Quality detection device and method for automobile hub surface

Abstract

The invention relates to the technical field of automobile part detection, in particular to a quality detection device and method for the surface of an automobile hub, and the device comprises a conveying unit, an outer surface regional detection unit, an overturning clamping unit and an inner surface detection unit. The conveying unit is sequentially connected with the outer surface sub-area detection unit, the overturning clamping unit and the inner surface detection unit and is used for sequentially conveying hubs to all detection areas; the outer surface regional detection unit comprises an aluminum profile support, a first linear guide rail and a third linear guide rail, the overturning clamping unit comprises a second linear guide rail, a third servo motor and a first sliding block, and the inner surface detection unit comprises a positioning assembly and a camera support. The conveying unit is sequentially connected with the outer surface regional detection unit, the overturning clamping unit and the inner surface detection unit, full-process automatic detection of the hub from the outer surface to the inner surface is achieved, and the detection efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile parts detection, in particular to a quality detection device and method for automobile wheel hub surface. Background Art

[0002] As an important component of the automobile, the wheel hub will inevitably have defects inside or on the surface during the casting, heat treatment, grinding and polishing processes. Internal defects of the wheel hub can generally be detected by analyzing its X-ray image. For surface defects such as scratches and abrasions that affect the appearance of the product, a considerable part of the industry currently relies on traditional manual inspection methods to detect wheel hub defects.

[0003] Traditional manual inspection uses visual observation to identify defects. That is, workers are fixed at a workstation on the production line and use their hands to lift, tilt, rotate, and other methods to change the angle of the wheel hub to identify defects with the naked eye. The highly boring and repetitive work and high labor intensity will have a certain degree of impact on the workers' inspection quality and efficiency. Combined with the uncertainties of inspection and the operator's subjective awareness, level, and experience, long-term work will lead to misjudgment and omission of wheel hub defects, thereby affecting the factory quality of wheel hub products.

[0004] With the development of industrial automation, some companies have tried to introduce automated inspection equipment, but the existing technology still has defects. For example, the equipment disclosed in patent CN119594887A can only complete the inspection of the front and back spoke areas of the automobile wheel hub, and cannot reach key parts such as the inner side of the rim. Although another patent CN215114607U mentions the self-rotation detection of the wheel hub, it does not provide a reliable fixing method. In actual application, the inner side of the rim cannot be inspected. In addition, most equipment adopts a single-station fixed inspection mode, which requires manual replacement of the wheel hub. Each time a wheel hub is inspected, the assembly line needs to be interrupted for 5-8 seconds, which seriously restricts production capacity. Secondly, the self-rotation mechanism of patent CN215114607U cannot be synchronized with the assembly line, and manual intervention is still required for loading and unloading, which fails to achieve true unmanned operation. At this stage, a quality inspection device and method for the surface of automobile wheel hubs is needed. Summary of the Invention

[0005] In order to solve the problems of low detection efficiency and limited detection range in traditional detection devices, the present invention provides a quality detection device and method for the surface of an automobile wheel hub.

[0006] In a first aspect, the present invention provides a quality inspection device for the surface of an automobile wheel hub, which adopts the following technical solution:

[0007] A quality inspection device for the surface of an automobile hub, comprising:

[0008] A conveying unit, an outer surface sub-region detection unit, a flipping and clamping unit, and an inner surface detection unit, wherein the conveying unit is sequentially connected to the outer surface sub-region detection unit, the flipping and clamping unit, and the inner surface detection unit, and is used to sequentially convey the wheel hub to each detection area;

[0009] The outer surface area detection unit includes an aluminum profile bracket, a first linear guide and a third linear guide, the first linear guide and the third linear guide are both fixed to the upper surface of the aluminum profile bracket, the flip clamping unit includes a second linear guide, a third servo motor and a first slider, one end of the second linear guide is connected to the output shaft of the third servo motor, the first slider is slidably connected to the inner surface of the second linear guide, the inner surface detection unit includes a positioning assembly and a camera bracket, the positioning assembly and the camera bracket are both connected to the aluminum profile bracket by bolts.

[0010] Furthermore, the conveying unit includes a conveyor belt, a driving motor, a tensioning mechanism and a limiting baffle. The output shaft of the driving motor is connected to the conveyor belt, and the tensioning mechanism abuts against the conveyor belt to maintain the tension of the conveyor belt. The limiting baffle is arranged on both sides of the conveyor belt. The tensioning mechanism includes a tensioning wheel, an elastic telescopic rod and a fixed bracket. The tensioning wheel is installed on one end of the elastic telescopic rod through a bearing, and the other end of the elastic telescopic rod is fixed to the fixed bracket. The tensioning wheel presses against the outer edge of the conveyor belt.

[0011] Furthermore, the aluminum profile bracket is fixedly connected to the first linear guide and the third linear guide by T-slot bolts respectively. The first linear guide adopts a ball screw guide. One end of the first linear guide is connected to the second servo motor through an elastic coupling. The other end of the first linear guide is fixed to the aluminum profile bracket to form an x-axis motion reference. The third linear guide adopts a sliding guide. The third linear guide cooperates with the flange-type linear bearing of the second slider to provide auxiliary support for the x-axis motion. The first linear guide includes an X-axis linear guide, a Y-axis linear guide and a Z-axis linear guide.

[0012] Furthermore, an M6 threaded hole is provided on the top surface of the second slider, and the second slider is fixed to the first servo motor by a hexagon socket head screw and a spring washer. The output shaft of the first servo motor is connected to the screw inside the Y-axis linear guide through an elastic coupling, and the Z-axis linear guide is connected to the output shaft of the fourth servo motor through an elastic coupling. A linear module slide is also provided on one side of the second slider, and the second slider is connected to the mounting plate of the first industrial camera through the linear module slide.

[0013] Furthermore, the outer surface area detection unit also includes a limit sensor, which is fixed to the side of the aluminum profile bracket through an L-shaped bracket. One end of the L-shaped bracket is connected to the T-slot of the aluminum profile bracket with an M6 bolt, and the other end of the L-shaped bracket is provided with a long strip adjustment hole for adjusting the vertical height of the limit sensor up and down.

[0014] Furthermore, the flip clamping unit also includes a flip bracket, the top surface of the flip bracket is fixed to the second linear guide rail by a countersunk screw and a positioning pin, the output shaft of the third servo motor is connected to one end of the screw rod of the second linear guide rail via a high-rigidity coupling, and the other end of the screw rod of the second linear guide rail is installed with an angular contact ball bearing for support, the bottom surface of the first slider is provided with a T-slot, and the first slider is connected to the cylinder bracket through the T-slot.

[0015] Furthermore, the cylinder bracket is fixedly connected to the flip cylinder through an L-shaped connecting plate, and guide sleeves and guide rods are provided on both sides of the L-shaped connecting plate. The flip cylinder is connected to the linear cylinder through a fisheye joint and a spherical bearing, and the piston rod of the linear cylinder is connected to the wheel hub clamp through a floating joint. The wheel hub clamp is a two-finger parallel opening and closing pneumatic finger, and the wheel hub clamp has a built-in pressure sensor.

[0016] Furthermore, the inner surface detection unit also includes a camera bracket, a fifth servo motor, a sixth servo motor, a fourth industrial camera, a connecting rod mechanism and a reciprocating linear guide rail. The fifth servo motor drives the positioning assembly to clamp the wheel hub through the connecting rod mechanism and the reciprocating linear guide rail. The sixth servo motor is connected to the positioning assembly through a keyway to drive the wheel hub to rotate. The fourth industrial camera is fixed to the aluminum profile bracket through the camera bracket, and the mounting plate of the fourth industrial camera is equipped with a shock-absorbing rubber pad.

[0017] In a second aspect, a method for detecting the quality of a surface of an automobile wheel hub comprises:

[0018] The original image acquired by the industrial camera is evenly sliced ​​to filter out the sub-images containing defect features;

[0019] Feature extraction and detection based on sub-images containing defect features, including the use of a Darknet-53 convolutional neural network combined with a three-level grid structure, generating multiple groups of anchor boxes through K-means clustering. Finally, after upsampling and feature fusion, each grid cell predicts the defect bounding box coordinates, confidence level, and category label of multiple anchor boxes.

[0020] Duplicate detection frames are eliminated through non-maximum suppression, and the detection results are transmitted to the PLC system through the TCP / IP protocol and synchronously stored in the MES system to generate defect analysis data.

[0021] Furthermore, the Darknet-53 convolutional neural network includes multiple layers of convolutional layers, residual blocks and maximum pooling layers. It extracts image features through 1×1 and 3×3 convolutional layers and residual blocks, and performs maximum pooling operation on the feature map through the SPP spatial pyramid pooling module.

[0022] In summary, the present invention has the following beneficial technical effects:

[0023] 1. The present invention connects the outer surface area detection unit, the flip clamping unit and the inner surface detection unit in sequence through the transmission unit to realize the full-process automated detection of the wheel hub from the outer surface to the inner surface. The driving motor of the transmission unit cooperates with the conveyor belt, and the tensioning mechanism ensures the stable operation of the conveyor belt. The limit baffle is used to precisely connect with each detection unit to ensure the seamless flow of the wheel hub in each area, reduce manual intervention, and significantly improve detection efficiency.

[0024] 2. The present invention utilizes the flip bracket of the flip clamping unit to fix the linear guide rail through countersunk screws and locating pins, and cooperates with high-rigidity couplings and angular contact ball bearings to ensure stable and reliable lifting movement. The special connection method between the cylinder bracket and the flip cylinder and linear cylinder, such as the L-shaped connecting plate with the guide sleeve and guide rod, the application of fisheye joints and spherical bearings, enhances the stability of the mechanism during the flipping process, and can adapt to the clamping requirements of wheel hubs of different specifications. The wheel hub clamp has a built-in pressure sensor to monitor the clamping force in real time to prevent damage to the wheel hub due to over-tightening, thereby ensuring the safety and integrity of the wheel hub during the flipping and clamping process.

[0025] 3. The inner surface detection unit of the present invention is based on the aluminum profile bracket. It cooperates with the positioning component and the servo motor to achieve precise positioning and rotation of the wheel hub, so that all parts of the inner side of the rim and the inner surface of the spoke can be inspected. Combined with the automated image acquisition and detection process, it can cover areas that are difficult for traditional detection equipment to reach, and realize all-round and high-precision detection of the inner surface of the wheel hub.

[0026] 4. The detection method of the present invention achieves rapid and accurate defect identification through steps such as raw image preprocessing, feature extraction and detection, and result processing. The slicing and filtering mechanisms of image preprocessing reduce ineffective computation and improve processing speed. The Darknet-53 convolutional neural network, combined with a three-level grid structure and anchor frames generated by K-means clustering, can effectively detect defects of varying sizes. Non-maximum suppression processing eliminates duplicate detection frames, and detection results are transmitted in real time via the TCP / IP protocol, ensuring high detection accuracy and fast response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The figure is a schematic diagram of the overall structure of a quality inspection device for automobile wheel hub surface according to an embodiment of the present invention.

[0028] Figure 2 The present invention is a schematic structural diagram of a flip clamping unit in a quality inspection device for an automobile wheel hub surface according to an embodiment of the present invention.

[0029] Figure 3 Schematic diagram of the x / y / z three-axis motion system structure of an embodiment of the present invention.

[0030] Figure 4 It is a top view of a quality inspection device for an automobile wheel hub surface according to an embodiment of the present invention.

[0031] Figure 5 The present invention is a schematic structural diagram of an inner surface detection unit in a quality detection device for an automobile wheel hub surface according to an embodiment of the present invention.

[0032] Figure 6 The present invention is a schematic diagram of the structure of an outer surface sub-area detection unit in a quality detection device for an automobile wheel hub surface according to an embodiment of the present invention.

[0033] Figure 7 This is a processing flow chart of a quality inspection method for an automobile wheel hub surface according to embodiment 2 of the present invention.

[0034] Among them, 1. transmission unit; 2. wheel hub; 3. first industrial camera; 4. aluminum profile bracket; 5. first linear guide; 6. first servo motor; 7. limit sensor; 8. second servo motor; 9. third servo motor; 10. second linear guide; 11. support frame; 12. flip cylinder; 13. linear cylinder; 14. wheel hub clamp; 15. camera bracket; 16. positioning assembly; 17. mechanical limit block; 18. fourth servo motor; 19. flip bracket; 20. first slider; 21. cylinder bracket; 22. workbench; 23. coupling; 24. third linear guide; 25. second industrial camera; 26. third industrial camera; 27. fourth industrial camera; 28. connecting rod; 29. ​​reciprocating linear guide; 30. connecting rod mechanism; 31. fifth servo motor; 32. sixth servo motor; 33. second slider. DETAILED DESCRIPTION

[0035] The present invention will be further described in detail below with reference to the accompanying drawings.

[0036] Example 1

[0037] Reference Figure 1 , a quality inspection device for the surface of an automobile wheel hub of this embodiment includes:

[0038] The conveying unit 1, the outer surface sub-region detection unit, the flipping and clamping unit and the inner surface detection unit are sequentially connected to the outer surface sub-region detection unit, the flipping and clamping unit and the inner surface detection unit, and are used to sequentially convey the wheel hub 2 to each detection area;

[0039] The outer surface area detection unit includes an aluminum profile bracket 4, a first linear guide 5 and a third linear guide 24, the first linear guide 5 and the third linear guide 24 are both fixed to the upper surface of the aluminum profile bracket 4, the flip clamping unit includes a second linear guide 10, a third servo motor 9 and a first slider 20, one end of the second linear guide 10 is connected to the output shaft of the third servo motor 9, the first slider 20 is slidably connected to the inner surface of the second linear guide 10, the inner surface detection unit includes a positioning assembly 16 and a camera bracket 15, the positioning assembly 16 and the camera bracket 15 are both connected to the aluminum profile bracket 4 by bolts.

[0040] Specifically, a quality inspection device for the surface of an automobile wheel hub includes the following contents:

[0041] like Figure 1 、 Figure 4 As shown, the transmission unit consists of a conveyor belt, a driving motor, a tensioning mechanism and a limit baffle. The driving motor drives the conveyor belt to run at a constant speed through a belt transmission system. Specifically, the output shaft of the driving motor is connected to the driving pulley through a flat key, and the driving pulley is transmitted to the driven pulley through a V-belt, thereby driving the conveyor belt to be sleeved on the outer surface of the driving roller and the driven roller for circulation. The tensioning mechanism ensures that the conveyor belt always maintains appropriate tension. It consists of a tensioning wheel, an elastic telescopic rod and a fixed bracket. The fixed bracket is fixed to the side of the conveying bracket. One end of the elastic telescopic rod is hinged to the fixed bracket, and the other end is mounted on the tensioning wheel through a bearing to press against the outer edge of the conveyor belt. The built-in compression spring provides pressure to maintain the tension of the conveyor belt within a reasonable range to avoid slipping and affecting the conveying accuracy.

[0042] Limit baffles are installed on both sides of the conveyor belt. They are made of L-shaped aluminum alloy plates and are fixed to the brackets on both sides of the conveyor belt by bolts. The position is precisely aligned with the mechanical limit block 17 of the outer surface area detection unit. The width of the channel formed by the two is set according to the diameter of the hub 2. The inner side is at a certain vertical distance from the conveyor belt surface. The hub is positioned by mechanical limit and sensor detection. When the hub moves with the conveyor belt to the entrance of the outer surface area detection unit, it first contacts the limit baffle and is decelerated by the mechanical limit. At the same time, the photoelectric sensor of the mechanical limit block 17 of the outer surface area detection unit detects the hub 2 and sends a signal to the PLC. The PLC controls the drive motor to decelerate, and the hub continues to move to the limit baffle. The gap with the mechanical limit block 17 is completely blocked. After receiving the arrival signal, the PLC instructs the drive motor to cut off the power and the conveyor belt to stop running, so that the wheel hub stays accurately at the starting position of the detection, providing a stable benchmark for subsequent regional detection. When working, the drive motor is energized and drives the conveyor belt to run through the belt drive. The frequency converter adjusts the motor speed to adapt to the detection rhythm. The tensioning mechanism adaptively adjusts the tensioning force through the elastic telescopic rod. The wheel hub flows from the front process into the rear conveyor belt and is transported. When it approaches the outer surface regional detection unit, the dual positioning mechanism is activated, and the outer surface regional detection unit is triggered to work after precise positioning. The whole process is highly automated, ensuring efficient coordination of the transmission and detection of the wheel hub 2.

[0043] like Figure 3 、 Figure 6 As shown, the external surface detection unit utilizes an aluminum profile bracket 4 as a reference to construct an x / y / z three-axis motion system. A first linear guide 5 serves as the core transmission component for the x-axis. One end of its ball screw is connected to a second servo motor 8 via an elastic coupling. When the motor rotates, the screw converts rotational motion into linear motion, driving the second slider 33 laterally along the x-axis (parallel to the width of the conveyor belt). A third linear guide 24 mates with the flanged linear bearing of the second slider 33 to form a dual-guide support structure, eliminating lateral torque associated with single-guide motion and ensuring smooth and linear x-axis motion.

[0044] A first servo motor 6, fixed to the top surface of the second slider 33, drives the leadscrew of the y-axis linear guide via an elastic coupling. This allows the second slider 33 to move longitudinally along the y-axis (parallel to the conveyor belt's direction of travel), in addition to its x-axis movement. This expands the longitudinal coverage of the inspection area. A fourth servo motor 18, connected to the leadscrew of the z-axis linear guide via an elastic coupling, drives the mounting plate of the first industrial camera 3 up and down along the z-axis (perpendicular to the xy plane), adjusting the distance between the camera and the outer surface of the wheel rim to ensure clear lens focus. The linear module slide and anti-collision buffer ensure that the elastic deformation of the buffer absorbs impact at the extreme z-axis movement, preventing camera shake.

[0045] When the wheel hub enters the outer surface detection unit along the conveyor belt, a limit sensor 7 is fixed to the side of the aluminum profile bracket 4 via an L-shaped bracket. Its detection end is precisely aligned with the wheel hub position through a long adjustment hole. If the wheel hub blocks the sensor beam, the sensor sends an electrical signal to the control system, triggering the three-axis motion mechanism to start. After receiving the control signal, the second servo motor 8 drives the x-axis linear guide, moving the second slider 33 along the x-axis to the left edge of the wheel rim's outer surface, determining the starting position for lateral detection. During this process, an elastic coupling compensates for any coaxiality errors between the motor and the lead screw, ensuring positioning accuracy.

[0046] The first servo motor 6 drives the y-axis linear guide to move the second slider 33 from the top to the bottom of the outer surface of the rim along the y-axis direction. At the same time, the fourth servo motor 18 adjusts the z-axis height to keep the camera lens at the optimal shooting distance from the rim surface. The three-axis linkage divides the outer surface of the rim into four areas: upper left, upper right, lower left, and lower right. Each area obtains a high-definition image by locally magnifying the area. The first industrial camera 3 performs overlapping shooting of each sub-area (overlap rate ≥ 30%). For example, after the upper left area is shot, the x-axis moves 120 mm to the right to the upper right area, and the y-axis synchronously adjusts the shooting starting position to ensure seamless splicing of images in adjacent areas. The error compensation function of the elastic coupling in the three-axis transmission ensures the geometric consistency of the spliced ​​image.

[0047] The limit sensor 7 can be adjusted up or down 20mm through the long adjustment hole (30mm long) of the L-shaped bracket to adapt to the height difference of 14-22 inch wheels. When the wheel reaches the entrance of area A, the sensor triggers a signal to the PLC control system, and the PLC executes the preset program:

[0048] 1. Send a pulse signal to the second servo motor 8 to control the x-axis to position to the left edge of the rim;

[0049] 2. Start the first servo motor 6 and the fourth servo motor 18 to make the y-axis and z-axis work together, and the camera starts scanning along the divided area path;

[0050] 3. After each area is shot, the three axes will automatically move to the next area according to the preset coordinates until all four areas are shot.

[0051] This linkage mechanism ensures that the position deviation of the wheel hub on the conveyor belt is accurately compensated by the three-axis movement. For example, when the wheel hub deviates by 0.8mm laterally, the x-axis motor moves an additional 0.8mm, so that the camera is always aligned with the inspection area on the outer surface of the rim to avoid missed inspections.

[0052] like Figure 2As shown, the third servo motor 9 of the flip clamping unit is the core power source, and the rotational motion is converted into the vertical lifting motion of the first slider 20 through the second linear guide 10, and supported by the flip bracket 19. Specifically, the output shaft of the third servo motor 9 is connected to the screw of the second linear guide 10 through a high-rigidity coupling 23. When the motor rotates, the screw rotates synchronously, driving the first slider 20 to lift vertically along the guide rail. The lifting accuracy is guaranteed by the screw pitch and the servo motor encoder. The T-slot on the bottom surface of the first slider 20 is connected to the cylinder bracket 21 through a T-bolt, allowing the cylinder bracket to be adjusted horizontally by ±10mm to adapt to the center height difference of the wheel hub 2 of different specifications. A support frame 11 is provided at the top of the flip bracket 19 to support the screw bearing seat of the linear guide (10) to reduce the sagging during long-stroke movement.

[0053] The flip cylinder 12 and the linear cylinder 13 form an articulated structure through a fisheye joint and a spherical bearing. When the flip cylinder piston rod is extended or retracted, the ball of the fisheye joint rotates in the spherical bearing, eliminating the radial force caused by the eccentric installation of the cylinder. This design allows the thrust of the flip cylinder to be transmitted to the linear cylinder only along the axial direction, driving the wheel hub clamp 14 to complete a 180° flipping action. The piston rod of the linear cylinder 13 and the wheel hub clamp 14 are connected by a floating joint, ensuring that the clamp is evenly stressed when clamping the wheel hub, avoiding local stress damage to the hub surface.

[0054] When the wheel hub 2 is transported by conveyor unit 1 to area B, third servo motor 9 drives first slide 20, which lowers cylinder support 21 above the wheel hub. The piston rod of linear cylinder 13 extends, closing wheel hub clamp 14 via a floating joint. A built-in pressure sensor in the clamp monitors the clamping force in real time. When the pressure sensor feedback value reaches 30-50N, the PLC control system instructs the linear cylinder to stop, ensuring a stable clamping of the wheel hub without damage from overtightening.

[0055] After clamping is complete, the tilting cylinder 12 activates, extending its piston rod to push the linear cylinder 13 and the clamping jaws to rotate about the hinge point. Guide sleeves and guide rods 28 on either side of the L-shaped connecting plate provide precise guidance during this process, limiting the tilting cylinder's runout and keeping the tilting angle within 180°±1°. During the tilting process, the linkage between the fisheye joint and the spherical plain bearing ensures that the linear cylinder remains perpendicular to the wheel hub axis, preventing the hub from shaking during the tilting process.

[0056] After the flipping action is completed, the third servo motor 9 rotates in the opposite direction, driving the first slide 20 to its initial position. The piston rod of the linear cylinder 13 retracts, the hub clamp 14 releases, and the hub 2 drops onto the conveyor belt of the transfer unit 1. The conveyor belt transports the hub 2 to area C for inner surface inspection. Throughout the entire process, a magnetic ring sensor monitors the position of the cylinder piston rod in real time to ensure the precise execution of each action. For example, when the flipping cylinder is in place, the magnetic ring sensor sends a signal to the PLC, which triggers the next upward action to avoid misoperation.

[0057] The third servo motor 9 is connected to a PLC (Siemens S7-1200) via a servo driver. The PLC sends pulse commands via the Profinet protocol to control the motor's speed and stroke. The motor's built-in encoder provides real-time position feedback, forming a closed-loop control system. For example, when the PLC instructs the slider to descend 50mm, the encoder detects the actual displacement and feeds it back to the PLC. If there is any deviation, the PLC immediately adjusts the pulse output to ensure positioning accuracy.

[0058] The movements of the tilting cylinder 12 and the linear cylinder 13 are controlled by a PLC via solenoid valves (two-position, five-way valves). Magnetic ring sensors on the cylinder bodies generate electrical signals when the piston rods reach a specified position. For example, when the linear cylinder is clamped in place, the magnetic ring sensor signal triggers the PLC to activate the tilting cylinder. This sensor feedback mechanism ensures the precise sequence of cylinder movements and avoids interference.

[0059] The pressure sensor built into the wheel hub clamp 14 is connected to the analog input module of the PLC through a shielded cable, converting the clamping force signal into a digital value in real time. When the clamping force exceeds 50N, the PLC immediately issues an alarm and stops the cylinder action to prevent the wheel hub from being clamped; if the clamping force is lower than 30N, it indicates that the clamp is not clamped and refuses to perform the flipping action to ensure safety.

[0060] When the outer surface area detection unit completes the inspection, the conveying unit transports the wheel hub to the specified position of the flip clamping unit. The limit sensor 7 detects that the wheel hub is in place and triggers the flip clamping unit to operate. After the flip clamping unit is flipped, the conveying unit continues to transport the wheel hub to the inner surface detection unit, where the positioning component 16 and the camera bracket 15 perform inner surface inspection. During the entire process, the PLC control system communicates with the sensors and actuators in each area through the TCP / IP protocol to ensure synchronization of the assembly line beat.

[0061] like Figure 5As shown, the inner surface detection unit uses the fifth servo motor 31 and the sixth servo motor 32 as the core power sources to achieve precise positioning and rotation of the wheel hub 2. The fifth servo motor 31 drives the positioning assembly 16 to clamp the wheel hub through the connecting rod mechanism 30. The positioning assembly 16 includes four rollers. Its core function is to drive the four rollers to move synchronously through the connecting rod mechanism through the servo motor to clamp and position the wheel hub from all sides. The connecting rod mechanism 30 adopts a symmetrical four-bar structure. The output shaft of the servo motor 31 is connected to one end of the active connecting rod, and the other end of the active connecting rod is connected to the intermediate transmission connecting rod through a hinge. The intermediate transmission connecting rod is then hinged to the four connecting rods that drive the rollers to move, forming a closed transmission chain. When the servo motor 31 rotates, it drives the active connecting rod to perform circular motion. The active connecting rod pushes the intermediate transmission connecting rod through the hinge point, and then transmits the motion to the connecting rod connected to the roller, so that the four rollers simultaneously converge toward the center or open to the surroundings, thereby achieving the clamping and loosening of the wheel hub.

[0062] The output shaft of the fifth servo motor 31 drives the crank to rotate through a coupling, and the crank pushes the slider to translate in the reciprocating linear guide 29 through the connecting rod 28. The linear motion of the slider is converted into vertical pressure of the clamping arm of the positioning component 16 through the wedge block. The maximum clamping force is 120N, which ensures that the wheel hub is firmly fixed. The sixth servo motor 32 is connected to the rotating shaft of the positioning component 16 through a flat keyway, and the speed is reduced to 250r / min through a reducer (reduction ratio 1:6), driving the wheel hub to rotate at a stable speed, so that the inner side of the rim and the inner surface of the spoke are exposed to the detection field of view in turn.

[0063] The second industrial camera 25 is fixed at the bottom of the three-coordinate z-axis guide rail, with the lens axis forming a 45° angle with the inner surface of the rim. It can be adjusted by ±25mm along the z-axis through the L-shaped mounting plate to adapt to different hub heights. When the hub rotates, the second industrial camera 25 continuously shoots the inner surface of the rim, and the image coverage range is from the edge of the rim to the mounting surface, ensuring that defects such as cracks and pits are not missed. The third industrial camera 26 is mounted on an x / y / z three-axis motion platform (with the same structure as the outer surface area detection unit) and realizes three-dimensional motion under the drive of a servo motor. The platform synchronously controls the movement of the third industrial camera 26 according to the rotation angle of the hub, and performs detailed shooting of the inner surface of the spoke in different areas, especially for complex structural areas such as bolt holes and ribs, covering areas that are difficult to reach with traditional equipment through multi-directional perspectives.

[0064] The fourth industrial camera 27 is fixed to the side of the aluminum profile bracket 4 through the camera bracket 15, with the lens vertically aligned with the inner surface of the spoke. The shock-absorbing rubber pad between the camera mounting plate and the bracket ensures a clear image. When the wheel hub rotates, the fourth industrial camera 27 continuously captures a panoramic image of the inner surface of the spoke, which complements the regional image of the third industrial camera 26 to improve the comprehensiveness of the detection.

[0065] The fifth servo motor 31 and the sixth servo motor 32 communicate with the PLC (Siemens S7-1200) via the Profinet protocol. The PLC sends pulse commands to control the motor movement and receives encoder feedback to form a closed loop. For example, every time the sixth servo motor 32 rotates 1°, the encoder sends a pulse to the PLC, which then synchronously triggers the camera to capture the image, ensuring that the image acquisition accurately matches the rotational position of the wheel hub and avoiding misalignment. The second industrial camera 25, the third industrial camera 26, and the fourth industrial camera 27 are connected to the PLC via the GPIO interface, and the PLC uniformly outputs a trigger signal. When the wheel hub rotates to a preset angle (such as 0°, 90°, 180°, and 270°), the PLC simultaneously sends a trigger signal to the three cameras to achieve multi-view image synchronization. The incremental encoder is installed on the rotating shaft of the positioning component 16, generating a pulse every 1° as a reference signal for camera triggering, ensuring a one-to-one correspondence between the image and the wheel hub position.

[0066] The three-axis motion platform of the third industrial camera 26 is linked to the rotational motion of the wheel hub. When the wheel hub rotates, the platform controls the movement of the third industrial camera 26 according to a preset trajectory, realizing a compound motion of "rotation + translation", ensuring that the camera lens is always vertically aligned with the area to be inspected on the inner surface of the spoke. For example, when inspecting the spoke ribs, the third industrial camera 26 translates along the x-axis while the wheel hub rotates synchronously, so that the rib surface is always in the center of the camera's field of view, improving the defect recognition rate. A 3mm thick shock-absorbing rubber pad is installed between the mounting plate of the fourth industrial camera 27 and the camera bracket 15 to effectively attenuate vibrations from the conveyor belt and the movement of the robotic arm. The dovetail structure at the bottom of the camera bracket cooperates with the T-slot of the aluminum profile bracket and is fixed with two sets of M8 bolts to ensure the stable position of the camera and avoid image blur caused by vibration.

[0067] After the wheel hub is transported by the flipping and clamping mechanism to the upper end of the worktable 22 of the inner surface inspection unit, the conveyor unit 1 pauses, the positioning assembly 16, driven by the fifth servo motor 31, clamps the wheel hub, and the sixth servo motor 32 starts rotating. The flipping and clamping unit and the inner surface inspection unit synchronize their operations via PLC communication: After the flipping and clamping unit completes flipping and transports the wheel hub to the designated position in the inner surface inspection unit, it sends a positioning signal to the inner surface inspection unit PLC. Upon receiving the signal, the inner surface inspection unit initiates the positioning and inspection process, ensuring synchronized assembly line rhythm and achieving fully automated continuous operation. After inspection is complete, the positioning assembly releases the wheel hub, and the conveyor unit resumes operation, transporting the wheel hub to the next process step, forming a closed-loop "positioning-inspection-transmission" process.

[0068] Example 2

[0069] This embodiment differs from the first embodiment in that this embodiment provides a method for inspecting the surface quality of an automobile wheel hub, comprising:

[0070] like Figure 7 As shown, S1, industrial cameras 25 (inside of the rim), and 27 (inner surface of the spoke) synchronously capture raw images from different angles. The resolution is set to 1280×720 or higher depending on the camera model to ensure clear defect details. The image acquisition trigger is synchronized with the rotation of the hub: the encoder signal on the rotation axis of the positioning component 16 (a pulse is sent every 1°) is used to control the camera shooting by the PLC to avoid motion blur.

[0071] The original image is divided into 16 sub-images according to the 4×4 grid rule. Based on the grayscale threshold and Canny edge detection algorithm, sub-images with edge mutations and texture abnormalities are automatically identified, and blank areas without defects are eliminated. The gradient amplitude of Canny edge detection is calculated as follows:

[0072]

[0073] Among them, G x and G y are the gradient operators in the x-direction and y-direction, respectively. I represents the preprocessed grayscale image matrix. The filtered sub-images are uniformly scaled to 416 × 416 pixels, and the letterbox filling method (adding grayscale borders) is used to maintain the original aspect ratio to avoid detection bias caused by target deformation.

[0074] S2, extracting features based on the preprocessed image data;

[0075] Feature extraction and defect detection are the core links of the entire algorithm. Their accuracy directly determines the precision of defect recognition. The YOLOv3-SPP algorithm adopts a three-level grid structure of 52×52, 26×26, and 13×13, corresponding to the detection of small (≤2 mm), medium (2-5 mm), and large (>5 mm) defects, respectively. To make the detection frame more consistent with the actual defect size, the bounding box size of historical defect samples is clustered based on the K-means clustering algorithm to generate 9 groups of anchor frames. Among them, small defects correspond to anchor frames such as (10×13), (16×30), and (33×23); medium defects correspond to anchor frames such as (30×61), (62×45), and (59×119); and large defects correspond to anchor frames such as (116×90), (156×198), and (373×326). The clustering process uses 1-IoU as the distance metric. Through multiple iterations, the center position of the anchor box is optimized to significantly improve the matching degree between the anchor box and the real defect, thereby effectively improving the defect recall rate.

[0076] Feature extraction is implemented using the Darknet-53 convolutional neural network, which consists of 53 convolutional layers. This network extracts image features layer by layer through a combination of 1×1 and 3×3 convolutional layers and residual blocks. The first 20 shallow layers primarily capture basic features such as edges and textures, while the next 33 deep layers focus on extracting advanced features such as shape and semantics. Ultimately, feature maps of three different scales are formed: 52×52, 26×26, and 13×13. To enhance the algorithm's robustness to multi-scale defects, the network incorporates an SPP (Spatial Pyramid Pooling) module. This module performs max pooling operations of 1×1, 5×5, 9×9, and 13×13 on the 13×13 feature map, respectively. This fusion of features at different scales enables the network to detect both subtle and large defects.

[0077] In the feature fusion and prediction stage, the deep feature map is fused with the shallow feature map through a 2x interpolation upsampling operation. For example, a 13×13 feature map is upsampled to 26×26 and then spliced ​​with a 26×26 feature map. This fusion method not only retains the semantic information of the deep network, but also combines the detailed features of the shallow network. Each grid unit predicts three anchor boxes. The output parameters include the bounding box coordinates (x, y, w, h), defect confidence (0-1) and category labels (such as cracks, pits, scratches, etc.). The coordinate prediction error is controlled within 0.5 pixels to ensure the accuracy of defect positioning. Finally, through non-maximum suppression (NMS) processing, the IoU threshold is set to 0.45, and repeated detection boxes of the same defect are filtered out. Only the detection results with the highest confidence are retained, reducing the false detection rate to below 1.5%.

[0078] S3, test result application and data closure stage;

[0079] The inspection results are transmitted to the production line's PLC system in real time via the TCP / IP protocol to ensure real-time response. For unqualified wheels, the PLC calculates the robot's grasping path based on the defect coordinates and maps the image pixel coordinates (u, v) to the robot's coordinate system (X, Y, Z):

[0080]

[0081] Among them, (u0, v0) represents the coordinates of the origin of the image coordinate system, that is, the coordinates of the pixel point at the center of the image, k x 、k y and k z They represent the physical size conversion coefficients of pixel coordinates in the x, y, and z directions respectively. X0 and Y0 represent the x-axis and y-axis coordinates corresponding to the image center in the robot coordinate system respectively. Z0 represents the reference height of the robot's z-axis. f represents the camera parameters, which drives the robot to accurately grasp it and move it to the rework line. Qualified products are then transported to the next process by a conveyor belt controlled by the PLC.

[0082] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A quality inspection device for the surface of an automobile hub, characterized in that: include: A conveying unit (1), an outer surface sub-region detection unit, a flipping and clamping unit, and an inner surface detection unit, wherein the conveying unit (1) is sequentially connected to the outer surface sub-region detection unit, the flipping and clamping unit, and the inner surface detection unit, and is used to sequentially convey the wheel hub (2) to each detection area; The outer surface sub-area detection unit includes an aluminum profile bracket (4), a first linear guide rail (5) and a third linear guide rail (24), wherein the first linear guide rail (5) and the third linear guide rail (24) are both fixed to the upper surface of the aluminum profile bracket (4), the flip clamping unit includes a second linear guide rail (10), a third servo motor (9) and a first slider (20), one end of the second linear guide rail (10) is connected to the output shaft of the third servo motor (9), and the first slider (20) is slidably connected to the inner surface of the second linear guide rail (10), and the inner surface detection unit includes a positioning assembly (16) and a camera bracket (15), wherein the positioning assembly (16) and the camera bracket (15) are both connected to the aluminum profile bracket (4) by bolts.

2. The quality inspection device for automobile wheel hub surface according to claim 1, characterized in that: The conveying unit (1) comprises a conveyor belt, a driving motor, a tensioning mechanism and a limiting baffle, wherein the output shaft of the driving motor is connected to the conveyor belt, the tensioning mechanism abuts against the conveyor belt and is used to maintain the tension of the conveyor belt, the limiting baffle is arranged on both sides of the conveyor belt, the tensioning mechanism comprises a tensioning wheel, an elastic telescopic rod and a fixed bracket, the tensioning wheel is installed on one end of the elastic telescopic rod through a bearing, the other end of the elastic telescopic rod is fixed to the fixed bracket, and the tensioning wheel abuts against the outer edge of the conveyor belt.

3. The quality inspection device for automobile wheel hub surface according to claim 1, characterized in that: The aluminum profile bracket (4) is fixedly connected to the first linear guide (5) and the third linear guide (24) respectively through T-slot bolts, the first linear guide (5) adopts a ball screw guide, one end of the first linear guide (5) is connected to the second servo motor (8) through an elastic coupling, and the other end of the first linear guide (5) is fixed to the aluminum profile bracket (4) to form an x-axis motion reference, the third linear guide (24) adopts a sliding guide, the third linear guide (24) cooperates with the flange-type linear bearing of the second slider (33) to provide auxiliary support for the x-axis motion, and the first linear guide (5) includes an X-axis linear guide, a Y-axis linear guide and a Z-axis linear guide.

4. The quality inspection device for automobile wheel hub surface according to claim 3, characterized in that: An M6 threaded hole is provided on the top surface of the second slider (33), and the second slider (33) is fixedly installed with the first servo motor (6) through a hexagon socket head screw and a spring washer. The output shaft of the first servo motor (6) is connected to the screw rod inside the Y-axis linear guide through an elastic coupling, and the Z-axis linear guide is connected to the output shaft of the fourth servo motor (18) through an elastic coupling. A linear module slide is also provided on one side of the second slider (33), and the second slider (33) is connected to the mounting plate of the first industrial camera (3) through the linear module slide.

5. The quality inspection device for automobile wheel hub surface according to claim 1, characterized in that: The outer surface sub-area detection unit further includes a limit sensor (7), which is fixed to the side of the aluminum profile bracket (4) via an L-shaped bracket, one end of the L-shaped bracket being connected to the T-slot of the aluminum profile bracket (4) using an M6 bolt, and the other end of the L-shaped bracket being provided with a long strip adjustment hole for adjusting the vertical height of the limit sensor (7) up and down.

6. The quality inspection device for automobile wheel hub surface according to claim 1, characterized in that: The flip clamping unit also includes a flip bracket (19), the top surface of the flip bracket (19) is fixed to the second linear guide (10) by a countersunk screw and a positioning pin, the output shaft of the third servo motor (9) is connected to one end of the screw rod of the second linear guide (10) via a high-rigidity coupling, the other end of the screw rod of the second linear guide (10) is installed with an angular contact ball bearing for support, the bottom surface of the first slider (20) is provided with a T-slot, and the first slider (20) is connected to the cylinder bracket (21) through the T-slot.

7. The quality inspection device for automobile wheel hub surface according to claim 6, characterized in that: The cylinder bracket (21) is fixedly connected to the flip cylinder (12) through an L-shaped connecting plate, and guide sleeves and guide rods are provided on both sides of the L-shaped connecting plate. The flip cylinder (12) is connected to the linear cylinder (13) through a fisheye joint and a joint bearing. The piston rod of the linear cylinder (13) is connected to the hub clamp (14) through a floating joint. The hub clamp (14) is a two-finger parallel opening and closing pneumatic finger, and the hub clamp (14) has a built-in pressure sensor.

8. The quality inspection device for automobile wheel hub surface according to claim 1, characterized in that: The inner surface detection unit also includes a camera bracket (15), a fifth servo motor (31), a sixth servo motor (32), a fourth industrial camera (27), a connecting rod mechanism (30) and a reciprocating linear guide rail (29), wherein the fifth servo motor (31) drives the positioning assembly (16) to clamp the wheel hub (2) through the connecting rod mechanism (30) and the reciprocating linear guide rail (29), and the sixth servo motor (32) is connected to the positioning assembly (16) through a keyway to drive the wheel hub (2) to rotate, and the fourth industrial camera (27) is fixed to the aluminum profile bracket (4) through the camera bracket (15), and the mounting plate of the fourth industrial camera (27) is equipped with a shock-absorbing rubber pad.

9. A method for inspecting the surface quality of an automobile wheel hub, based on the device according to claim 1, characterized in that: include: The original image acquired by the industrial camera is evenly sliced ​​to filter out the sub-images containing defect features; Feature extraction and detection based on sub-images containing defect features, including the use of a Darknet-53 convolutional neural network combined with a three-level grid structure, generating multiple groups of anchor boxes through K-means clustering. Finally, after upsampling and feature fusion, each grid cell predicts the defect bounding box coordinates, confidence level, and category label of multiple anchor boxes. Duplicate detection frames are eliminated through non-maximum suppression, and the detection results are transmitted to the PLC system through the TCP / IP protocol and synchronously stored in the MES system to generate defect analysis data.

10. A method for inspecting the surface quality of an automobile hub according to claim 9, characterized in that: The Darknet-53 convolutional neural network includes multiple convolutional layers, residual blocks, and maximum pooling layers. It extracts image features through 1×1 and 3×3 convolutional layers and residual blocks, and performs maximum pooling operations on feature maps through the SPP spatial pyramid pooling module.

Images