repeated, until no point cloud in the seed point cloud set is moved into the ground set G, to obtain a final leaf point cloud set Bc, wherein the points p j belong to the standard point cloud; More specifically, for each seed point cloud s j ( x sj ,y si , z sj ), a covariance matrix is constructed , eigenvalue decomposition is performed on the covariance matrix, and the eigenvector corresponding to the minimum eigenvalue is the normal vector of the seed point cloud s j The normal vector between the point cloud p j and the seed point cloud s j is represented as , wherein n j is the normal vector of the point cloud p j , and the normal vector solving process is the same as that of the seed point cloud s j , except that the neighborhood of the point cloud p j is different from that of the seed point cloud s j , wherein k 1 is the number of point clouds in the neighborhood , and is the center point of the neighborhood . Step S13, the final leaf point cloud set Bc is clustered by using a base density noise application spatial clustering algorithm to obtain an initial template point cloud, wherein the base density noise application spatial clustering algorithm is a prior art.
[0023] Step S2, uniform down-sampling is performed on the initial template point cloud to obtain a template point cloud, and in the down-sampling process, different sampling intervals are determined according to the relationship between the average Gaussian curvature of the initial template point cloud and the principal curvatures of each point cloud; the point clouds in the initial template point cloud whose relationship between the average Gaussian curvature and the principal curvature meets are classified into the first type of point clouds, the point clouds whose relationship meets are classified into the second type of point clouds, and the point clouds whose relationship meets The point cloud is divided into the third category of point clouds. The sampling interval ratio of the first category of point clouds, the second category of point clouds and the third category of point clouds is set to 3:2:1, so as to achieve adaptive curvature downsampling and enhance adaptability. , n represents the number of point clouds in the initial template point cloud, Represents the point cloud in the initial template point cloud p i' The principal curvature of .
[0024] Get point cloud p i' The specific process of the principal curvature is as follows: p i' , search for the nearest neighbor through KD tree K points to form a neighborhood, traverse all point clouds in the neighborhood, if the normal vector angle between two adjacent point clouds is greater than 90°, one of the normal vectors is reversed to ensure local consistency; p i' The normal vector is Z Axis, build local coordinate system ( X , Y , Z ), projecting the point cloud in the neighborhood to the local coordinate system XY Plane, based on Euler's formula, by least squares method Fitting to obtain the maximum principal curvature K 1. Minimum principal curvature K 2 and the main direction angle , principal curvature K i' = K 1 K 2.
[0025] Step S3: driving a transport truck equipped with a cutting robot to the location of the wind turbine blade to be cut, moving the wind turbine blade to be segmented to a location where the robot can cut it, scanning the wind turbine blade to be segmented using a 3D camera to obtain an actual point cloud, and preprocessing the actual point cloud to remove the ground point cloud and downsample it to obtain a processed point cloud containing only the point cloud of the blade to be cut; The transport truck is provided with a control room, a visual system and a cutting system, the control room comprises a water tank, a water softener, a sand supply machine, a high-pressure water pump and an electric control cabinet, the visual system comprises a 3D camera, and the cutting system comprises a truck-mounted crane, a robot, a high-pressure water jet head and a water receiving pool; the water tank provides water source for high-pressure water cutting, waste water is discharged to the site, the high-pressure water pump is connected with the water tank, the water softener is connected between the high-pressure water pump and the high-pressure water jet head to soften hard water, so as to prevent carbonate from causing fouling and affecting the use stability and service life of equipment and parts, the sand supply machine is connected with a sand storage container of the high-pressure water jet head, sand (abrasive) is delivered into the sand storage container by using air pressure sand delivery mode, sand enters the high-pressure water jet head under the action of high-pressure water negative pressure through a sand delivery pipe, and the purpose of continuously supplying sand during cutting is achieved. The truck-mounted crane is fixedly arranged at the lower end of the transport truck to assist in carrying the wind power blade, the robot base is provided with a seventh shaft and is horizontally slidably arranged on the transport truck, and the high-pressure water jet head is arranged at the end of the robot wrist. Main electric control devices and a robot control module are arranged in the electric control cabinet. The robot drives the high-pressure water jet head to realize flexible control cutting, the positioning accuracy requirement of the wind power blade is reduced, so that an additional mechanical transmission machine is not needed, and the wind power blade does not need to be completely positioned.
[0026] Before scanning the wind power blade to be cut, the correspondence between the camera coordinate system and the robot coordinate system is established through hand-eye calibration, then the actual point cloud is acquired by using the 3D camera after hand-eye calibration, the actual point cloud comprises blade point cloud and ground point cloud, the actual point cloud is first filtered by using straight-through filtering, radius filtering and / or statistical filtering, then the actual point cloud is segmented to remove the ground point cloud, and the point cloud after removing the ground point cloud is clustered and down-sampled by using a method such as fast Euclidean clustering to obtain the point cloud to be processed.
[0027] Step S4: three initial cutting points are marked in the template point cloud according to requirements, the template point cloud is registered with the point cloud to be processed to obtain a rotation and translation matrix, and three actual cutting points respectively corresponding to the three initial cutting points in the point cloud to be processed are obtained according to the rotation and translation matrix; The template point cloud is registered with the point cloud to be processed by using the RANSCA+ICP algorithm to obtain a rotation and translation matrix T, the first actual cutting point is represented as m , wherein, m =1, 2, 3.
[0028] Step S5: a trajectory circle is fitted according to the three actual cutting points, a blade cutting trajectory is formed by points on the trajectory circle, for any point on the blade cutting trajectory, an orthogonal coordinate system is established at the point to provide attitude information for the robot, the robot cuts the wind power blade to be cut according to the cutting trajectory and the attitude information, and the wind power blade to be cut is transported to a designated location by the transport truck after cutting. Specifically include: Step S51: Fit a trajectory circle based on the three actual cutting points. The points on the trajectory circle form the blade cutting trajectory. The coordinates of each point on the blade cutting trajectory are calculated as follows: the center, radius, and normal vector of the trajectory circle are calculated. The normal vector is normalized within the trajectory circle to calculate two unit vectors with the center as the origin and perpendicular to each other. The coordinates of any point on the trajectory circle can then be calculated. Step S52: For the points on the cutting trajectory q , with a point q Three orthogonal unit vectors are established for the origin, and the rotation matrix of the robot posture is obtained from the three orthogonal unit vectors and the normal vector of the trajectory circle. The rotation matrix is converted into Euler angles to obtain the robot's posture information; among them, the specific processes of fitting the trajectory circle, calculating the coordinates of any point on the trajectory circle, and obtaining the robot's posture information are all existing technologies.
[0029] The above description is merely a preferred embodiment of the present invention and therefore cannot be used to limit the scope of the present invention. In other words, equivalent changes and modifications made according to the scope of the patent application and the contents of the specification should still fall within the scope of the patent of the present invention.
Claims
1. A vision-guided wind turbine blade cutting method, characterized by: The steps include: Step S1: In a specific location, a wind turbine blade serving as a template is scanned at multiple angles using a 3D camera to obtain a standard point cloud, and the standard point cloud is preprocessed to remove the ground point cloud to obtain an initial template point cloud containing only the blade point cloud; Step S2: uniformly downsample the initial template point cloud to obtain a template point cloud. During the downsampling process, different sampling intervals are determined based on the relationship between the average Gaussian curvature of the initial template point cloud and the principal curvature of each point cloud. Step S3: driving a transport truck equipped with a cutting robot to the location of the wind turbine blade to be cut, moving the wind turbine blade to be segmented to a location where the robot can cut it, scanning the wind turbine blade to be segmented using a 3D camera to obtain an actual point cloud, and preprocessing the actual point cloud to remove the ground point cloud and downsample it to obtain a processed point cloud containing only the point cloud of the blade to be cut; Step S4: Mark three initial cutting points in the template point cloud as required, align the template point cloud with the point cloud to be processed to obtain a rotation and translation matrix, and obtain three actual cutting points in the point cloud to be processed corresponding to the three initial cutting points according to the rotation and translation matrix; Step S5: Fit a trajectory circle based on the three actual cutting points, and the points on the trajectory circle constitute the blade cutting trajectory. For any point on the blade cutting trajectory, an orthogonal coordinate system is established with this point to provide posture information for the robot. The robot cuts the wind turbine blade to be cut according to the cutting trajectory and posture information, and after cutting, it is transported to the designated location by a transport truck.
2. The visually guided wind turbine blade cutting method according to claim 1, characterized in that: In step S1, the wind turbine blade is shifted multiple times at a set rotation angle, and a 3D camera is used to scan after each rotation. The data obtained after each scan are spliced to obtain a standard point cloud.
3. The visually guided wind turbine blade cutting method according to claim 2, characterized in that: In step S1, preprocessing the standard point cloud includes: Step S11: traverse all point clouds in the standard point cloud. i Point Cloud p i =( x i ,y i , z i )of z i If the value is greater than the set height threshold, the point cloud is attributed to the leaf point cloud set; otherwise, the point cloud is attributed to the ground set G; Step S12: Select the ground set G z The smallest coordinate k Point clouds form a seed point cloud set. For each seed point cloud s j , if the point cloud within the spherical neighborhood of the set radius p j The angle between the normal vector of the seed point cloud and the seed point cloud is not greater than the set angle threshold, then the seed point cloud is moved into the ground set G. When there is no point cloud in the seed point cloud set and it is moved into the ground set G, the final leaf point cloud set Bc is obtained, where point p j Belongs to standard point cloud; Step S13: Using a density-based noise spatial clustering algorithm to cluster the final leaf point cloud set Bc to obtain an initial template point cloud.
4. The visually guided wind turbine blade cutting method according to claim 3, characterized in that: In step S2, the average Gaussian curvature of the initial template point cloud is With principal curvature The relationship between The point cloud is divided into the first type of point cloud, The point cloud is divided into the second type of point cloud, The point cloud is divided into the third category of point cloud, and the sampling interval ratio of the first category of point cloud, the second category of point cloud and the third category of point cloud is set to 3:2:1, where , n represents the number of point clouds in the initial template point cloud, Represents the point cloud in the initial template point cloud p i' The principal curvature of .
5. The vision-guided wind turbine blade cutting method according to claim 4, characterized in that: In step S4, the template point cloud and the point cloud to be processed are registered using the RANSCA+ICP algorithm to obtain the rotation and translation matrix T. m The actual cutting point is expressed as ,in, m =1,2,3.
6. The vision-guided wind turbine blade cutting method according to claim 5, characterized in that: The step S5 specifically includes the following steps: Step S51: Fit a trajectory circle based on the three actual cutting points. The points on the trajectory circle form the blade cutting trajectory. The coordinates of each point on the blade cutting trajectory are calculated as follows: the center, radius, and normal vector of the trajectory circle are calculated. The normal vector is normalized within the trajectory circle to calculate two unit vectors with the center as the origin and perpendicular to each other. The coordinates of any point on the trajectory circle can then be calculated. Step S52: For the points on the cutting trajectory q , with a point q Three orthogonal unit vectors are established for the origin. The rotation matrix of the robot posture is obtained from the three orthogonal unit vectors and the normal vector of the trajectory circle. The rotation matrix is converted into Euler angles to obtain the robot posture information.
7. The vision-guided wind turbine blade cutting method according to any one of claims 3 to 6, characterized in that: The pre-processing of step S1 further includes filtering the standard point cloud using voxel filtering, and after filtering, the process proceeds to step S11.
8. The vision-guided wind turbine blade cutting method according to any one of claims 1 to 6, characterized in that: In step S3, before scanning the wind turbine blade to be segmented, a correspondence between the camera coordinate system and the robot coordinate system is established through hand-eye calibration.
9. The vision-guided wind turbine blade cutting method according to any one of claims 1 to 6, characterized in that: The transport truck is provided with a cutting system, which includes a truck crane, a robot, a high-pressure water jet head and a water receiving pool. The lower end of the truck crane is fixedly arranged on the transport truck, the robot is horizontally slidably arranged on the transport truck, and the high-pressure water jet head is arranged at the end of the robot wrist.
10. The vision-guided wind turbine blade cutting method according to any one of claim 9, characterized in that: A control room is also provided on the transport truck, which includes a water tank, a water softener, a sand supply machine and a high-pressure water pump. The high-pressure water pump is connected to the water tank, the water softener is connected between the high-pressure water pump and the high-pressure water jet head to soften hard water, and the sand supply machine is connected to the sand storage container of the high-pressure water jet head.