A kind of rotating friction welding spindle support device

Abstract

The invention discloses a rotating friction welding spindle supporting device, which relates to the technical field of friction welding, comprising a spindle supporting seat, a supporting bearing being installed inside the spindle supporting seat, a control module being loaded on the spindle supporting seat, a magnetic control mechanism being installed at one end of the spindle supporting seat, a hydraulic cylinder being arranged between the magnetic control mechanism and the spindle supporting seat, a spindle body being inserted inside the supporting bearing, the spindle body comprising a fixed shaft and a movable shaft, and a magnetic ring being nested on the side of the movable shaft; the fixed shaft will not be subjected to axial force during high-speed rotation, so that the load of the supporting bearing is greatly reduced during operation, thus prolonging the service life of the supporting bearing, and the magnetic control mechanism will not contact the magnetic ring when the magnetic repulsive force is applied to the magnetic ring by the magnetic control mechanism during the process of the magnetic repulsion force being applied to the magnetic ring, so that when the movable shaft drives the magnetic ring to rotate at high speed, the magnetic control mechanism will not affect the rotation of the magnetic ring, which will greatly reduce the energy consumption of the spindle body for friction welding.

Description

Technical Field

[0001] The invention relates to the technical field of friction welding, and in particular to a rotary friction welding spindle supporting device. Background Art

[0002] Rotary friction welding connects two workpieces together through rotation and friction. This technology uses frictional heat to plasticize the contact surface without using filler materials, and then achieves welding through axial pressure. Therefore, rotary friction welding requires a spindle that can rotate at high speed. The spindle support device is a device that stabilizes the spindle through bearings;

[0003] Existing rotary friction welding spindle support devices, such as the friction welding electric spindle of the Chinese patent application with publication number CN103084730A and the stir friction welding seam welding and spot welding integrated spindle of the Chinese patent application with publication number CN105598574A, need to apply an axial force to the weldment when the spindle is working. This force will react on the support bearing of the spindle, causing the support bearing to greatly increase the workload and reduce the life of the bearing. Summary of the invention

[0004] In order to overcome the above-mentioned technical problems, the purpose of the present invention is to provide a rotary friction welding spindle support device to solve the problem in the prior art that when the spindle is working, an axial force needs to be applied to the weldment, and this force will react on the support bearing of the spindle, thereby greatly increasing the workload of the support bearing and shortening the service life of the support bearing.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] Specifically, a rotary friction welding spindle support device is provided, including a spindle support seat, a support bearing is installed inside the spindle support seat, a control module is loaded on the spindle support seat, a magnetic control mechanism is installed at one end of the spindle support seat, a hydraulic cylinder is arranged between the magnetic control mechanism and the spindle support seat, a spindle body is inserted on the inner side of the support bearing, the spindle body includes a fixed axis and a movable axis, a magnetic ring is nested on the side of the movable axis, the control module collects a historical training data set in advance, and based on the historical training data set, trains a machine learning model for predicting the magnetic repulsion force value applied by the magnetic control mechanism to the magnetic ring, the control module generates a fusion distance change curve of the weldment based on the diameter of the weldment, the rotation speed value of the spindle body and the change of the magnetic repulsion force value applied by the magnetic control mechanism to the magnetic ring, the control module collects the diameter of the weldment, the rotation speed value of the spindle body and the fusion distance of the weldment in real time, and uses the machine learning model to predict the magnetic repulsion force value applied by the magnetic control mechanism to the magnetic ring based on the diameter of the weldment, the rotation speed value of the spindle body and the fusion distance of the weldment.

[0007] As a further solution of the present invention: the historical training data set includes N groups of training data, N is a positive integer, each group of training data includes feature data and label data, the feature data includes the diameter of the weldment, the rotation speed value of the spindle body and the fusion distance of the weldment;

[0008] The diameter of the weldment is the friction diameter of the weldment when each set of training data is collected;

[0009] The rotation speed value of the spindle body is the rotation speed value of the spindle body driving the weldment when collecting each set of training data;

[0010] The fusion distance of the weldment is the difference between the total length of the weldment before welding and the total length of the weldment after welding when collecting each set of training data;

[0011] The label data is the magnetic repulsion value applied by the magnetic control mechanism to the magnetic ring when collecting each set of training data.

[0012] As a further solution of the present invention: the control module trains a machine learning model for predicting the magnetic repulsive force value applied by the magnetic control mechanism to the magnetic ring as follows:

[0013] The feature data in each set of training data is converted into a feature vector, and the feature vector is used as the input of the machine learning model. The magnetic repulsion value predicted by the machine learning model for the feature data is used as the output, and the magnetic repulsion value in the label data corresponding to the feature data is used as the prediction target. The training target is to minimize the sum of all prediction accuracies. The calculation formula for prediction accuracy is ui=(ai-bi) 2 , where ui is the prediction accuracy, ai is the predicted magnetic repulsion value corresponding to the feature data of the i-th group of training data, and bi is the magnetic repulsion value in the label data of the i-th group of training data. The machine learning model is trained until the sum of the prediction accuracies converges and the training is stopped.

[0014] As a further solution of the present invention: the machine learning model adopts a recurrent neural network model.

[0015] As a further solution of the present invention: a moving groove is provided on the side of one end of the fixed shaft close to the moving shaft, and a moving column is fixedly connected to one end of the moving shaft close to the fixed shaft, and the moving column matches the moving groove.

[0016] As a further solution of the present invention: the magnetic control mechanism includes an electromagnetic ring, on which an infrared sensor is installed, and the infrared sensor is used to collect the magnetic distance between the electromagnetic ring and the magnetic ring.

[0017] As a further solution of the present invention: the calculation method of the magnetic repulsion value applied by the magnetic control mechanism to the magnetic ring is:

[0018] F=μ×(m1+m2) / r2 ;

[0019] Among them, F is the magnetic repulsion value applied by the magnetic control mechanism to the magnetic ring, μ is the magnetic permeability, m1 is the magnetic moment of the magnetic control mechanism, m2 is the magnetic moment of the magnetic ring, and r is the magnetic spacing.

[0020] As a further solution of the present invention: the method for the control module to generate the fusion distance variation curve is:

[0021] S1: Prepare weldments X1, X2, ...X with the same diameter in advance n , X n represents the nth weldment;

[0022] S2: weld parts X1, X2...X n Friction welding is performed using the spindle body at the same rotation speed;

[0023] S3: Get weldment X1, X2...X n The corresponding magnetic repulsion values ​​applied by the magnetic control mechanism to the magnetic ring are f1, f2, ... f n , where f1>f2>...>f n ;f n Indicates the magnetic repulsion value corresponding to the nth weldment;

[0024] S4: Use computer to generate weldments X1, X2, ... X n The variation curve of the fusion distance d with the working time t of the spindle body l 1 、l 2…… l n , l n It represents the curve corresponding to the nth weldment changing with the working time t of the spindle body;

[0025] S5: Use the computer to l 1 、l 2…… l n The fusion distance variation curve is generated in the same rectangular coordinate system.

[0026] As a further solution of the present invention: the control module uses an image collector to collect weldments X1, X2...X n After the high-definition pictures P1, P2...Pn are welded at a fusion distance d, the control module identifies the symmetry Sy1, Sy2...Sy of the high-definition pictures P1, P2...Pn based on the high-definition pictures P1, P2...Pn. n , select the smallest symmetry Sy minThe corresponding magnetic repulsion value is used as the magnetic repulsion value of the label data, Pn represents the high-definition picture of the nth weldment after welding, Sy n Indicates the degree of symmetry corresponding to the nth weldment.

[0027] As a further solution of the present invention: the control module identifies the symmetry of the high-definition pictures based on the high-definition pictures P1, P2...Pn in the following manner:

[0028] Y1: Convert high-definition images into grayscale images;

[0029] Y2: Split the grayscale image into 2m based on the size of the grayscale image 2 pixel areas, m is the number of pixel areas on the side length of the grayscale image;

[0030] Y3: Take the central axis of the weldment in the grayscale image as the axis of symmetry and divide the grayscale image into two parts, each of which occupies m 2 pixel areas, number the pixel areas in the two parts, and calculate the grayscale difference of the pixel area at each corresponding position:

[0031] ;

[0032] Where I (i, j) and I' (i, j) are the grayscale values ​​of the corresponding positions on both sides of the symmetry axis, and i and j are the number values ​​of the pixel area.

[0033] Beneficial effects of the present invention:

[0034] 1. In the present invention, the magnetic control mechanism can apply magnetic repulsion to the magnetic ring, that is, the magnetic ring will be subjected to a force away from one end of the spindle support seat. When the spindle body performs friction welding on the weldment, the spindle body needs to apply a certain extrusion force to the weldment to ensure that the weldment generates heat by friction during high-speed rotation. At this time, the movable shaft applies an axial force to the weldment, while the fixed shaft does not apply an axial force to the weldment. Therefore, the fixed shaft will not be subjected to axial force during high-speed rotation, and the support bearing installed with the fixed shaft will not be subjected to axial force either, so that the load of the support bearing is greatly reduced during operation, thereby extending the service life of the support bearing. When the magnetic control mechanism applies magnetic repulsion to the magnetic ring, the magnetic control mechanism will not contact the magnetic ring. Therefore, when the movable shaft drives the magnetic ring to rotate at high speed, the magnetic control mechanism will not affect the rotation of the magnetic ring, which will greatly reduce the energy consumption of the spindle body for friction welding.

[0035] 2. In the present invention, the machine learning model trained by the control module predicts the magnetic repulsion force value applied by the magnetic control mechanism to the magnetic ring, which can ensure full automatic control during friction welding of the spindle body and ensure the accuracy of the force applied by the moving axis to the weldment, greatly improving the quality of friction welding of the weldment and further reducing the scrap rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention will be further described below in conjunction with the accompanying drawings:

[0037] Figure 1 It is a structural schematic diagram of the rotary friction welding spindle support device of the present invention;

[0038] Figure 2 is a front view of the rotary friction welding spindle support device of the present invention;

[0039] Figure 3 It is a structural schematic diagram of the main shaft body in the present invention;

[0040] Figure 4 It is a curve diagram of the change of the fusion distance in the present invention;

[0041] Figure 5 It is a flow chart of the method for the control module in the present invention to identify the symmetry of a high-definition picture.

[0042] Reference numerals:

[0043] 1. Spindle support seat; 11. Support bearing; 2. Magnetic control mechanism; 21. Hydraulic cylinder; 3. Spindle body; 31. Fixed axis; 32. Moving axis; 33. Moving groove; 34. Moving column; 35. Concentric column; 4. Magnetic ring. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] Example 1

[0046] like Figure 1 - Figure 4As shown, the present invention discloses a rotary friction welding spindle support device, including a spindle support seat 1, a support bearing 11 is installed inside the spindle support seat 1, a control module is loaded on the spindle support seat 1, it should be noted that a microprocessor can be installed on the spindle support seat 1, and the control module is loaded on the microprocessor. A magnetic control mechanism 2 is installed at one end of the spindle support seat 1, and a hydraulic cylinder 21 is arranged between the magnetic control mechanism 2 and the spindle support seat 1. The output end of the hydraulic cylinder 21 is connected to the magnetic control mechanism 2. When in use, the hydraulic cylinder 21 can be opened by the control module, and the output of the hydraulic cylinder 21 The end can push the magnetic control mechanism 2 through the hydraulic rod to adjust the position of the magnetic control mechanism 2. The main shaft body 3 is inserted into the inner side of the support bearing 11. One end of the main shaft body 3 is connected to the drive motor (not shown in the drawings of the specification) through a coupling. When the main shaft body 3 is used for friction welding of the weldment, the drive motor is turned on first. The output shaft of the drive motor can drive the main shaft body 3 through the coupling, and the main shaft body 3 can drive the weldment to rotate. It should also be noted that the rotation speed of the main shaft body 3 is usually very high, so the specifications of the drive motor and the drive motor and The transmission parts between the main shaft body 3 are adaptively selected by technicians in this field according to the specifications of the main shaft body 3. The main shaft body 3 includes a fixed shaft 31 and a movable shaft 32. The fixed shaft 31 is installed inside the main shaft support seat 1 through a support bearing 11. The axial position of the fixed shaft 31 along the main shaft support seat 1 needs to be fixed, while the axial position of the movable shaft 32 along the main shaft support seat 1 does not need to be fixed. A magnetic ring 4 is embedded on the side of the movable shaft 32. When in use, the magnetic control mechanism 2 can apply a magnetic repulsion force to the magnetic ring 4, that is, the magnetic ring 4 will be affected by the end away from the main shaft support seat 1. When the spindle body 3 performs friction welding on the weldment, the spindle body 3 needs to apply a certain extrusion force to the weldment to ensure that the weldment generates friction heat during high-speed rotation. At this time, the movable shaft 32 applies an axial force to the weldment, while the fixed shaft 31 does not apply an axial force to the weldment. Therefore, the fixed shaft 31 will not be subjected to axial force during high-speed rotation, and the support bearing 11 installed with the fixed shaft 31 will not be subjected to axial force, so that the load of the support bearing 11 is greatly reduced during operation, thereby extending the service life of the support bearing 11.

[0047] It should also be noted that during the process of the magnetic control mechanism 2 applying magnetic repulsion to the magnetic ring 4, the magnetic control mechanism 2 will not contact the magnetic ring 4. Therefore, when the movable shaft 32 drives the magnetic ring 4 to rotate at high speed, the magnetic control mechanism 2 will not affect the rotation of the magnetic ring 4, which will greatly reduce the energy consumption of the spindle body 3 for friction welding.

[0048] Example 2

[0049] like Figure 1 - Figure 5As shown, the control module collects a historical training data set in advance, and the historical training data set includes N groups of training data, where N is a positive integer, and each group of training data includes feature data and label data, and the feature data includes the diameter of the weldment, the rotation speed value of the spindle body 3, and the fusion distance of the weldment;

[0050] The diameter of the weldment is the friction diameter of the weldment when collecting each set of training data. It should be noted that the weldment is divided into two parts, the fixed part and the rotating part, and the friction diameter of the weldment refers to the diameter of the rotating part of the weldment. Usually, the diameter of the fixed part and the diameter of the rotating part of the weldment are the same;

[0051] The rotation speed value of the spindle body 3 is the rotation speed value of the weldment driven by the spindle body 3 when collecting each set of training data. It should be noted that the rotation speed of the spindle body 3 is adaptively adjusted by designers in this field according to the material and specifications of the weldment. For example, when the melting point of the weldment is high, a higher rotation speed is usually required. When the friction diameter of the weldment is large, a higher rotation speed is usually required. Conversely, a lower rotation speed is required. For example, the rotation speed of weldments made of aluminum alloy is usually between 1000 and 3000 rpm, and the rotation speed of weldments made of stainless steel is usually between 600 and 1800 rpm. The specific situation is adjusted by technicians in this field according to the actual situation on site;

[0052] The fusion distance of the weldment is the difference between the total length of the weldment before welding and the total length of the weldment after welding when collecting each set of training data. It should be noted that when performing friction welding of weldments, if the total length of the fixed part and the rotating part of the weldment before welding is L 前 , and the required weldment length is L 需 , then the fusion distance should be less than or equal to L 前 -L 需 , to ensure that the weldment has redundant length after welding, which is specifically selected by technical personnel in the field according to the conditions of the weldment;

[0053] The label data is the magnetic repulsive force value applied by the magnetic control mechanism 2 to the magnetic ring 4 when collecting each set of training data.

[0054] Based on the historical training data set, a machine learning model for predicting the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4 is trained. The machine learning model adopts a recurrent neural network model. The control module trains a machine learning model for predicting the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4 in the following method:

[0055] The feature data in each set of training data is converted into a feature vector, and the feature vector is used as the input of the machine learning model. The magnetic repulsion value predicted by the machine learning model for the feature data is used as the output, and the magnetic repulsion value in the label data corresponding to the feature data is used as the prediction target. The training target is to minimize the sum of all prediction accuracies. The calculation formula for prediction accuracy is ui=(ai-bi) 2 , where ui is the prediction accuracy, ai is the predicted magnetic repulsion value corresponding to the feature data of the i-th group of training data, and bi is the magnetic repulsion value in the label data of the i-th group of training data. The machine learning model is trained until the sum of the prediction accuracies converges. It should be noted that the converged label is set by a technician in this field according to the specific model training situation.

[0056] The control module collects the diameter of the weld, the rotation speed of the spindle body 3 and the fusion distance of the weld in real time, and uses a machine learning model to predict the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4 based on the diameter of the weld, the rotation speed of the spindle body 3 and the fusion distance of the weld. It should be noted that the machine learning model trained by the control module predicts the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4, which can ensure that the spindle body 3 is automatically controlled throughout the friction welding process, and ensures the accuracy of the force applied by the moving shaft 32 to the weld, which greatly improves the quality of friction welding of the weld and further reduces the scrap rate.

[0057] Example 3

[0058] like Figure 1 - Figure 3 As shown, a moving groove 33 is provided on the side surface of one end of the fixed shaft 31 close to the moving shaft 32, and a moving column 34 is fixedly connected to the end of the moving shaft 32 close to the fixed shaft 31. The moving column 34 matches the moving groove 33, and the moving column 34 can slide freely in the moving groove 33 along the groove wall direction of the moving groove 33. A concentric column 35 is fixedly connected to the center position of one end of the moving shaft 32 close to the fixed shaft 31, and a concentric groove matching the concentric column 35 is provided on the end face of the fixed shaft 31 close to the concentric column 35. During the movement of the moving shaft 32, the cooperation between the concentric column 35 and the concentric groove can improve the rotation stability of the moving shaft 32.

[0059] The magnetic control mechanism 2 includes an electromagnetic ring, on which an infrared sensor is installed. The infrared sensor is used to collect the magnetic distance between the electromagnetic ring and the magnetic ring 4. It should be noted that a power supply is arranged on the spindle support seat 1, and the power supply can provide power to the electromagnetic ring. The circuit switch between the power supply and the electromagnetic ring is controlled by the control module, that is, the control module can open or close the circuit on the electromagnetic ring through the power supply. When the circuit on the electromagnetic ring is opened, the electromagnetic ring will generate a magnetic field. When the circuit on the electromagnetic ring is closed, the electromagnetic ring will not generate a magnetic field.

[0060] The calculation method of the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4 is:

[0061] F=μ×(m1+m2) / r 2 ;

[0062] Wherein, F is the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4, μ is the magnetic permeability, m1 is the magnetic moment of the magnetic control mechanism 2, m2 is the magnetic moment of the magnetic ring 4, and r is the magnetic spacing. It should be noted that the magnetic moment of the magnetic control mechanism 2 is controlled by the specifications of the electromagnetic ring and the current of the power supply, and the magnetic moment of the magnetic ring 4 can be fixed. For example, the magnetic moment of the magnetic ring 4 made of a permanent magnet is measured by a person skilled in the art, according to the formula "F=μ×(m1+m2) / r 2 It can be seen that the magnitude of the magnetic repulsive force exerted by the magnetic control mechanism 2 on the magnetic ring 4 is inversely proportional to the square of the distance between the magnetic control mechanism 2 and the magnetic ring 4, so the distance between the magnetic control mechanism 2 and the magnetic ring 4 needs to be precisely controlled.

[0063] Example 4

[0064] like Figure 1 - Figure 4 As shown, the control module generates a fusion distance variation curve of the weldment based on the diameter of the weldment, the rotation speed value of the spindle body 3 and the change of the magnetic repulsion value applied by the magnetic control mechanism 2 to the magnetic ring 4. The method for the control module to generate the fusion distance variation curve is:

[0065] S1: Prepare weldments X1, X2, ...X with the same diameter in advance n ,It should be noted that the material of the weldment also needs to be the same.

[0066] S2: weld parts X1, X2...X n The main spindle body 3 is used to perform friction welding at the same rotation speed. It should be noted that, first, the rotating part of the weldment is clamped by the movable shaft 32 on the main spindle body 3, and the fixed part of the weldment is clamped by other clamps to ensure that the rotating part of the weldment and the fixed part of the weldment are on the same horizontal axis. Then, the drive motor is turned on, and the output shaft of the drive motor drives the fixed shaft 31 to rotate through the coupling. The rotating fixed shaft 31 drives the movable shaft 32 to rotate synchronously. The rotating movable shaft 32 drives the rotating part of the weldment to rotate at a high speed, generating friction with the fixed part of the weldment, thereby generating frictional heat.

[0067] S3: weldment X1, X2...X n The corresponding magnetic repulsive force values ​​applied by the magnetic control mechanism 2 to the magnetic ring 4 are f1, f2, ..., f n , where f1>f2>...>f n, where f1 > f2 >... > f n , f n represents the magnetic repulsive force value corresponding to the nth welded part. It should be noted that the magnitudes of f1, f2... f n are preset within a certain range by those skilled in the art according to the material and diameter of the welded parts, and then n values are evenly selected within this range. For example, 4 values can be selected, that is, f1 > f2 > f3 > f4, and the corresponding number of welded parts is also 4;

[0068] S4: Use a computer to generate the variation curve of the fusion distance d of the welded parts X1, X2, X 3、 X4 with respect to the working time t of the main shaft body 3 l 1、 l 2、 l 3、 l 4;

[0069] S5: Use a computer to l 1 、l 2... l n Generate in the same rectangular coordinate system to form a fusion distance variation curve. It should be noted that the fusion distance variation curve is as Figure 4 shown, Figure 4 In the coordinate system, the X-axis is the time length, the Y-axis is the fusion distance length, t1, t2, t3, t4 represent different working times, and the time lengths increase in sequence, that is, t1 < t2 < t3 < t4. The greater the magnetic repulsive force value applied by the magnetic force control mechanism 2 to the magnetic force ring 4, the faster the fusion distance changes. Conversely, the smaller the magnetic repulsive force value applied by the magnetic force control mechanism 2 to the magnetic force ring 4, the slower the fusion distance changes.

[0070] Embodiment 5

[0071] The control module uses an image collector to collect the high-definition pictures P1, P2... Pn of the welded parts X1, X2... X n after welding with the fusion distance d. The control module identifies the symmetry degrees Sy1, Sy2... Sy of the high-definition pictures P1, P2... Pn based on the high-definition pictures P1, P2... Pn n , and selects the smallest symmetry degree Sy min corresponding magnetic repulsive force value as the magnetic repulsive force value of the labeled data. Pn represents the high-definition picture of the nth welded part after welding; Sy nIndicates the degree of symmetry corresponding to the nth weldment. It should be noted that the image acquisition device can be a high-definition camera, that is, the technical personnel in this field installs a high-definition camera that can move along the axial direction of the spindle support seat 1 above the spindle support seat 1, and after the spindle body 3 completes the friction welding of the weldment, the high-definition camera is moved to the top of the welding surface of the weldment to collect its high-definition picture.

[0072] like Figure 5 As shown, the control module identifies the symmetry of the high-definition pictures based on the high-definition pictures P1, P2...Pn in the following manner:

[0073] Y1: Convert the HD image to a grayscale image. It should be noted that the grayscale image divides the color of the HD image into 256 chromaticities, that is, [0,255], where 0 represents all black and 255 represents all white.

[0074] Y2: Split the grayscale image into 2m based on the size of the grayscale image 2 pixel area, m is the number of pixel areas on the side length of the grayscale image. If the pixel area contains one chromaticity, the grayscale value of the pixel area is equal to the chromaticity. If the pixel area contains multiple chromaticities, the grayscale value of the pixel area is calculated as follows:

[0075] I=S1×h1+S2×h2……S n ×h n ;

[0076] Where I is the gray value, S1 is the percentage of the pixel area occupied by the chromaticity h1 corresponding to the first weldment, S2 is the percentage of the pixel area occupied by the chromaticity h2 corresponding to the second weldment, and S n is the chromaticity h corresponding to the nth weldment n The percentage of pixel area.

[0077] Y3: Take the central axis of the weldment in the grayscale image as the axis of symmetry and divide the grayscale image into two parts, each of which occupies m 2 Pixel regions, the pixel regions in the two parts are numbered. For example, when both parts contain 4 pixel regions, the pixel regions in the two parts are numbered (1, 1), (1, 2), (2, 1), (2, 2), and so on for multiple pixel regions. The grayscale difference of the pixel region at each corresponding position is calculated:

[0078] ;

[0079] Among them, I (i, j) and I' (i, j) are the grayscale values ​​of the corresponding positions on both sides of the symmetry axis, and i and j are the number values ​​of the pixel area. It should be noted that the smaller the symmetry of the grayscale image, the better the symmetry of the weldment corresponding to the grayscale image, and the better the welding effect of the weldment. Therefore, the magnetic repulsion value corresponding to the grayscale image with the best symmetry is selected as the magnetic repulsion value of the label data. This can ensure that after the machine learning model is trained, the effect of controlling the friction welding of the spindle body 3 on the weldment will be better.

[0080] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A rotary friction welding spindle support device, characterized in that: include: A main shaft support seat (1), a support bearing (11) is installed inside the main shaft support seat (1), and a control module is loaded on the main shaft support seat (1); A magnetic control mechanism (2) is mounted on one end of the spindle support seat (1), and a hydraulic cylinder (21) is provided between the magnetic control mechanism (2) and the spindle support seat (1); A main shaft body (3) is inserted into the inner side of the support bearing (11), the main shaft body (3) comprises a fixed shaft (31) and a movable shaft (32), a movable groove (33) is provided on the side surface of one end of the fixed shaft (31) close to the movable shaft (32), a movable column (34) is fixedly connected to one end of the movable shaft (32) close to the fixed shaft (31), the movable column (34) matches the movable groove (33), the movable column (34) can slide freely in the movable groove (33) along the groove wall direction of the movable groove (33), a concentric column (35) is fixedly connected to the center position of one end of the movable shaft (32) close to the fixed shaft (31), and a concentric groove matching the concentric column (35) is provided on the end surface of the fixed shaft (31) close to the concentric column (35); A magnetic ring (4) which is nested on the side of the moving shaft (32); The control module collects a historical training data set in advance, and trains a machine learning model for predicting the magnetic repulsive force value applied by the magnetic control mechanism (2) to the magnetic ring (4) based on the historical training data set; The control module generates a fusion distance variation curve of the weldment based on the diameter of the weldment, the rotation speed of the spindle body (3), and the change in the magnetic repulsive force value applied by the magnetic control mechanism (2) to the magnetic ring (4); The control module collects the diameter of the weldment, the rotation speed of the spindle body (3) and the fusion distance of the weldment in real time, and uses a machine learning model to predict the magnetic repulsion force value applied by the magnetic control mechanism (2) to the magnetic ring (4) based on the diameter of the weldment, the rotation speed of the spindle body (3) and the fusion distance of the weldment; The historical training data set includes N groups of training data, where N is a positive integer, and each group of training data includes feature data and label data, and the feature data includes the diameter of the weldment, the rotation speed value of the spindle body (3), and the fusion distance of the weldment; The diameter of the weldment is the friction diameter of the weldment when each set of training data is collected; The rotation speed value of the spindle body (3) is the rotation speed value of the spindle body (3) driving the weldment when collecting each set of training data; The fusion distance of the weldment is the difference between the total length of the weldment before welding and the total length of the weldment after welding when collecting each set of training data; The label data is the value of the magnetic repulsive force applied by the magnetic control mechanism (2) to the magnetic ring (4) when collecting each set of training data; The control module trains a machine learning model for predicting the value of the magnetic repulsive force applied by the magnetic control mechanism (2) to the magnetic ring (4) in the following manner: The feature data in each set of training data is converted into a feature vector, and the feature vector is used as the input of the machine learning model. The magnetic repulsion value predicted by the machine learning model for the feature data is used as the output, and the magnetic repulsion value in the label data corresponding to the feature data is used as the prediction target. The training target is to minimize the sum of all prediction accuracies. The calculation formula for prediction accuracy is ui=(ai-bi) 2 , where ui is the prediction accuracy, ai is the predicted magnetic repulsion value corresponding to the feature data in the i-th group of training data, and bi is the magnetic repulsion value in the label data of the i-th group of training data. The machine learning model is trained until the sum of the prediction accuracies converges and the training is stopped.

2. A rotary friction welding spindle support device according to claim 1, characterized in that: The machine learning model adopts a recurrent neural network model.

3. A rotary friction welding spindle support device according to claim 1, characterized in that: The magnetic force control mechanism (2) comprises an electromagnetic ring, on which an infrared sensor is mounted, and the infrared sensor is used to collect the magnetic distance between the electromagnetic ring and the magnetic ring (4).

4. A rotary friction welding spindle support device according to claim 3, characterized in that: The calculation method of the magnetic repulsive force value applied by the magnetic control mechanism (2) to the magnetic ring (4) is: F=μ×(m1+m2) / r 2 ; Wherein, F is the magnetic repulsion value applied by the magnetic control mechanism (2) to the magnetic ring (4), μ is the magnetic permeability, m1 is the magnetic moment of the magnetic control mechanism (2), m2 is the magnetic moment of the magnetic ring (4), and r is the magnetic spacing.

5. The rotary friction welding spindle support device according to claim 1, characterized in that: The method for the control module to generate the fusion distance variation curve is: S1: Prepare weldments X1, X2, ...X with the same diameter in advance n , X n represents the nth weldment; S2: weld parts X1, X2...X n The friction welding is performed using the spindle body (3) at the same rotation speed; S3: Get weldment X1, X2...X n The corresponding magnetic repulsive force values ​​applied by the magnetic control mechanism (2) to the magnetic ring (4) are f1, f2, ..., f n , where f1>f2>...>f n ;f n Indicates the magnetic repulsion value corresponding to the nth weldment; S4: Use computer to generate weldments X1, X2, ... X n The variation curve of the fusion distance d with the working time t of the spindle body (3) l 1 、l 2…… l n ; l n represents the curve corresponding to the nth weldment changing with the working time t of the spindle body (3); S5: Use the computer to l 1 、l 2…… l n The fusion distance variation curve is generated in the same rectangular coordinate system.

6. A rotary friction welding spindle support device according to claim 5, characterized in that: The control module uses an image collector to collect weldments X1, X2...X n After the high-definition pictures P1, P2...Pn are welded at a fusion distance d, the control module identifies the symmetry Sy1, Sy2...Sy of the high-definition pictures P1, P2...Pn based on the high-definition pictures P1, P2...Pn. n , select the smallest symmetry Sy min The corresponding magnetic repulsion value is used as the magnetic repulsion value of the label data; Pn represents the high-definition picture of the nth weldment after welding; Sy n Indicates the degree of symmetry corresponding to the nth weldment.

7. A rotary friction welding spindle support device according to claim 6, characterized in that: The control module identifies the symmetry of the high-definition pictures based on the high-definition pictures P1, P2...Pn in the following manner: Y1: Convert high-definition images into grayscale images; Y2: Split the grayscale image into 2m based on the size of the grayscale image 2 pixel area; m is the number of pixel areas on the side length of the grayscale image; Y3: Take the central axis of the weldment in the grayscale image as the axis of symmetry and divide the grayscale image into two parts, each of which occupies m 2 pixel areas, number the pixel areas in the two parts, and calculate the grayscale difference of the pixel areas at the corresponding positions on both sides of each symmetry axis: ； where I(i, j) and It is the grayscale value of the corresponding position on both sides of the symmetry axis, and i and j are the number values ​​of the pixel area.

Images