A method and equipment for offline automatic laser cleaning of tire molds.

Abstract

This invention belongs to the field of laser cleaning technology, specifically a method and equipment for offline automatic laser cleaning of a tire mold. Its structure includes: a robotic arm, a 3D camera, an additional rotating axis, a laser cleaning head, and a sliding platform. The process involves capturing a 3D point cloud of the tire mold's interior, stitching the point cloud into the robotic arm's coordinate system based on hand-eye calibration; dividing the cylindrical surface of the point cloud into several concentric grid regions, obtaining the coordinates of the center point and normal vector of each grid region through plane fitting; selecting a cross-section of the normal vector of a grid region, flattening and aligning the arc, and calculating the height difference between the highest and lowest points to obtain the tread pattern height; then obtaining the tread pattern spacing by calculating the distance between the tops of each tread pattern; calculating the coordinates and rotation angle of the robotic arm's TCP (Temperature Control Point) based on the coordinates of the plane center point, the normal vector, and the tread pattern height and spacing; and determining the rotation angle of the additional rotating axis at the end of the robotic arm based on the angle between the normal vector and the central axis of the tire mold.

Description

Technical Field

[0001] This invention belongs to the field of laser cleaning technology, specifically relating to an offline automatic laser cleaning method and equipment for the entire tire mold. Background Technology

[0002] The production of vehicle tires requires the use of different types of molds, and the main production process is injection molding. During production, rubber, sulfides, and other residues inevitably adhere to the surface of the tire molds. Therefore, the tire molds must be cleaned periodically; otherwise, the quality of the tire products will be affected. Traditional tire mold cleaning methods are mainly manual, using handheld sandblasting, dry ice, and chemical reagents to remove residues. This method is inefficient, produces poor quality, is environmentally unfriendly, involves harsh working conditions, and has high material costs. In contrast, laser cleaning offers advantages such as high efficiency, low energy consumption, precision, environmental friendliness, and ease of automation.

[0003] Furthermore, tire mold cleaning is divided into online cleaning and offline cleaning. Online cleaning refers to cleaning the mold directly at the production station, while offline cleaning refers to removing the mold from the production line and placing it in a dedicated cleaning station. Traditional offline cleaning equipment often requires disassembling the entire tire mold into multiple tread blocks for separate cleaning, which is complex and inefficient. In addition, existing technologies neglect the problems of point cloud loss caused by tread block occlusion and incomplete cleaning due to laser line obstruction during cleaning.

[0004] Therefore, there is a need to provide an offline cleaning method for the entire tire mold to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an offline automatic laser cleaning method and equipment for tire molds. By using point cloud stitching and pattern recognition, the accuracy of point cloud and the coordinates and angles of the robotic arm TCP are improved, effectively overcoming the problem of pattern occlusion and improving the cleaning effect.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for generating an offline automatic laser cleaning trajectory for a tire mold. The laser cleaning device includes a robotic arm, a 3D camera mounted at the end of the robotic arm, and an additional rotating axis. A laser cleaning head is mounted on the additional rotating axis. The trajectory generation method includes the following steps: S1. Capture the 3D point cloud inside the tire mold and stitch the 3D point cloud onto the robotic arm coordinate system according to hand-eye calibration; S2. Divide the cylindrical surface of the 3D point cloud into several rings, and further divide each ring into several grid regions. For the point cloud of each grid region, obtain the coordinates of the center point of the plane of the grid region and the corresponding normal vector through plane fitting. S3. Select a cross section in the normal direction of the grid region and flatten and align the arc. Using the normal vector direction as the height direction, calculate the height difference between the highest and lowest points of the point cloud to obtain the tread height h of the tire mold. Then, connect each highest point in the grid region into a horizontal line, take the set of points d away from the horizontal line to obtain the top point set of the tread, and then obtain the tread spacing w through the distance between each tread top. S4. Calculate the coordinates and rotation angle of the robotic arm TCP based on the coordinates of the center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. Determine the rotation angle of the additional rotation axis at the end of the robotic arm based on the angle between the normal vector and the direction of the tire mold center axis.

[0007] Furthermore, in step S3, the laser incident angle = arctan(0.5w / h) is determined according to the pattern height h and the spacing w, and the coordinates and rotation angle of the robotic arm TCP are adjusted according to the laser incident angle, wherein the laser incident angle is the angle between the laser emission direction and the normal vector; The d = pattern height h 10%, and the size of each grid region is equal to the linewidth of the laser emitted from the laser cleaning head.

[0008] Furthermore, in step S1, the 3D point cloud inside the tire mold captured is a local point cloud of the tire mold, and the coverage area of ​​the local point cloud is greater than or equal to the smallest repeating structural unit of the tire mold; the local point cloud is sequentially processed by downsampling, clustering, extracting the whole mold axis of the tire mold, and filling in the missing point cloud to obtain the complete point cloud of the whole tire mold; the complete point cloud is used to establish a cylindrical coordinate system according to the whole mold axis, where the z-axis is the direction of the whole mold axis, the x-axis is the direction of laser emission, and the y-axis is parallel to the axis of the additional rotation axis.

[0009] Furthermore, the extraction of the overall model axis specifically includes: firstly performing a normal filter on the clustered point cloud to remove points whose normals have an angle greater than 30° with the z-axis; then sorting the points from largest to smallest, for the first 50% of the point cloud, each point cloud is first evaluated for its flatness using the RANSAC plane fitting algorithm, and points with flatness below a preset threshold are removed; then, for the remaining point cloud, edges are extracted, and RANSAC arc fitting is performed on the edge points to find several fitted arcs; finally, the several arcs are evaluated, and the point cloud with the most points whose arc degree is higher than the preset threshold is selected, and the arc axis is calculated as the overall model axis.

[0010] Furthermore, the process of completing the missing point cloud specifically includes: inputting the captured local point cloud into a trained neural network model, completing the missing parts of the local point cloud region, and then rotating the array around the central axis of the whole model and removing outliers to obtain the complete point cloud of the whole model. The neural network model is obtained by scanning the internal point cloud of various tire molds with a 3D camera, manually hollowing out the internal point cloud to form a missing point cloud as the input of the training set, and using the unhollowed point cloud as the output to construct the training set, and training the neural network model.

[0011] A second aspect of the present invention provides an offline automatic laser cleaning trajectory generation system for tire molds, comprising: The point cloud stitching module is used to capture 3D point clouds inside the tire mold and stitch the 3D point clouds into the coordinate system of the robotic arm according to hand-eye calibration. The region division module is used to divide the cylindrical surface of the 3D point cloud into several rings, and each ring is further divided into several grid regions. The coordinates of the center point of the plane of each grid region and the corresponding normal vector are obtained by plane fitting for the point cloud of each grid region. The pattern recognition module is used to select a cross section in the normal direction of the grid area, flatten and align the arc, calculate the height difference between the highest and lowest points of the point cloud with the normal vector direction as the height direction, and obtain the pattern height h of the tire mold; then connect each highest point in the grid area into a horizontal line, take the point set d away from the horizontal line to obtain the pattern top point set, and then obtain the pattern spacing w through the distance between each pattern top; The trajectory generation module is used to calculate the coordinates and rotation angle of the robotic arm tcp based on the coordinates of the plane center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. The rotation angle of the additional rotation axis at the end of the robotic arm is determined based on the angle between the normal vector and the direction of the tire mold central axis.

[0012] The third aspect of the present invention provides a method for offline automatic laser cleaning of a tire mold, comprising: generating the coordinates and rotation angles of each revolution of the robotic arm TCP of the tire mold to be cleaned, as well as the rotation angle of the additional rotation axis, using the aforementioned method for generating the offline automatic laser cleaning trajectory of the tire mold; Then, according to the coordinates and rotation angles, the laser emitted by the laser cleaning head is used to scan and clean each grid area of ​​each circle in sequence.

[0013] Furthermore, it also includes: converting the RGB values ​​of the 3D point cloud to grayscale values, determining the thickness of the residue on each ring of the tire mold surface based on the grayscale values, setting an interval threshold according to the grayscale values, increasing the cleaning count for that ring by 1 for each ring exceeding a threshold, and repeating the cleaning trajectory for each ring according to the cleaning count.

[0014] Furthermore, the scanning and cleaning process specifically includes: first, adjusting the laser beam direction to align with the normal vector of the grid region; then, adjusting the incident angle so that the angle between the beam direction and the normal vector equals the incident angle, thus obtaining the final TCP coordinates and angle; finally, connecting the trajectory points of each grid region at the same height sequentially to obtain a cleaning trajectory; for each trajectory point, solving the inverse kinematics of the robotic arm, and performing interference calculations between the 3D model of the robotic arm and the complete point cloud of the tire mold; if interference occurs, adjusting the additional axis rotation angle and TCP rotation angle until there is no interference. Each cleaning trajectory is cleaned at least twice: the first time at a positive incident angle, and the second time at a negative incident angle.

[0015] The fourth aspect of the present invention provides an offline automatic laser cleaning device for tire molds, comprising: a robotic arm, a sliding platform and a control unit, wherein a 3D camera and an additional rotating shaft are installed at the end of the robotic arm, and a laser cleaning head is installed on the additional rotating shaft; The sliding platform is used to transport the entire tire mold to the cleaning station. The control unit is used to control the operation of the robotic arm, 3D camera, additional rotating shaft and laser cleaning head according to the above-described tire mold offline automatic laser cleaning method.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention mainly possess the following technical advantages: 1. The tire mold offline automatic laser cleaning trajectory generation method provided by this invention utilizes point cloud stitching to obtain the coordinate distribution of the inner wall of the tire mold. By segmenting into grid regions, it is easy to determine the cleaning trajectory points of each region based on the coordinates of the center point and the normal vector. Based on the point cloud, the tread height and width are identified, facilitating the calculation of the laser incident angle, thereby ensuring that the laser is not blocked and achieving thorough cleaning. Therefore, this invention can complete cleaning without disassembling the mold. Compared with traditional offline cleaning methods, whole-mold cleaning is simpler to operate, has a higher degree of automation, and is more efficient.

[0017] 2. This invention determines the thickness of the residue based on the grayscale value of the point cloud, and then sets the number of cleaning cycles accordingly to ensure complete cleaning. Furthermore, each cleaning trajectory is cleaned at least twice: the first time at a forward incident angle, and the second time at a reverse incident angle, to ensure that the inner wall of the pattern is cleaned.

[0018] 3. This invention obtains a complete point cloud by capturing local point clouds and then extracting and rotating the array using the whole model axis. This method reduces data processing volume, increases efficiency, and facilitates missing data completion for local captures, preventing point cloud gaps caused by factors such as reflections from inner walls. Furthermore, this invention utilizes a pre-trained neural network for point cloud completion, achieving good completion results and high efficiency. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall structure of the cleaning equipment according to an example of the present invention; Figure 2 This is a schematic cross-sectional view of the tire mold of the present invention; Figure 3 This is a flowchart of the trajectory generation process of the present invention; Figure 4 This is a schematic diagram of the cleaning equipment and trajectory generation coordinate system of the present invention; Figure 5 This is a schematic diagram of the process for extracting the whole model axis in the point cloud processing algorithm of the present invention; Figure 6 This is a schematic diagram illustrating the process of completing missing point clouds and dividing grid regions in this invention; Figure 7 A schematic diagram of the structure used to determine the incident angle for pattern cleaning. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 Please see Figure 1-7 A method for generating an offline automatic laser cleaning trajectory for a tire mold, wherein the laser cleaning device includes a robotic arm, a 3D camera mounted at the end of the robotic arm, and an additional rotating axis, wherein a laser cleaning head is mounted on the additional rotating axis, and the trajectory generation method includes the following steps: S1. Capture the 3D point cloud inside the tire mold and stitch the 3D point cloud onto the robotic arm coordinate system according to hand-eye calibration; S2. Divide the cylindrical surface of the 3D point cloud into several rings, and further divide each ring into several grid regions. For the point cloud of each grid region, obtain the coordinates of the center point of the plane of the grid region and the corresponding normal vector through plane fitting. S3. Select a cross section in the normal direction of the grid region and flatten and align the arc. Using the normal vector direction as the height direction, calculate the height difference between the highest and lowest points of the point cloud to obtain the tread height h of the tire mold. Then, connect each highest point in the grid region into a horizontal line, take the set of points d away from the horizontal line to obtain the top point set of the tread, and then obtain the tread spacing w through the distance between each tread top. S4. Calculate the coordinates and rotation angle of the robotic arm TCP based on the coordinates of the center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. Determine the rotation angle of the additional rotation axis at the end of the robotic arm based on the angle between the normal vector and the direction of the tire mold center axis.

[0022] Specifically, a sliding platform is used to transport the entire tire mold to the workstation, and point cloud acquisition, trajectory generation, and subsequent cleaning are performed directly on the sliding platform.

[0023] A schematic diagram of the overall cross-sectional view of the tire mold described in this invention is shown below. Figure 2 As shown, the tire mold is often composed of multiple tread blocks. The term "complete mold" refers to the state in which all the tread blocks of the tire mold are installed together to form a whole. The tread blocks of the tire mold are installed within the mold frame, and multiple tread blocks form a complete circle, roughly cylindrical in shape.

[0024] In step S1, the 3D point cloud inside the tire mold captured is a local point cloud of the tire mold, and the coverage area of ​​this local point cloud is greater than or equal to the smallest repeating structural unit of the tire mold, that is, the tire mold has a symmetrical structure. Capturing its local point cloud and rotating it allows for the acquisition of the entire point cloud. This reduces the data processing complexity of capturing all point clouds. Furthermore, because the mold has patterns, there may be missed shots leading to partial gaps, and completing all point clouds is also quite complex. The local point cloud is then sequentially downsampled, clustered, and the entire mold axis of the tire mold is extracted. Missing point clouds are then filled in to obtain the complete point cloud of the entire tire mold. A cylindrical coordinate system is established based on the entire mold axis, where the z-axis is the direction of the entire mold axis (perpendicular to the worktable plane), the x-axis is the laser emission direction, and the y-axis is parallel to the axis of the additional rotation axis. Figure 4 The additional rotating axis can rotate around the y-axis, thereby changing the direction of the laser beam, that is, changing the elevation angle of the laser.

[0025] Specifically, point cloud processing algorithms include point cloud stitching, downsampling, and clustering, which are considered preprocessing steps. Point cloud stitching mainly involves a robotic arm driving a 3D camera to photograph different areas of a tire mold. Then, using hand-eye calibration and the pose of the photographed points, the point cloud is rotated, translated, and stitched into the robot's coordinate system. Downsampling utilizes 3D voxel sampling. Clustering mainly divides the point cloud into multiple point sets according to specific intervals.

[0026] like Figure 5As shown, the extraction of the overall model axis specifically includes: firstly, performing a normal filter on the clustered point cloud to remove points whose normals have an angle greater than 30° with the z-axis, the purpose of which is to remove point clouds on surfaces parallel to the axis direction and retain as many point clouds as possible that are parallel to the horizontal plane; then, sorting the point clouds by the number of points from largest to smallest, for the first 50% of the point clouds, each point cloud is first evaluated for its flatness using the RANSAC plane fitting algorithm, and point clouds with flatness below a preset threshold are removed; then, for the remaining point clouds, edges are extracted, and RANSAC arc fitting is performed on the edge points to find several fitted arcs; finally, several arcs are evaluated, and the point cloud with the most points whose arc degree is higher than the preset threshold is selected, and the axis of the arc is calculated, which is taken as the overall model axis.

[0027] like Figure 6 The process of completing the missing point cloud specifically includes: during cleaning, inputting the captured local point cloud into a trained neural network model, outputting the completed point cloud to complete the missing parts of the local point cloud region, and then using the whole model axis as the central axis to rotate the array and correct outliers to obtain the complete point cloud of the whole model.

[0028] The neural network model is obtained by scanning the internal point cloud of various tire molds with a high-precision 3D camera before cleaning, manually hollowing out the internal point cloud to form a missing point cloud as the input of the training set, and using the unhollowed point cloud as the output to construct the training set, and then training the neural network model.

[0029] In step S2, the grid area is divided as follows: Figure 6 As shown in the blue boxes, each ring corresponds to one ring of the inner wall of the mold. Each ring is a square region with a side length equal to the laser linewidth, ensuring that each scan covers the region. RANSAC plane fitting is performed on the patchy point cloud of each region to obtain the plane center point and normal vector, and the point set is recorded.

[0030] In step S3, the laser incident angle is determined as arctan(0.5w / h) based on the pattern height h and spacing w. The coordinates and rotation angle of the robotic arm TCP are then adjusted according to this laser incident angle, which is the angle between the laser emission direction and the normal vector. Adjusting the incident angle in this way ensures that the laser is not blocked by the pattern, thus improving cleaning quality.

[0031] like Figure 7 The red area represents the top point set of the pattern, where d = pattern height h. 10%, and the size of each grid region is equal to the linewidth of the laser emitted from the laser cleaning head.

[0032] In step S4, using the coordinates of the center point of the point cloud mesh region and the normal vector obtained in the previous step, the xyz coordinates and xyz rotation angle of the robotic arm TCP (and the center point of the laser line) are calculated. Based on the angle between the normal vector and the z-axis, the rotation angle of the additional rotation axis at the end of the robotic arm is calculated, thus simultaneously obtaining the trajectory point of the robotic arm and the corresponding additional axis rotation angle. Specifically, the rotation angle of the additional rotation axis is adjusted so that the laser emission direction is parallel to the normal vector of the mesh region. The xyz coordinates and xyz rotation angle of the robotic arm TCP ensure that the angle between the emission direction and the normal vector during the scanning process is always equal to the incident angle = arctan(0.5w / h).

[0033] Example 2 A tire mold offline automatic laser cleaning trajectory generation system includes: The point cloud stitching module is used to capture 3D point clouds inside the tire mold and stitch the 3D point clouds into the coordinate system of the robotic arm according to hand-eye calibration. The region division module is used to divide the cylindrical surface of the 3D point cloud into several rings, and each ring is further divided into several grid regions. The coordinates of the center point of the plane of each grid region and the corresponding normal vector are obtained by plane fitting for the point cloud of each grid region. The pattern recognition module is used to select a cross section in the normal direction of the grid area, flatten and align the arc, calculate the height difference between the highest and lowest points of the point cloud with the normal vector direction as the height direction, and obtain the pattern height h of the tire mold; then connect each highest point in the grid area into a horizontal line, take the point set d away from the horizontal line to obtain the pattern top point set, and then obtain the pattern spacing w through the distance between each pattern top; The trajectory generation module is used to calculate the coordinates and rotation angle of the robotic arm tcp based on the coordinates of the plane center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. The rotation angle of the additional rotation axis at the end of the robotic arm is determined based on the angle between the normal vector and the direction of the tire mold central axis.

[0034] Example 3 A method for offline automatic laser cleaning of a tire mold includes: generating the coordinates and rotation angles of each revolution of the robotic arm TCP of the tire mold to be cleaned, as well as the rotation angle of the additional rotation axis, using the offline automatic laser cleaning trajectory generation method for the tire mold described in Example 1, i.e., generating the cleaning motion trajectory of the robotic arm and the additional rotation axis. Then, according to the coordinates and angles, the robotic arm is guided to execute the cleaning trajectory, and the laser emitted from the laser cleaning head is used to scan and clean each grid area of ​​each circle in sequence to clean the tire mold.

[0035] Specifically, the cleaning method also includes: since point clouds contain RGB information in addition to XYZ information, the RGB values ​​of the 3D point cloud are converted to grayscale values. The thickness of the residue on each ring of the tire mold surface is determined based on the grayscale value. A threshold is set according to the grayscale value. Each time the threshold is exceeded, the cleaning count for that ring is increased by one. This cleaning process is repeated for each ring until all cleaning rings are completed. The residue inside the tire mold is typically a relatively uniform 0.1-1mm thick vulcanized rubber composite film. It should be noted that at the same height, the distribution of residue in each repeating unit is considered to be basically consistent. The thickness of each ring can be determined based on the average grayscale value of that ring. In this way, the present invention utilizes the influence of residue thickness on grayscale; generally, the thicker the residue, the darker the color. Therefore, a relationship between grayscale value and thickness can be established, allowing for a rough determination of the thickness range. The number of cleaning cycles can then be set based on the thickness to ensure thorough cleaning.

[0036] Furthermore, the scanning and cleaning specifically includes: first adjusting the laser emission direction to be consistent with the normal vector of the grid region (this direction is adjusted by rotating an additional rotation axis around the y-axis), and then adjusting the incident angle (e.g., Figure 7 This process involves aligning the angle between the emitted light direction and the normal vector with the incident angle to obtain the final TCP coordinates and angles of the robotic arm. When cleaning each grid region, the midpoint of the emitted laser line is aligned with the center point of the plane. Then, according to... Figure 6 The laser scans horizontally, grid by grid, in the direction of the middle arrow. Please complete one scan before moving to the next. The purpose of adjusting the laser incident angle in this invention is to ensure that the laser is not blocked by the sidewalls of the pattern, thus preventing the blocked areas from being cleaned.

[0037] Finally, the trajectory points of each grid region at the same height are connected sequentially to obtain a cleaning trajectory. In particular, for each trajectory point, the inverse solution of the robotic arm is solved, and the interference calculation between the 3D model of the robotic arm and the complete point cloud of the tire mold is performed. If interference occurs, the additional axis rotation angle and TCP rotation angle are adjusted until there is no interference.

[0038] Each cleaning path is cleaned at least twice: the first time at a forward incident angle and the second time at a reverse incident angle, to ensure that the inner wall of the pattern is cleaned.

[0039] In actual production, the equipment's industrial control computer runs offline laser cleaning control software for tire molds. Operators only need to click the operation buttons on the software interface to drive the robotic arm and laser to automatically perform the cleaning.

[0040] The trajectory planning method described in this invention allows for the saving of a template after the cleaning trajectory for a specific type of tire mold has been planned. If another tire mold of the same type needs cleaning later, the saved trajectory template can be directly called, and corresponding rotation and translation transformations can be performed based on the actual point cloud position to complete the trajectory planning.

[0041] Example 4 An offline automatic laser cleaning device for tire molds includes: a robotic arm, a sliding platform, and a control unit. The end of the robotic arm is equipped with a 3D camera and an additional rotating shaft, and a laser cleaning head is mounted on the additional rotating shaft. The sliding platform is used to transport the entire tire mold to the cleaning station, and to place the entire tire mold to be cleaned on the sliding platform for point cloud collection and cleaning. The control unit is used to control the operation of the robotic arm, 3D camera, additional rotating shaft and laser cleaning head according to the tire mold offline automatic laser cleaning method described in Example 3.

[0042] The main process of offline automatic laser cleaning of tire molds described in this invention is as follows: 1. First, the tire mold is hoisted onto a sliding platform, and then the sliding platform transports the entire tire mold to the cleaning station; 2. The robotic arm drives a 3D camera to capture the 3D point cloud inside the mold; 3. Using laser cleaning vision guidance software installed on the industrial control computer, a cleaning motion trajectory of the robotic arm and an additional rotating axis is generated using a point cloud processing algorithm; 4. The robotic arm executes the cleaning trajectory, and the laser cleaning head emits light while the robotic arm is moving, cleaning the tire mold thoroughly.

[0043] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating an offline automatic laser cleaning trajectory for a tire mold, characterized in that, The device used for laser cleaning includes a robotic arm, a 3D camera mounted at the end of the robotic arm, and an additional rotating axis. A laser cleaning head is mounted on the additional rotating axis. The trajectory generation method includes the following steps: S1. Capture the 3D point cloud inside the tire mold, and stitch the 3D point cloud into the coordinate system of the robotic arm according to the hand-eye calibration. S2. Divide the cylindrical surface of the 3D point cloud into several rings, and further divide each ring into several grid regions. For the point cloud of each grid region, obtain the coordinates of the center point of the plane of the grid region and the corresponding normal vector through plane fitting. S3. Select a cross section in the normal direction of the grid region and flatten and align the arc. Using the normal vector direction as the height direction, calculate the height difference between the highest and lowest points of the point cloud to obtain the tread height h of the tire mold. Then, connect each highest point in the grid region into a horizontal line, take the set of points d away from the horizontal line to obtain the top point set of the tread, and then obtain the tread spacing w through the distance between each tread top. S4. Calculate the coordinates and rotation angle of the robotic arm TCP based on the coordinates of the center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. Determine the rotation angle of the additional rotation axis at the end of the robotic arm based on the angle between the normal vector and the direction of the tire mold center axis.

2. The method for generating offline automatic laser cleaning trajectory for the entire tire mold according to claim 1, characterized in that, In step S3, the laser incident angle = arctan(0.5w / h) is determined according to the pattern height h and the spacing w. The coordinates and rotation angle of the robotic arm TCP are adjusted according to the laser incident angle, where the laser incident angle is the angle between the laser emission direction and the normal vector. The d = pattern height h 10%, and the size of each grid region is equal to the linewidth of the laser emitted from the laser cleaning head.

3. The method for generating offline automatic laser cleaning trajectory for the entire tire mold according to claim 1, characterized in that, In step S1, the 3D point cloud inside the tire mold captured by the camera is a local point cloud of the tire mold, and the coverage area of ​​the local point cloud is greater than or equal to the smallest structural repeating unit of the tire mold. The local point cloud is then processed sequentially through downsampling, clustering, extraction of the entire mold axis of the tire mold, and completion of missing point clouds to obtain the complete point cloud of the entire tire mold. A cylindrical coordinate system is established for the complete point cloud according to the entire mold axis, where the z-axis is the direction of the entire mold axis, the x-axis is the direction of laser emission, and the y-axis is parallel to the axis of the additional rotation axis.

4. The method for generating offline automatic laser cleaning trajectory for the entire tire mold according to claim 3, characterized in that, The extraction of the overall model axis specifically includes: firstly, performing a normal filter on the clustered point cloud to remove points whose normals have an angle greater than 30° with the z-axis; then, sorting the points from largest to smallest, for the first 50% of the point cloud, using the RANSAC plane fitting algorithm to evaluate the flatness of each point cloud, removing point clouds with flatness below a preset threshold; then, for the remaining point cloud, extracting edges, performing RANSAC arc fitting on the edge points to find several fitted arcs; finally, evaluating these arcs, selecting the point cloud with the highest arc degree above the preset threshold and the most points, and calculating the arc axis as the overall model axis.

5. The method for generating offline automatic laser cleaning trajectory for the entire tire mold according to claim 3, characterized in that, The process of completing the missing point cloud specifically includes: inputting the captured local point cloud into a trained neural network model, completing the missing parts of the local point cloud region, and then rotating the array around the central axis of the whole model and removing outliers to obtain the complete point cloud of the whole model. The neural network model is obtained by scanning the internal point cloud of various tire molds with a 3D camera, manually hollowing out the internal point cloud to form a missing point cloud as the input of the training set, and using the unhollowed point cloud as the output to construct the training set, and training the neural network model.

6. A tire mold offline automatic laser cleaning trajectory generation system, characterized in that, include: The point cloud stitching module is used to capture 3D point clouds inside the tire mold and stitch the 3D point clouds into the coordinate system of the robotic arm according to hand-eye calibration. The region division module is used to divide the cylindrical surface of the 3D point cloud into several rings, and each ring is further divided into several grid regions. The coordinates of the center point of the plane of each grid region and the corresponding normal vector are obtained by plane fitting for the point cloud of each grid region. The pattern recognition module is used to select a cross section in the normal direction of the grid area, flatten and align the arc, calculate the height difference between the highest and lowest points of the point cloud with the normal vector direction as the height direction, and obtain the pattern height h of the tire mold; then connect each highest point in the grid area into a horizontal line, take the point set d away from the horizontal line to obtain the pattern top point set, and then obtain the pattern spacing w through the distance between each pattern top; The trajectory generation module is used to calculate the coordinates and rotation angle of the robotic arm tcp based on the coordinates of the plane center point of all grid areas in each circle, the normal vector, and the pattern height and spacing. The rotation angle of the additional rotation axis at the end of the robotic arm is determined based on the angle between the normal vector and the direction of the tire mold central axis.

7. A method for offline automatic laser cleaning of a tire mold, characterized in that, include: The method for generating the offline automatic laser cleaning trajectory of the entire tire mold as described in any one of claims 1-5 is used to generate the coordinates and rotation angles of each rotation of the robotic arm TCP of the tire mold to be cleaned, as well as the rotation angle of the additional rotation axis. Then, according to the coordinates and rotation angles, the laser emitted by the laser cleaning head is used to scan and clean each grid area of ​​each circle in sequence.

8. The offline automatic laser cleaning method for the entire tire mold according to claim 7, characterized in that, Also includes: The RGB values ​​of the 3D point cloud are converted to grayscale values. The thickness of the residue on each ring of the tire mold surface is determined based on the grayscale values. An interval threshold is set according to the grayscale values. Each time the threshold is exceeded, the cleaning count for that ring is increased by 1. The cleaning trajectory for each ring is repeated according to the number of cleaning counts.

9. The offline automatic laser cleaning method for the entire tire mold according to claim 7, characterized in that, The scanning and cleaning process specifically includes: first, adjusting the laser beam direction to align with the normal vector of the grid region; then, adjusting the incident angle so that the angle between the beam direction and the normal vector equals the incident angle, thus obtaining the final TCP coordinates and angles; finally, connecting the trajectory points of each grid region at the same height sequentially to obtain a cleaning trajectory; for each trajectory point, solving the inverse kinematics of the robotic arm, and performing interference calculations between the 3D model of the robotic arm and the complete point cloud of the tire mold; if interference occurs, adjusting the additional axis rotation angle and TCP rotation angle until there is no interference. Each cleaning trajectory is cleaned at least twice: the first time at a positive incident angle, and the second time at a negative incident angle.

10. An offline automatic laser cleaning device for tire molds, characterized in that, include: The robotic arm, sliding platform, and control unit are provided. The end of the robotic arm is equipped with a 3D camera and an additional rotating axis, on which a laser cleaning head is mounted. The sliding platform is used to transport the entire tire mold to the cleaning station. The control unit is used to control the operation of the robotic arm, 3D camera, additional rotating shaft and laser cleaning head in the tire mold offline automatic laser cleaning method according to any one of claims 7-9.

Images