A composite material tee joint polishing system based on visual positioning

Abstract

The application discloses a kind of based on visual guidance's composite material three-way pipe fittings polishing system, including rotatable main shaft, polishing device including force sensor, depth camera, polishing robot, host computer, specifically includes the following steps: polishing robot carries depth camera and automatically looks for image acquisition optimal pose, depth camera installed on polishing robot gathers the image information of composite material three-way pipe fittings fixed on main shaft and sends to host computer, and host computer processes image information by YOLOv8-CGF visual identification algorithm and calculates pipe fitting center point coordinates, pose, controls polishing robot to drive polishing device including force sensor to carry out polishing to composite material three-way pipe fittings according to the polishing method designed.The present application realizes the accurate positioning of different sizes of composite material three-way pipe fittings and automatic polishing of surface, compared with manual polishing, polishing effect is better, more stable, can effectively improve the efficiency and quality of polishing.

Description

Technical Field

[0001] This invention relates to the field of composite material pipe grinding, and particularly to a visual positioning-based composite material tee grinding system. Background Technology

[0002] Composite material tee fittings are characterized by their light weight, high strength, easy processing and molding, chemical corrosion resistance, good weather resistance, and strong design flexibility, making them widely used in aerospace, automotive, construction, and fitness equipment industries. In practical applications, processes such as spraying and adding insulation layers to the surface of tee fittings are required, which places certain demands on the surface finish. However, traditional tee fitting grinding methods primarily rely on manual grinding. Manual grinding is inefficient and prone to errors, resulting in inconsistent product quality and hindering production.

[0003] CN202310094418.2 discloses a structured light 3D vision positioning method for grinding multi-type variable castings. It locates the grinding position of multi-type variable castings through vision, but it requires multiple scans of the casting surface to obtain the casting point cloud model, which is computationally complex and time-consuming. The present invention only needs to take surface images of tee fittings with a depth camera and identify the center points of the main pipe section and branch pipe section of the tee fitting through a visual recognition algorithm. It has a small computational load and fast recognition.

[0004] CN202211238361.0 discloses an adaptive grinding method for top cover brazing based on 3D line laser vision guidance. It uses 3D laser vision to scan the weld and segment the image. The grinding robot's trajectory is planned by connecting the coordinates of the center point of each image. However, the method is computationally complex, cumbersome to operate, and can only grind the fixed position of the top cover brazing weld. This invention can grind the surface of tee fittings of different sizes by allowing the operator to input different data of the tee fitting.

[0005] To address these issues, this patent proposes a novel grinding system for tee fittings. The key feature of this method is the use of machine vision to identify the center point of the tee fitting's opening for precise positioning of the composite material tee fitting. Compared to manual grinding, this method offers better and more stable grinding results, effectively improving grinding efficiency and quality. Furthermore, this method can automatically adjust grinding parameters based on the specific shape and size of the tee fitting, further enhancing grinding efficiency and quality. By combining machine vision and automation technologies, the grinding process for tee fittings has been optimized and upgraded. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a vision-based composite material tee pipe grinding system, solving the problems mentioned in the background section.

[0007] The present invention adopts the following technical solution:

[0008] This invention discloses a visual positioning-based composite material tee pipe polishing system. The polishing system includes: a rotatable spindle, a polishing device with a force sensor, a depth camera, a polishing robot, a robot base, and a host computer. The system comprises the following steps: establishing the pose relationship between the depth camera and the polishing device; the host computer sending a polishing start command to the polishing robot; the polishing robot automatically finding the optimal pose for image acquisition; the depth camera acquiring images of the composite material tee pipe fixed on the spindle and sending the image information back to the host computer; the host computer using the YOLOv8-CGF visual recognition algorithm to determine the coordinates of the center point of the pipe opening of the composite material tee pipe; calculating the current posture of the composite material tee pipe; the spindle driving the composite material tee pipe to rotate from the current posture to a preset posture; updating the tee pipe model based on the winding information of the composite material tee pipe; determining the initial polishing position and planning the polishing trajectory based on the polishing method; and the host computer controlling the polishing robot to drive the polishing device with the force sensor to polish each part of the composite material tee pipe fixed on the rotatable spindle according to the preset polishing method.

[0009] Furthermore, the pose relationship between the depth camera coordinate system and the grinding device is established. First, the grinding robot and the depth camera are firmly assembled. A 6×9 black and white checkered calibration board is placed in the field of view of the depth camera. The pose of the end effector of the grinding robot is adjusted, and 16 images of the calibration board fully exposed in the camera's field of view are acquired in four different poses. The pose relationship between the depth camera and the grinding device is obtained using the following formula.

[0010]

[0011] Where O represents the calibration plate coordinate system, E i E j C represents different orientations in the coordinate system of the grinding device, which includes the force sensor. i C j Representative and E i E j The corresponding pose is the depth camera pose, where B represents the base coordinates of the grinding robot. A matrix of the form is used to represent the relative pose relationship between the Y coordinate system and the X coordinate system.

[0012] Furthermore, the method for the grinding robot to automatically find the optimal pose for image acquisition is as follows: a depth camera installed on the grinding robot acquires images of the composite material tee fitting, and the distance between the grinding robot and the composite material tee fitting and the pose of the grinding robot are adjusted according to the size and position of the composite material tee fitting in the image until the composite material tee fitting is located at the center of the image captured by the depth camera.

[0013] Furthermore, the YOLOv8-CGF visual recognition algorithm is as follows: Based on the YOLOv8 network model, an attention mechanism with a CoorAtt structure is added. Channel attention is decomposed into two 1D feature encoding processes. Features of the composite material tee fitting feature map are aggregated along two spatial directions, and the resulting feature maps are encoded into a pair of direction-aware and position-sensitive attention maps, which are then applied to the input feature map to enhance the detection accuracy of the composite material tee fitting nozzle. The Neck part in the YOLOv8 network model is modified to a Gather-and-Distribute algorithm. The algorithm's structure is modified by altering the original YOLOv8 algorithm's Neck section into low-order and high-order clustering and distribution branches. The high-order branch is dedicated to identifying composite tee fittings over long distances, while the low-order branch is dedicated to locating the center point of the fitting's opening at close range. Finally, Focaler-IoU is used as the algorithm's loss function. When IoU is less than a lower threshold, the loss is 0; when IoU is greater than an upper threshold, the loss is 1; and when IoU is between the lower and upper thresholds, the loss is a function that increases linearly with the IoU value.

[0014] Furthermore, the steps for identifying the coordinates of the center point of the composite material tee fitting nozzle and calculating its current orientation using YOLOv8-CGF are as follows:

[0015] a. Construct a dataset by collecting 800 images of composite material tee fittings. Use the software Labelimg to calibrate the images of composite material tee fittings, manually divide the pipe opening portion of the tee fitting in the images, and randomly divide the divided images into training set, validation set and test set according to a ratio of 6:2:2. Use the YOLOv8-CGF network model to train and obtain the training weights for detecting the center point of the pipe opening of the composite material tee fitting.

[0016] b. After the grinding robot (5) carrying the depth camera (4) moves to the designated position, the depth camera (4) collects the color image information and depth image information of the composite material tee fitting and sends them into the YOLOv8-CGF network model loaded with training weights. After the collected color image extracts the image features through the backbone of the model, it is sent into the Gather-and-Distribute structure. The image is classified by the Gather-and-Distribute structure and sent to the detection head.

[0017] c. During detection, the attention mechanism of the CoorAtt structure is applied, enabling the model to adaptively learn different regions of the image. By improving the localization accuracy of the detection box, the detection accuracy of the composite material tee fitting opening is improved. The Focaler-IoU is used as the loss function of the YOLOv8-CGF algorithm at the output end to further improve the localization accuracy of the detection box.

[0018] d. The formula for calculating the center point coordinates of composite material tee fittings is:

[0019]

[0020] Where x1 is the X-axis coordinate of the center point of the main pipe section 1, l1 is the length of the main pipe section 1, l is the length of the intersection, y1 is the Y-axis coordinate of the center point of the main pipe section 1, y2 is the Y-axis coordinate of the center point of the main pipe section 2, z1 is the Z-axis coordinate of the center point of the main pipe section 1, and z2 is the Z-axis coordinate of the center point of the main pipe section 2.

[0021] e. Taking the center point of the main pipe section as the origin O of the target coordinate system, and the straight line passing through the center point of the main pipe section and the center point of the identified branch pipe opening as the Y-axis of the target coordinate system, with the positive direction being from the center point of the main pipe section to the center point of the branch pipe opening, and the principal axis as the X-axis of the target coordinate system, with the positive direction being from main pipe section 1 to main pipe section 2, and the Z-axis being determined by the right-hand rule, after constructing the target coordinate system, the target coordinate system is transformed into the end coordinate system of the grinding robot through coordinate transformation. The transformation matrix is ​​the posture of the composite material tee fitting.

[0022] Furthermore, the preset posture is a posture in which the Z-axis of the target coordinate system is parallel to the Z-axis of the end coordinate system of the grinding robot.

[0023] Furthermore, the model of the composite material tee fitting will change after winding. The tee fitting model is updated based on the winding information to obtain updated model data. The grinding path planned based on the new model data is more accurate. The formula for calculating the updated model thickness is as follows:

[0024]

[0025]

[0026] Where, m i (i = 1, 2, 3, 4) are undetermined coefficients, r i The radius of the latitude circle, ρ is the compressibility coefficient of the fiber-reinforced layer unit, n is the number of fiber layers in the unit, and ρ a V is the unit density. f ρ is the fiber volume fraction, and ρ is the fiber density.

[0027] Furthermore, the composite material tee fitting is divided into four parts: main pipe section 1, intersection section, branch pipe section, and main pipe section 2. The parameters of each part can be customized according to the actual situation.

[0028] Furthermore, the grinding start positions of the branch pipe section, the main pipe section 1, and the main pipe section 2 are the intersection points of the target coordinate system XOY plane with the branch pipe section opening, the target coordinate system XOY plane with the main pipe section 1 opening, and the target coordinate system XOY plane with the main pipe section 2 opening, respectively. The grinding start position of the intersection is the intersection point of the cross section at the boundary between the intersection and the main pipe section 1 with the target coordinate system XOY plane.

[0029] Furthermore, the first grinding strategy involves a grinding robot driving a grinding device containing a force sensor to begin horizontal movement from the starting position of grinding the main pipe section of the composite material tee fitting. At the same time, the spindle drives the composite material tee fitting to rotate. The grinding device stops at the junction of the main pipe section and the intersection, and then moves horizontally in the opposite direction. This completes one reciprocating motion, and the reciprocating motion is repeated multiple times until the surface of the main pipe section is ground.

[0030] Furthermore, the ratio k of the horizontal movement speed of the grinding robot driving the grinding device containing the force sensor from the initial position to the rotational movement speed of the spindle is constant, and k satisfies:

[0031] k=2πrcotα

[0032] Where r represents the cross-sectional radius of the composite material tee fitting, and a represents the angle between the grinding trajectory and the horizontal line, with 0°... <a<90°。

[0033] Furthermore, in the second grinding strategy, the composite material tee fitting is rotated by the main shaft, and the grinding robot drives the grinding device containing the force sensor to perform a horizontal reciprocating motion from left to right starting from the grinding start position of the branch pipe section. The motion trajectory coordinates of the grinding robot satisfy the following formula:

[0034]

[0035] Among them, θ, θ′, It is a function of the cross-sectional radius r of the branch pipe of the composite material tee fitting.

[0036] Furthermore, the third grinding strategy involves rotating the composite material tee fitting via the main axis. The grinding robot drives the grinding device, which includes a force sensor, to start grinding from the grinding start position at the intersection. The grinding robot avoids collision between the grinding device and the branch pipe section by moving to the arc transition part of the composite material tee fitting. After grinding is completed, the grinding robot drives the grinding device, which includes a force sensor, back to the initial position.

[0037] Furthermore, in the third polishing strategy, the ratio K of the horizontal motion speed of the polishing robot driving the polishing device containing the force sensor at the starting position to the rotational motion speed of the spindle satisfies the following:

[0038] K = 35.36sin(0.01σ - 1.6) + 4.78

[0039] Where σ represents the rotation angle of the spindle.

[0040] Furthermore, the force sensor plays a role in the grinding process to prevent the grinding wheel from colliding with the surface of the composite material tee fitting. When the contact force detected by the force sensor exceeds the threshold, the grinding robot drives the grinding device containing the force sensor to separate from the surface of the composite material tee fitting to avoid collision.

[0041] This invention provides a vision-based positioning system for grinding composite material tee fittings, offering the following advantages: By using machine vision to detect the pipe opening position of the composite material tee fitting, the center point and orientation of the tee fitting can be calculated quickly and accurately, thus achieving precise positioning of the workpiece for grinding. Compared to manual grinding, this method provides better and more stable grinding results, effectively improving grinding efficiency and quality. Furthermore, this method can automatically adjust grinding parameters according to the specific shape and size of the tee fitting, further enhancing grinding efficiency and quality. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the grinding equipment structure in this invention;

[0043] In the diagram: 1-spindle, 2-composite material tee fitting, 3-grinding device including force sensor, 4-depth camera, 5-grinding robot, 6-robot base, 7-host computer.

[0044] Figure 2 This is a flowchart of the polishing system in this invention.

[0045] Figure 3 This is a schematic diagram showing the division of composite material tee fittings in this invention.

[0046] In the diagram: 8-Main pipe section 1, 9-Intersection, 10-Branch pipe section, 11-Main pipe section 2.

[0047] Figure 4 This is a schematic diagram of the grinding trajectory angle in this invention.

[0048] Figure 5 This is a schematic diagram of the grinding trajectory in this invention;

[0049] In the diagram: 12 - starting position for grinding branch pipe section, 13 - starting position for grinding intersection, 14 - starting position for grinding main pipe section 1, 15 - starting position for grinding main pipe section 2.

[0050] Figure 6 In this invention Figure 6 Schematic diagram of pose recognition principle. Detailed Implementation

[0051] The process flow of the present invention will now be described in detail with reference to the accompanying drawings. Many technical details are provided to enable the reader to better understand the present invention; however, even without these technical details, those skilled in the art can implement the technical solutions protected by the claims of the present invention, including variations or modifications based on these details.

[0052] Figure 1 This embodiment discloses a visual positioning-based composite material tee pipe fitting grinding system, which includes a schematic diagram of the grinding equipment structure, comprising:

[0053] The components include a rotatable spindle 1, a grinding device 3 with a force sensor, a depth camera 4, a grinding robot 5, a robot base 6, and a host computer 7. The spindle 1 is used to fix the composite material tee fitting 2 and can drive it to rotate.

[0054] Figure 2 This is a flowchart of a visual positioning-based composite material tee pipe fitting grinding system disclosed in this embodiment. This embodiment uses a composite material tee pipe fitting with a main pipe section 1 that is 90mm long, 70mm in diameter, and 10mm thick, a main pipe section 2 that is 75mm long, 70mm in diameter, and 10mm thick, a branch pipe section that is 85mm long, 68mm in diameter, and 12mm thick, a length of 100mm from the right dividing line of the main pipe section 1 to the left dividing line of the main pipe section 2, and a length of 95mm from the dividing line of the branch pipe section to the bottom of the composite material tee pipe fitting as an example.

[0055] S1: Establish the pose relationship between the depth camera coordinate system and the grinding device using the checkerboard calibration method. First, firmly assemble the grinding robot and the depth camera, place the checkerboard pattern within the depth camera's field of view, adjust the pose of the grinding robot's end effector, and acquire complete checkerboard image data exposed to the camera's field of view in different poses. Then, use the formula:

[0056]

[0057] The pose relationship between the camera and the grinding device is obtained:

[0058]

[0059] S2: The host computer sends a grinding start command to the grinding robot. The grinding robot automatically moves to the position where the composite material tee fitting is located in the center of the image captured by the depth camera, accurately identifying the center point of the pipe opening. It then sends a ready signal to the host computer. Figure 6 The coordinates of the end of the grinding robot, with the base as the coordinate system, are (0, 300, 280).

[0060] S3: The host computer sends a visual recognition command to the depth camera. The depth camera acquires images of the composite material tee fitting fixed on the spindle and sends the image information back to the host computer.

[0061] S4: The host computer uses the YOLOv8-CGF visual recognition algorithm to identify the center point of the composite material tee fitting. The coordinates of the center point of the main pipe section 1 are (-140, 600, 280), the coordinates of the center point of the main pipe section 2 are (125, 600, 280), and the coordinates of the center point of the branch pipe section are (0, 716.9, 212.5). The calculated center point coordinates of the composite material tee fitting are (0, 600, 280), and the attitude matrix is:

[0062]

[0063] The spindle drives the composite material tee fitting to rotate from its current posture to a posture where the Z-axis of the target coordinate system is parallel to the Z-axis of the end-effector coordinate system of the grinding robot, according to the formula:

[0064]

[0065]

[0066] Where, m i (i = 1, 2, 3, 4) are undetermined coefficients, r i The radius of the latitude circle, ρ is the compressibility coefficient of the fiber-reinforced layer unit, n is the number of fiber layers in the unit, and ρ a V is the unit density. f ρ represents the fiber volume fraction, and ρ represents the fiber density. The updated tee fitting model, combined with a grinding strategy to plan the grinding trajectory, divides the composite material tee fitting into four parts: main pipe section 1, intersection section, branch pipe section, and main pipe section 2.

[0067] The main pipe sections 1 and 2 employ the first grinding strategy: a grinding robot drives a grinding device containing a force sensor to begin horizontal movement from the grinding start position of the main pipe section 1 of the composite material tee fitting, while the spindle rotates the composite material tee fitting. The grinding device stops at the junction of the main pipe section 1 and the intersection, then reverses direction and moves horizontally until it reaches the grinding start position of the main pipe section 1 again. The ratio k of the horizontal movement speed to the spindle rotation speed satisfies:

[0068] k=2πrcotα

[0069] Where r represents the cross-sectional radius of the composite material tee fitting, and a is as follows: Figure 4 The angle shown represents the angle between the polishing trajectory and the horizontal line in a counterclockwise direction, where 0° is the angle between the two sides. <a<90°。

[0070] The branch pipe section adopts the second grinding strategy, that is, the spindle drives the composite material tee fitting to rotate, and the grinding robot drives the grinding device containing force sensors to move horizontally from left to right starting from the grinding start position of the branch pipe section. The motion trajectory coordinates of the grinding robot satisfy the following formula:

[0071]

[0072] Among them, θ, θ′, It is a function of the cross-sectional radius r of the branch pipe of the composite material tee fitting.

[0073] The intersection section employs a third grinding strategy: the spindle drives the composite material tee fitting to rotate, and the grinding robot, carrying a grinding device containing a force sensor, begins grinding from the starting position of the intersection. The grinding robot avoids collisions between the grinding device and the branch pipe section by moving to the arc transition part of the composite material tee fitting. The ratio K of the horizontal movement speed to the spindle rotation speed satisfies:

[0074] K = 35.36sin(0.01σ - 1.6) + 4.78

[0075] Where σ represents the rotation angle of the spindle.

[0076] S5: The grinding robot grinds the composite material tee fittings according to the grinding trajectory planned by the host computer. The trajectory is as follows: Figure 5 As shown, after polishing is completed, a polishing completion signal is sent to the host computer and the device returns to the initial position.

[0077] The above embodiments are only used to further illustrate the visual positioning-based composite material tee pipe grinding system of the present invention. However, the present invention is not limited to the embodiments. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. A vision-based composite material tee pipe fitting grinding system, characterized in that, The grinding system includes: a rotatable spindle (1), a grinding device (3) with a force sensor, a depth camera (4), a grinding robot (5), a robot base (6), and a host computer (7). The system comprises the following steps: establishing the pose relationship between the depth camera (4) and the grinding device (3); the host computer (7) sending a grinding start command to the grinding robot (5); the grinding robot (5) automatically finding the optimal pose for image acquisition; the depth camera (4) acquiring images of the composite material tee fitting (2) fixed on the spindle (1) and sending the image information back to the host computer (7); and the host computer (7) using YOLO... The v8-CGF visual recognition algorithm determines the coordinates of the center point of the pipe opening of the composite material tee fitting (2), calculates the current posture of the composite material tee fitting (2), the spindle (1) drives the composite material tee fitting (2) to rotate from the current posture to the preset posture, updates the model according to the winding information of the composite material tee fitting, determines the initial position of grinding and plans the grinding trajectory in combination with the grinding method, the host computer (7) controls the grinding robot (5) to drive the grinding device (3) with force sensor, and grinds each part of the composite material tee fitting (2) fixed on the rotatable spindle (1) according to the preset grinding method; The YOLOv8-CGF visual recognition algorithm is as follows: a. Add the CoorAtt structure attention mechanism to the YOLOv8 network model, decompose the channel attention into two 1D feature encoding processes, aggregate the features of the composite material tee fitting feature map along two spatial directions respectively, and encode the resulting feature map into a pair of direction-aware and position-sensitive attention maps respectively, which are applied to the input feature map to enhance the detection accuracy of the composite material tee fitting port; b. Modify the Neck part in the YOLOv8 network model to a Gather-and-Distribute structure. This is achieved by modifying the original YOLOv8 algorithm's Neck part into a low-order clustering and distribution branch and a high-order clustering and distribution branch. The high-order clustering and distribution branch is dedicated to the identification of composite tee fittings at long distances, while the low-order clustering and distribution branch is dedicated to the localization of the center point of the pipe opening of composite tee fittings at close distances. c. Focaler-IoU is used as the loss function of the YOLOv8-CGF algorithm. When IoU is less than a lower threshold, the loss is 0. When IoU is greater than an upper threshold, the loss is 1. When IoU is between the lower and upper thresholds, the loss is a function that increases linearly with the IoU value.

2. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The pose relationship between the depth camera (4) and the polishing device (3) is obtained by the following formula: ； Where O represents the calibration plate coordinate system, E i E j C represents different orientations in the coordinate system of the grinding device, which includes the force sensor. i C j Representative and E i E j The corresponding pose is the pose of the depth camera, where B represents the base coordinates of the grinding robot, and a matrix of the form is used to represent the relative pose relationship between the Y coordinate system and the X coordinate system.

3. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The method for the grinding robot (5) to automatically find the optimal pose for image acquisition is as follows: the depth camera (4) installed on the grinding robot (5) acquires images of the composite material tee fitting (2), and the grinding robot (5) continuously adjusts its pose according to the image information until the composite material tee fitting (2) is located at the center of the image captured by the depth camera (4).

4. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The steps for identifying the coordinates of the center point of the composite material tee fitting nozzle and calculating its current orientation using YOLOv8-CGF are as follows: a. Using the created composite material tee fitting dataset, the YOLOv8-CGF network model was trained to obtain the training weights for detecting the center point of the pipe opening of the composite material tee fitting. 2b. After the grinding robot (5) carrying the depth camera (4) moves to the optimal pose for image acquisition, the depth camera (4) acquires color image information and depth image information of the composite material tee fitting and sends them to the YOLOv8-CGF network model loaded with training weights. After extracting image features from the acquired color image through the backbone of the model, it is sent to the Gather-and-Distribute structure. c. During detection, the attention mechanism of the CoorAtt structure is applied, enabling the model to adaptively learn different regions of the image. By improving the localization accuracy of the detection box, the detection accuracy of the composite material tee fitting opening is improved. The Focaler-IoU is used as the loss function of the YOLOv8-CGF algorithm at the output end to further improve the localization accuracy of the detection box. d. The formula for calculating the center point coordinates of composite material tee fittings is: ； Where x1 is the X-axis coordinate of the center point of the main pipe section 1, l1 is the length of the main pipe section 1, l is the length of the intersection, y1 is the Y-axis coordinate of the center point of the main pipe section 1, y2 is the Y-axis coordinate of the center point of the main pipe section 2, z1 is the Z-axis coordinate of the center point of the main pipe section 1, and z2 is the Z-axis coordinate of the center point of the main pipe section 2. e. Take the center point of the main pipe section as the origin O of the target coordinate system, take the straight line passing through the center point of the main pipe section and the center point of the identified branch pipe opening as the Y-axis of the target coordinate system, with the positive direction being from the center point of the main pipe section to the center point of the branch pipe opening, take the main axis as the X-axis of the target coordinate system, with the positive direction being from the main pipe section 1 (8) to the main pipe section 2 (11), and the Z-axis is determined by the right-hand rule. After constructing the target coordinate system, the target coordinate system is transformed into the end coordinate system of the grinding robot through coordinate transformation. The transformation matrix is ​​the posture of the composite material tee fitting.

5. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The preset posture is a posture in which the Z-axis of the target coordinate system is parallel to the Z-axis of the end coordinate system of the grinding robot.

6. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The thickness update calculation formula for the model updated based on the winding information of composite material tee fittings is as follows: ; ; Where, m i (i = 1, 2, 3, 4) are undetermined coefficients, r i Let ζ be the radius of the latitude circle. i ρ is the compressibility coefficient of the fiber-reinforced layer unit, n is the number of fiber layers in the unit, and ρ a V is the unit density. f ρ is the fiber volume fraction, and ρ is the fiber density.

7. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The grinding method involves dividing the composite material tee fitting into four parts: main pipe section 1 (8), intersection section (9), branch pipe section (10), and main pipe section 2 (11). The main pipe section 1 (8) and main pipe section 2 (11) are ground using the first grinding method, the branch pipe section (10) is ground using the second grinding method, and the intersection section (9) is ground using the third grinding method. The grinding order and number of times for the four parts of the composite material tee fitting can be changed.

8. The composite material tee pipe fitting grinding system based on vision positioning according to claim 1, characterized in that, The grinding start position (12) of the branch pipe section (10), the grinding start position (14) of the main pipe section 1 (8), and the grinding start position (15) of the main pipe section 2 (11) are the intersection points of the target coordinate system XOY plane with the pipe opening of the branch pipe section (10), the target coordinate system XOY plane with the pipe opening of the main pipe section 1 (8), and the target coordinate system XOY plane with the pipe opening of the main pipe section 2 (11), respectively. The grinding start position (13) of the intersection (9) is the intersection point of the cross section at the boundary between the intersection (9) and the main pipe section 1 (8) and the target coordinate system XOY plane.

9. A visual positioning-based composite material tee pipe polishing system according to claim 7, characterized in that, The first grinding method involves a grinding robot (5) driving a grinding device (3) containing a force sensor to move horizontally from the grinding start position (14) of the main pipe section 1 (8) of the composite material tee fitting (2). At the same time, the spindle (1) drives the composite material tee fitting (2) to rotate. When the grinding device (3) moves to the junction of the main pipe section 1 (8) and the intersection (9), it stops and then moves horizontally in the opposite direction. When it moves to the grinding start position (14) of the main pipe section 1 of the composite material tee fitting (2), it stops, completing one reciprocating motion. The above process is repeated until the main pipe section 1 (8) is ground. The grinding process of the main pipe section 2 (11) is similar to that of the main pipe section 1 (8).

10. A visual positioning-based composite material tee pipe polishing system according to claim 9, characterized in that, In the first grinding method, the grinding robot (5) drives the grinding device (3) with a force sensor. The ratio k of the horizontal movement speed and the rotational movement speed of the spindle from the starting position (14) satisfies: k = 2πrcotα; Where r represents the cross-sectional radius of the composite material tee fitting, and a represents the angle between the grinding trajectory and the horizontal line in the counterclockwise direction, with 0°. <a<90°。 11. A visual positioning-based composite material tee pipe polishing system according to claim 7, characterized in that, The second grinding method uses the main shaft (1) to drive the composite material tee fitting (2) to rotate. The grinding robot (5) drives the grinding device (3) with force sensor to start horizontal reciprocating motion from left to right from the grinding starting position (12) of the branch pipe section (10). The motion trajectory coordinates of the grinding robot (5) satisfy the following formula: ; Among them, θ, θ′, It is a function of the cross-sectional radius r of the branch pipe of the composite material tee fitting.

12. The composite material tee pipe fitting grinding system based on vision positioning according to claim 7, characterized in that, The third grinding method uses the main shaft (1) to drive the composite material tee fitting (2) to rotate. The grinding robot (5) drives the grinding device (3) containing the force sensor to move horizontally from the grinding start position (13) of the intersection (9). The grinding robot (5) avoids the grinding device (3) from colliding with the branch pipe section (10) by moving to the arc transition part of the composite material tee fitting (2). After grinding is completed, the grinding robot (5) drives the grinding device (3) containing the force sensor to return to the initial position.

13. The composite material tee pipe fitting grinding system based on vision positioning according to claim 12, characterized in that, In the third polishing method, the polishing robot (5) drives the polishing device (3) containing a force sensor. The ratio K of the horizontal movement speed and the spindle rotation speed from the starting position (13) satisfies: K=35.36sin(0.01σ-1.6)+4.78; Where σ represents the rotation angle of the spindle.

Images